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 void dir_init(struct dir_struct
*dir
)
59 memset(dir
, 0, sizeof(*dir
));
62 struct dirent
*readdir_skip_dot_and_dotdot(DIR *dirp
)
66 while ((e
= readdir(dirp
)) != NULL
) {
67 if (!is_dot_or_dotdot(e
->d_name
))
73 int count_slashes(const char *s
)
82 int fspathcmp(const char *a
, const char *b
)
84 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
87 int fspatheq(const char *a
, const char *b
)
89 return !fspathcmp(a
, b
);
92 int fspathncmp(const char *a
, const char *b
, size_t count
)
94 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
97 unsigned int fspathhash(const char *str
)
99 return ignore_case
? strihash(str
) : strhash(str
);
102 int git_fnmatch(const struct pathspec_item
*item
,
103 const char *pattern
, const char *string
,
107 if (ps_strncmp(item
, pattern
, string
, prefix
))
112 if (item
->flags
& PATHSPEC_ONESTAR
) {
113 int pattern_len
= strlen(++pattern
);
114 int string_len
= strlen(string
);
115 return string_len
< pattern_len
||
116 ps_strcmp(item
, pattern
,
117 string
+ string_len
- pattern_len
);
119 if (item
->magic
& PATHSPEC_GLOB
)
120 return wildmatch(pattern
, string
,
122 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0));
124 /* wildmatch has not learned no FNM_PATHNAME mode yet */
125 return wildmatch(pattern
, string
,
126 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0);
129 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
130 const char *string
, int stringlen
,
134 struct strbuf pat_buf
= STRBUF_INIT
;
135 struct strbuf str_buf
= STRBUF_INIT
;
136 const char *use_pat
= pattern
;
137 const char *use_str
= string
;
139 if (pattern
[patternlen
]) {
140 strbuf_add(&pat_buf
, pattern
, patternlen
);
141 use_pat
= pat_buf
.buf
;
143 if (string
[stringlen
]) {
144 strbuf_add(&str_buf
, string
, stringlen
);
145 use_str
= str_buf
.buf
;
149 flags
|= WM_CASEFOLD
;
150 match_status
= wildmatch(use_pat
, use_str
, flags
);
152 strbuf_release(&pat_buf
);
153 strbuf_release(&str_buf
);
158 static size_t common_prefix_len(const struct pathspec
*pathspec
)
164 * ":(icase)path" is treated as a pathspec full of
165 * wildcard. In other words, only prefix is considered common
166 * prefix. If the pathspec is abc/foo abc/bar, running in
167 * subdir xyz, the common prefix is still xyz, not xyz/abc as
170 GUARD_PATHSPEC(pathspec
,
179 for (n
= 0; n
< pathspec
->nr
; n
++) {
180 size_t i
= 0, len
= 0, item_len
;
181 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
183 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
184 item_len
= pathspec
->items
[n
].prefix
;
186 item_len
= pathspec
->items
[n
].nowildcard_len
;
187 while (i
< item_len
&& (n
== 0 || i
< max
)) {
188 char c
= pathspec
->items
[n
].match
[i
];
189 if (c
!= pathspec
->items
[0].match
[i
])
195 if (n
== 0 || len
< max
) {
205 * Returns a copy of the longest leading path common among all
208 char *common_prefix(const struct pathspec
*pathspec
)
210 unsigned long len
= common_prefix_len(pathspec
);
212 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
215 int fill_directory(struct dir_struct
*dir
,
216 struct index_state
*istate
,
217 const struct pathspec
*pathspec
)
222 unsigned exclusive_flags
= DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
;
223 if ((dir
->flags
& exclusive_flags
) == exclusive_flags
)
224 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
227 * Calculate common prefix for the pathspec, and
228 * use that to optimize the directory walk
230 prefix_len
= common_prefix_len(pathspec
);
231 prefix
= prefix_len
? pathspec
->items
[0].match
: "";
233 /* Read the directory and prune it */
234 read_directory(dir
, istate
, prefix
, prefix_len
, pathspec
);
239 int within_depth(const char *name
, int namelen
,
240 int depth
, int max_depth
)
242 const char *cp
= name
, *cpe
= name
+ namelen
;
248 if (depth
> max_depth
)
255 * Read the contents of the blob with the given OID into a buffer.
256 * Append a trailing LF to the end if the last line doesn't have one.
259 * -1 when the OID is invalid or unknown or does not refer to a blob.
260 * 0 when the blob is empty.
261 * 1 along with { data, size } of the (possibly augmented) buffer
264 * Optionally updates the given oid_stat with the given OID (when valid).
266 static int do_read_blob(const struct object_id
*oid
, struct oid_stat
*oid_stat
,
267 size_t *size_out
, char **data_out
)
269 enum object_type type
;
276 data
= read_object_file(oid
, &type
, &sz
);
277 if (!data
|| type
!= OBJ_BLOB
) {
283 memset(&oid_stat
->stat
, 0, sizeof(oid_stat
->stat
));
284 oidcpy(&oid_stat
->oid
, oid
);
292 if (data
[sz
- 1] != '\n') {
293 data
= xrealloc(data
, st_add(sz
, 1));
297 *size_out
= xsize_t(sz
);
303 #define DO_MATCH_EXCLUDE (1<<0)
304 #define DO_MATCH_DIRECTORY (1<<1)
305 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
308 * Does the given pathspec match the given name? A match is found if
310 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
311 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
312 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
313 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
315 * Return value tells which case it was (1-4), or 0 when there is no match.
317 * It may be instructive to look at a small table of concrete examples
318 * to understand the differences between 1, 2, and 4:
321 * | a/b | a/b/ | a/b/c
322 * ------+-----------+-----------+------------
323 * a/b | EXACT | EXACT[1] | LEADING[2]
324 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
325 * a/b/c | RECURSIVE | RECURSIVE | EXACT
327 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
328 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
330 static int match_pathspec_item(struct index_state
*istate
,
331 const struct pathspec_item
*item
, int prefix
,
332 const char *name
, int namelen
, unsigned flags
)
334 /* name/namelen has prefix cut off by caller */
335 const char *match
= item
->match
+ prefix
;
336 int matchlen
= item
->len
- prefix
;
339 * The normal call pattern is:
340 * 1. prefix = common_prefix_len(ps);
341 * 2. prune something, or fill_directory
342 * 3. match_pathspec()
344 * 'prefix' at #1 may be shorter than the command's prefix and
345 * it's ok for #2 to match extra files. Those extras will be
348 * Suppose the pathspec is 'foo' and '../bar' running from
349 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
350 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
351 * user does not want XYZ/foo, only the "foo" part should be
352 * case-insensitive. We need to filter out XYZ/foo here. In
353 * other words, we do not trust the caller on comparing the
354 * prefix part when :(icase) is involved. We do exact
355 * comparison ourselves.
357 * Normally the caller (common_prefix_len() in fact) does
358 * _exact_ matching on name[-prefix+1..-1] and we do not need
359 * to check that part. Be defensive and check it anyway, in
360 * case common_prefix_len is changed, or a new caller is
361 * introduced that does not use common_prefix_len.
363 * If the penalty turns out too high when prefix is really
364 * long, maybe change it to
365 * strncmp(match, name, item->prefix - prefix)
367 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
368 strncmp(item
->match
, name
- prefix
, item
->prefix
))
371 if (item
->attr_match_nr
&&
372 !match_pathspec_attrs(istate
, name
, namelen
, item
))
375 /* If the match was just the prefix, we matched */
377 return MATCHED_RECURSIVELY
;
379 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
380 if (matchlen
== namelen
)
381 return MATCHED_EXACTLY
;
383 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
384 return MATCHED_RECURSIVELY
;
385 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
386 match
[matchlen
- 1] == '/' &&
387 namelen
== matchlen
- 1 &&
388 !ps_strncmp(item
, match
, name
, namelen
))
389 return MATCHED_EXACTLY
;
391 if (item
->nowildcard_len
< item
->len
&&
392 !git_fnmatch(item
, match
, name
,
393 item
->nowildcard_len
- prefix
))
394 return MATCHED_FNMATCH
;
396 /* Perform checks to see if "name" is a leading string of the pathspec */
397 if ( (flags
& DO_MATCH_LEADING_PATHSPEC
) &&
398 !(flags
& DO_MATCH_EXCLUDE
)) {
399 /* name is a literal prefix of the pathspec */
400 int offset
= name
[namelen
-1] == '/' ? 1 : 0;
401 if ((namelen
< matchlen
) &&
402 (match
[namelen
-offset
] == '/') &&
403 !ps_strncmp(item
, match
, name
, namelen
))
404 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
406 /* name doesn't match up to the first wild character */
407 if (item
->nowildcard_len
< item
->len
&&
408 ps_strncmp(item
, match
, name
,
409 item
->nowildcard_len
- prefix
))
413 * name has no wildcard, and it didn't match as a leading
414 * pathspec so return.
416 if (item
->nowildcard_len
== item
->len
)
420 * Here is where we would perform a wildmatch to check if
421 * "name" can be matched as a directory (or a prefix) against
422 * the pathspec. Since wildmatch doesn't have this capability
423 * at the present we have to punt and say that it is a match,
424 * potentially returning a false positive
425 * The submodules themselves will be able to perform more
426 * accurate matching to determine if the pathspec matches.
428 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
435 * do_match_pathspec() is meant to ONLY be called by
436 * match_pathspec_with_flags(); calling it directly risks pathspecs
437 * like ':!unwanted_path' being ignored.
439 * Given a name and a list of pathspecs, returns the nature of the
440 * closest (i.e. most specific) match of the name to any of the
443 * The caller typically calls this multiple times with the same
444 * pathspec and seen[] array but with different name/namelen
445 * (e.g. entries from the index) and is interested in seeing if and
446 * how each pathspec matches all the names it calls this function
447 * with. A mark is left in the seen[] array for each pathspec element
448 * indicating the closest type of match that element achieved, so if
449 * seen[n] remains zero after multiple invocations, that means the nth
450 * pathspec did not match any names, which could indicate that the
451 * user mistyped the nth pathspec.
453 static int do_match_pathspec(struct index_state
*istate
,
454 const struct pathspec
*ps
,
455 const char *name
, int namelen
,
456 int prefix
, char *seen
,
459 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
471 if (!ps
->recursive
||
472 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
474 return MATCHED_RECURSIVELY
;
476 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
477 return MATCHED_EXACTLY
;
485 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
488 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
489 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
492 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
495 * Make exclude patterns optional and never report
496 * "pathspec ':(exclude)foo' matches no files"
498 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
499 seen
[i
] = MATCHED_FNMATCH
;
500 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
503 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
504 ps
->max_depth
!= -1 &&
505 how
&& how
!= MATCHED_FNMATCH
) {
506 int len
= ps
->items
[i
].len
;
507 if (name
[len
] == '/')
509 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
510 how
= MATCHED_EXACTLY
;
517 if (seen
&& seen
[i
] < how
)
524 static int match_pathspec_with_flags(struct index_state
*istate
,
525 const struct pathspec
*ps
,
526 const char *name
, int namelen
,
527 int prefix
, char *seen
, unsigned flags
)
529 int positive
, negative
;
530 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
531 prefix
, seen
, flags
);
532 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
534 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
536 flags
| DO_MATCH_EXCLUDE
);
537 return negative
? 0 : positive
;
540 int match_pathspec(struct index_state
*istate
,
541 const struct pathspec
*ps
,
542 const char *name
, int namelen
,
543 int prefix
, char *seen
, int is_dir
)
545 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
546 return match_pathspec_with_flags(istate
, ps
, name
, namelen
,
547 prefix
, seen
, flags
);
551 * Check if a submodule is a superset of the pathspec
553 int submodule_path_match(struct index_state
*istate
,
554 const struct pathspec
*ps
,
555 const char *submodule_name
,
558 int matched
= match_pathspec_with_flags(istate
, ps
, submodule_name
,
559 strlen(submodule_name
),
562 DO_MATCH_LEADING_PATHSPEC
);
566 int report_path_error(const char *ps_matched
,
567 const struct pathspec
*pathspec
)
570 * Make sure all pathspec matched; otherwise it is an error.
573 for (num
= 0; num
< pathspec
->nr
; num
++) {
574 int other
, found_dup
;
579 * The caller might have fed identical pathspec
580 * twice. Do not barf on such a mistake.
581 * FIXME: parse_pathspec should have eliminated
582 * duplicate pathspec.
584 for (found_dup
= other
= 0;
585 !found_dup
&& other
< pathspec
->nr
;
587 if (other
== num
|| !ps_matched
[other
])
589 if (!strcmp(pathspec
->items
[other
].original
,
590 pathspec
->items
[num
].original
))
592 * Ok, we have a match already.
599 error(_("pathspec '%s' did not match any file(s) known to git"),
600 pathspec
->items
[num
].original
);
607 * Return the length of the "simple" part of a path match limiter.
609 int simple_length(const char *match
)
614 unsigned char c
= *match
++;
616 if (c
== '\0' || is_glob_special(c
))
621 int no_wildcard(const char *string
)
623 return string
[simple_length(string
)] == '\0';
626 void parse_path_pattern(const char **pattern
,
631 const char *p
= *pattern
;
636 *flags
|= PATTERN_FLAG_NEGATIVE
;
640 if (len
&& p
[len
- 1] == '/') {
642 *flags
|= PATTERN_FLAG_MUSTBEDIR
;
644 for (i
= 0; i
< len
; i
++) {
649 *flags
|= PATTERN_FLAG_NODIR
;
650 *nowildcardlen
= simple_length(p
);
652 * we should have excluded the trailing slash from 'p' too,
653 * but that's one more allocation. Instead just make sure
654 * nowildcardlen does not exceed real patternlen
656 if (*nowildcardlen
> len
)
657 *nowildcardlen
= len
;
658 if (*p
== '*' && no_wildcard(p
+ 1))
659 *flags
|= PATTERN_FLAG_ENDSWITH
;
664 int pl_hashmap_cmp(const void *unused_cmp_data
,
665 const struct hashmap_entry
*a
,
666 const struct hashmap_entry
*b
,
669 const struct pattern_entry
*ee1
=
670 container_of(a
, struct pattern_entry
, ent
);
671 const struct pattern_entry
*ee2
=
672 container_of(b
, struct pattern_entry
, ent
);
674 size_t min_len
= ee1
->patternlen
<= ee2
->patternlen
679 return strncasecmp(ee1
->pattern
, ee2
->pattern
, min_len
);
680 return strncmp(ee1
->pattern
, ee2
->pattern
, min_len
);
683 static char *dup_and_filter_pattern(const char *pattern
)
687 char *result
= xstrdup(pattern
);
693 /* skip escape characters (once) */
713 static void add_pattern_to_hashsets(struct pattern_list
*pl
, struct path_pattern
*given
)
715 struct pattern_entry
*translated
;
718 const char *prev
, *cur
, *next
;
720 if (!pl
->use_cone_patterns
)
723 if (given
->flags
& PATTERN_FLAG_NEGATIVE
&&
724 given
->flags
& PATTERN_FLAG_MUSTBEDIR
&&
725 !strcmp(given
->pattern
, "/*")) {
730 if (!given
->flags
&& !strcmp(given
->pattern
, "/*")) {
735 if (given
->patternlen
< 2 ||
736 *given
->pattern
== '*' ||
737 strstr(given
->pattern
, "**")) {
738 /* Not a cone pattern. */
739 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
743 prev
= given
->pattern
;
744 cur
= given
->pattern
+ 1;
745 next
= given
->pattern
+ 2;
748 /* Watch for glob characters '*', '\', '[', '?' */
749 if (!is_glob_special(*cur
))
752 /* But only if *prev != '\\' */
756 /* But allow the initial '\' */
758 is_glob_special(*next
))
761 /* But a trailing '/' then '*' is fine */
767 /* Not a cone pattern. */
768 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
777 if (given
->patternlen
> 2 &&
778 !strcmp(given
->pattern
+ given
->patternlen
- 2, "/*")) {
779 if (!(given
->flags
& PATTERN_FLAG_NEGATIVE
)) {
780 /* Not a cone pattern. */
781 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
785 truncated
= dup_and_filter_pattern(given
->pattern
);
787 translated
= xmalloc(sizeof(struct pattern_entry
));
788 translated
->pattern
= truncated
;
789 translated
->patternlen
= given
->patternlen
- 2;
790 hashmap_entry_init(&translated
->ent
,
792 strihash(translated
->pattern
) :
793 strhash(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
,
823 strihash(translated
->pattern
) :
824 strhash(translated
->pattern
));
826 hashmap_add(&pl
->recursive_hashmap
, &translated
->ent
);
828 if (hashmap_get_entry(&pl
->parent_hashmap
, translated
, ent
, NULL
)) {
829 /* we already included this at the parent level */
830 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
832 hashmap_remove(&pl
->parent_hashmap
, &translated
->ent
, &data
);
840 warning(_("disabling cone pattern matching"));
841 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
842 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
843 pl
->use_cone_patterns
= 0;
846 static int hashmap_contains_path(struct hashmap
*map
,
847 struct strbuf
*pattern
)
849 struct pattern_entry p
;
851 /* Check straight mapping */
852 p
.pattern
= pattern
->buf
;
853 p
.patternlen
= pattern
->len
;
854 hashmap_entry_init(&p
.ent
,
856 strihash(p
.pattern
) :
858 return !!hashmap_get_entry(map
, &p
, ent
, NULL
);
861 int hashmap_contains_parent(struct hashmap
*map
,
863 struct strbuf
*buffer
)
867 strbuf_setlen(buffer
, 0);
870 strbuf_addch(buffer
, '/');
872 strbuf_addstr(buffer
, path
);
874 slash_pos
= strrchr(buffer
->buf
, '/');
876 while (slash_pos
> buffer
->buf
) {
877 strbuf_setlen(buffer
, slash_pos
- buffer
->buf
);
879 if (hashmap_contains_path(map
, buffer
))
882 slash_pos
= strrchr(buffer
->buf
, '/');
888 void add_pattern(const char *string
, const char *base
,
889 int baselen
, struct pattern_list
*pl
, int srcpos
)
891 struct path_pattern
*pattern
;
896 parse_path_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
897 if (flags
& PATTERN_FLAG_MUSTBEDIR
) {
898 FLEXPTR_ALLOC_MEM(pattern
, pattern
, string
, patternlen
);
900 pattern
= xmalloc(sizeof(*pattern
));
901 pattern
->pattern
= string
;
903 pattern
->patternlen
= patternlen
;
904 pattern
->nowildcardlen
= nowildcardlen
;
905 pattern
->base
= base
;
906 pattern
->baselen
= baselen
;
907 pattern
->flags
= flags
;
908 pattern
->srcpos
= srcpos
;
909 ALLOC_GROW(pl
->patterns
, pl
->nr
+ 1, pl
->alloc
);
910 pl
->patterns
[pl
->nr
++] = pattern
;
913 add_pattern_to_hashsets(pl
, pattern
);
916 static int read_skip_worktree_file_from_index(struct index_state
*istate
,
918 size_t *size_out
, char **data_out
,
919 struct oid_stat
*oid_stat
)
924 pos
= index_name_pos(istate
, path
, len
);
927 if (!ce_skip_worktree(istate
->cache
[pos
]))
930 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
934 * Frees memory within pl which was allocated for exclude patterns and
935 * the file buffer. Does not free pl itself.
937 void clear_pattern_list(struct pattern_list
*pl
)
941 for (i
= 0; i
< pl
->nr
; i
++)
942 free(pl
->patterns
[i
]);
945 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
946 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
948 memset(pl
, 0, sizeof(*pl
));
951 static void trim_trailing_spaces(char *buf
)
953 char *p
, *last_space
= NULL
;
955 for (p
= buf
; *p
; p
++)
975 * Given a subdirectory name and "dir" of the current directory,
976 * search the subdir in "dir" and return it, or create a new one if it
977 * does not exist in "dir".
979 * If "name" has the trailing slash, it'll be excluded in the search.
981 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
982 struct untracked_cache_dir
*dir
,
983 const char *name
, int len
)
986 struct untracked_cache_dir
*d
;
989 if (len
&& name
[len
- 1] == '/')
993 while (last
> first
) {
994 int cmp
, next
= first
+ ((last
- first
) >> 1);
996 cmp
= strncmp(name
, d
->name
, len
);
997 if (!cmp
&& strlen(d
->name
) > len
)
1009 FLEX_ALLOC_MEM(d
, name
, name
, len
);
1011 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
1012 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
1013 dir
->dirs_nr
- first
);
1015 dir
->dirs
[first
] = d
;
1019 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
1023 dir
->untracked_nr
= 0;
1024 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1025 do_invalidate_gitignore(dir
->dirs
[i
]);
1028 static void invalidate_gitignore(struct untracked_cache
*uc
,
1029 struct untracked_cache_dir
*dir
)
1031 uc
->gitignore_invalidated
++;
1032 do_invalidate_gitignore(dir
);
1035 static void invalidate_directory(struct untracked_cache
*uc
,
1036 struct untracked_cache_dir
*dir
)
1041 * Invalidation increment here is just roughly correct. If
1042 * untracked_nr or any of dirs[].recurse is non-zero, we
1043 * should increment dir_invalidated too. But that's more
1047 uc
->dir_invalidated
++;
1050 dir
->untracked_nr
= 0;
1051 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1052 dir
->dirs
[i
]->recurse
= 0;
1055 static int add_patterns_from_buffer(char *buf
, size_t size
,
1056 const char *base
, int baselen
,
1057 struct pattern_list
*pl
);
1059 /* Flags for add_patterns() */
1060 #define PATTERN_NOFOLLOW (1<<0)
1063 * Given a file with name "fname", read it (either from disk, or from
1064 * an index if 'istate' is non-null), parse it and store the
1065 * exclude rules in "pl".
1067 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1068 * stat data from disk (only valid if add_patterns returns zero). If
1069 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1071 static int add_patterns(const char *fname
, const char *base
, int baselen
,
1072 struct pattern_list
*pl
, struct index_state
*istate
,
1073 unsigned flags
, struct oid_stat
*oid_stat
)
1081 if (flags
& PATTERN_NOFOLLOW
)
1082 fd
= open_nofollow(fname
, O_RDONLY
);
1084 fd
= open(fname
, O_RDONLY
);
1086 if (fd
< 0 || fstat(fd
, &st
) < 0) {
1088 warn_on_fopen_errors(fname
);
1093 r
= read_skip_worktree_file_from_index(istate
, fname
,
1099 size
= xsize_t(st
.st_size
);
1102 fill_stat_data(&oid_stat
->stat
, &st
);
1103 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
1104 oid_stat
->valid
= 1;
1109 buf
= xmallocz(size
);
1110 if (read_in_full(fd
, buf
, size
) != size
) {
1119 if (oid_stat
->valid
&&
1120 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
1121 ; /* no content change, oid_stat->oid still good */
1123 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
1124 !ce_stage(istate
->cache
[pos
]) &&
1125 ce_uptodate(istate
->cache
[pos
]) &&
1126 !would_convert_to_git(istate
, fname
))
1127 oidcpy(&oid_stat
->oid
,
1128 &istate
->cache
[pos
]->oid
);
1130 hash_object_file(the_hash_algo
, buf
, size
,
1131 "blob", &oid_stat
->oid
);
1132 fill_stat_data(&oid_stat
->stat
, &st
);
1133 oid_stat
->valid
= 1;
1137 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1141 static int add_patterns_from_buffer(char *buf
, size_t size
,
1142 const char *base
, int baselen
,
1143 struct pattern_list
*pl
)
1148 hashmap_init(&pl
->recursive_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1149 hashmap_init(&pl
->parent_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1153 if (skip_utf8_bom(&buf
, size
))
1154 size
-= buf
- pl
->filebuf
;
1158 for (i
= 0; i
< size
; i
++) {
1159 if (buf
[i
] == '\n') {
1160 if (entry
!= buf
+ i
&& entry
[0] != '#') {
1161 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
1162 trim_trailing_spaces(entry
);
1163 add_pattern(entry
, base
, baselen
, pl
, lineno
);
1166 entry
= buf
+ i
+ 1;
1172 int add_patterns_from_file_to_list(const char *fname
, const char *base
,
1173 int baselen
, struct pattern_list
*pl
,
1174 struct index_state
*istate
,
1177 return add_patterns(fname
, base
, baselen
, pl
, istate
, flags
, NULL
);
1180 int add_patterns_from_blob_to_list(
1181 struct object_id
*oid
,
1182 const char *base
, int baselen
,
1183 struct pattern_list
*pl
)
1189 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
1193 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1197 struct pattern_list
*add_pattern_list(struct dir_struct
*dir
,
1198 int group_type
, const char *src
)
1200 struct pattern_list
*pl
;
1201 struct exclude_list_group
*group
;
1203 group
= &dir
->exclude_list_group
[group_type
];
1204 ALLOC_GROW(group
->pl
, group
->nr
+ 1, group
->alloc
);
1205 pl
= &group
->pl
[group
->nr
++];
1206 memset(pl
, 0, sizeof(*pl
));
1212 * Used to set up core.excludesfile and .git/info/exclude lists.
1214 static void add_patterns_from_file_1(struct dir_struct
*dir
, const char *fname
,
1215 struct oid_stat
*oid_stat
)
1217 struct pattern_list
*pl
;
1219 * catch setup_standard_excludes() that's called before
1220 * dir->untracked is assigned. That function behaves
1221 * differently when dir->untracked is non-NULL.
1223 if (!dir
->untracked
)
1224 dir
->unmanaged_exclude_files
++;
1225 pl
= add_pattern_list(dir
, EXC_FILE
, fname
);
1226 if (add_patterns(fname
, "", 0, pl
, NULL
, 0, oid_stat
) < 0)
1227 die(_("cannot use %s as an exclude file"), fname
);
1230 void add_patterns_from_file(struct dir_struct
*dir
, const char *fname
)
1232 dir
->unmanaged_exclude_files
++; /* see validate_untracked_cache() */
1233 add_patterns_from_file_1(dir
, fname
, NULL
);
1236 int match_basename(const char *basename
, int basenamelen
,
1237 const char *pattern
, int prefix
, int patternlen
,
1240 if (prefix
== patternlen
) {
1241 if (patternlen
== basenamelen
&&
1242 !fspathncmp(pattern
, basename
, basenamelen
))
1244 } else if (flags
& PATTERN_FLAG_ENDSWITH
) {
1245 /* "*literal" matching against "fooliteral" */
1246 if (patternlen
- 1 <= basenamelen
&&
1247 !fspathncmp(pattern
+ 1,
1248 basename
+ basenamelen
- (patternlen
- 1),
1252 if (fnmatch_icase_mem(pattern
, patternlen
,
1253 basename
, basenamelen
,
1260 int match_pathname(const char *pathname
, int pathlen
,
1261 const char *base
, int baselen
,
1262 const char *pattern
, int prefix
, int patternlen
,
1269 * match with FNM_PATHNAME; the pattern has base implicitly
1272 if (*pattern
== '/') {
1279 * baselen does not count the trailing slash. base[] may or
1280 * may not end with a trailing slash though.
1282 if (pathlen
< baselen
+ 1 ||
1283 (baselen
&& pathname
[baselen
] != '/') ||
1284 fspathncmp(pathname
, base
, baselen
))
1287 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
1288 name
= pathname
+ pathlen
- namelen
;
1292 * if the non-wildcard part is longer than the
1293 * remaining pathname, surely it cannot match.
1295 if (prefix
> namelen
)
1298 if (fspathncmp(pattern
, name
, prefix
))
1301 patternlen
-= prefix
;
1306 * If the whole pattern did not have a wildcard,
1307 * then our prefix match is all we need; we
1308 * do not need to call fnmatch at all.
1310 if (!patternlen
&& !namelen
)
1314 return fnmatch_icase_mem(pattern
, patternlen
,
1320 * Scan the given exclude list in reverse to see whether pathname
1321 * should be ignored. The first match (i.e. the last on the list), if
1322 * any, determines the fate. Returns the exclude_list element which
1323 * matched, or NULL for undecided.
1325 static struct path_pattern
*last_matching_pattern_from_list(const char *pathname
,
1327 const char *basename
,
1329 struct pattern_list
*pl
,
1330 struct index_state
*istate
)
1332 struct path_pattern
*res
= NULL
; /* undecided */
1336 return NULL
; /* undefined */
1338 for (i
= pl
->nr
- 1; 0 <= i
; i
--) {
1339 struct path_pattern
*pattern
= pl
->patterns
[i
];
1340 const char *exclude
= pattern
->pattern
;
1341 int prefix
= pattern
->nowildcardlen
;
1343 if (pattern
->flags
& PATTERN_FLAG_MUSTBEDIR
) {
1344 *dtype
= resolve_dtype(*dtype
, istate
, pathname
, pathlen
);
1345 if (*dtype
!= DT_DIR
)
1349 if (pattern
->flags
& PATTERN_FLAG_NODIR
) {
1350 if (match_basename(basename
,
1351 pathlen
- (basename
- pathname
),
1352 exclude
, prefix
, pattern
->patternlen
,
1360 assert(pattern
->baselen
== 0 ||
1361 pattern
->base
[pattern
->baselen
- 1] == '/');
1362 if (match_pathname(pathname
, pathlen
,
1364 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1365 exclude
, prefix
, pattern
->patternlen
,
1375 * Scan the list of patterns to determine if the ordered list
1376 * of patterns matches on 'pathname'.
1378 * Return 1 for a match, 0 for not matched and -1 for undecided.
1380 enum pattern_match_result
path_matches_pattern_list(
1381 const char *pathname
, int pathlen
,
1382 const char *basename
, int *dtype
,
1383 struct pattern_list
*pl
,
1384 struct index_state
*istate
)
1386 struct path_pattern
*pattern
;
1387 struct strbuf parent_pathname
= STRBUF_INIT
;
1388 int result
= NOT_MATCHED
;
1389 const char *slash_pos
;
1391 if (!pl
->use_cone_patterns
) {
1392 pattern
= last_matching_pattern_from_list(pathname
, pathlen
, basename
,
1395 if (pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1407 strbuf_addch(&parent_pathname
, '/');
1408 strbuf_add(&parent_pathname
, pathname
, pathlen
);
1410 if (hashmap_contains_path(&pl
->recursive_hashmap
,
1411 &parent_pathname
)) {
1412 result
= MATCHED_RECURSIVE
;
1416 slash_pos
= strrchr(parent_pathname
.buf
, '/');
1418 if (slash_pos
== parent_pathname
.buf
) {
1419 /* include every file in root */
1424 strbuf_setlen(&parent_pathname
, slash_pos
- parent_pathname
.buf
);
1426 if (hashmap_contains_path(&pl
->parent_hashmap
, &parent_pathname
)) {
1431 if (hashmap_contains_parent(&pl
->recursive_hashmap
,
1434 result
= MATCHED_RECURSIVE
;
1437 strbuf_release(&parent_pathname
);
1441 static struct path_pattern
*last_matching_pattern_from_lists(
1442 struct dir_struct
*dir
, struct index_state
*istate
,
1443 const char *pathname
, int pathlen
,
1444 const char *basename
, int *dtype_p
)
1447 struct exclude_list_group
*group
;
1448 struct path_pattern
*pattern
;
1449 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1450 group
= &dir
->exclude_list_group
[i
];
1451 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1452 pattern
= last_matching_pattern_from_list(
1453 pathname
, pathlen
, basename
, dtype_p
,
1454 &group
->pl
[j
], istate
);
1463 * Loads the per-directory exclude list for the substring of base
1464 * which has a char length of baselen.
1466 static void prep_exclude(struct dir_struct
*dir
,
1467 struct index_state
*istate
,
1468 const char *base
, int baselen
)
1470 struct exclude_list_group
*group
;
1471 struct pattern_list
*pl
;
1472 struct exclude_stack
*stk
= NULL
;
1473 struct untracked_cache_dir
*untracked
;
1476 group
= &dir
->exclude_list_group
[EXC_DIRS
];
1479 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1480 * which originate from directories not in the prefix of the
1481 * path being checked.
1483 while ((stk
= dir
->exclude_stack
) != NULL
) {
1484 if (stk
->baselen
<= baselen
&&
1485 !strncmp(dir
->basebuf
.buf
, base
, stk
->baselen
))
1487 pl
= &group
->pl
[dir
->exclude_stack
->exclude_ix
];
1488 dir
->exclude_stack
= stk
->prev
;
1489 dir
->pattern
= NULL
;
1490 free((char *)pl
->src
); /* see strbuf_detach() below */
1491 clear_pattern_list(pl
);
1496 /* Skip traversing into sub directories if the parent is excluded */
1501 * Lazy initialization. All call sites currently just
1502 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1503 * them seems lots of work for little benefit.
1505 if (!dir
->basebuf
.buf
)
1506 strbuf_init(&dir
->basebuf
, PATH_MAX
);
1508 /* Read from the parent directories and push them down. */
1509 current
= stk
? stk
->baselen
: -1;
1510 strbuf_setlen(&dir
->basebuf
, current
< 0 ? 0 : current
);
1512 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1516 while (current
< baselen
) {
1518 struct oid_stat oid_stat
;
1520 CALLOC_ARRAY(stk
, 1);
1525 cp
= strchr(base
+ current
+ 1, '/');
1527 die("oops in prep_exclude");
1530 lookup_untracked(dir
->untracked
, untracked
,
1532 cp
- base
- current
);
1534 stk
->prev
= dir
->exclude_stack
;
1535 stk
->baselen
= cp
- base
;
1536 stk
->exclude_ix
= group
->nr
;
1537 stk
->ucd
= untracked
;
1538 pl
= add_pattern_list(dir
, EXC_DIRS
, NULL
);
1539 strbuf_add(&dir
->basebuf
, base
+ current
, stk
->baselen
- current
);
1540 assert(stk
->baselen
== dir
->basebuf
.len
);
1542 /* Abort if the directory is excluded */
1545 dir
->basebuf
.buf
[stk
->baselen
- 1] = 0;
1546 dir
->pattern
= last_matching_pattern_from_lists(dir
,
1548 dir
->basebuf
.buf
, stk
->baselen
- 1,
1549 dir
->basebuf
.buf
+ current
, &dt
);
1550 dir
->basebuf
.buf
[stk
->baselen
- 1] = '/';
1552 dir
->pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1553 dir
->pattern
= NULL
;
1555 dir
->exclude_stack
= stk
;
1560 /* Try to read per-directory file */
1561 oidclr(&oid_stat
.oid
);
1563 if (dir
->exclude_per_dir
&&
1565 * If we know that no files have been added in
1566 * this directory (i.e. valid_cached_dir() has
1567 * been executed and set untracked->valid) ..
1569 (!untracked
|| !untracked
->valid
||
1571 * .. and .gitignore does not exist before
1572 * (i.e. null exclude_oid). Then we can skip
1573 * loading .gitignore, which would result in
1576 !is_null_oid(&untracked
->exclude_oid
))) {
1578 * dir->basebuf gets reused by the traversal, but we
1579 * need fname to remain unchanged to ensure the src
1580 * member of each struct path_pattern correctly
1581 * back-references its source file. Other invocations
1582 * of add_pattern_list provide stable strings, so we
1583 * strbuf_detach() and free() here in the caller.
1585 struct strbuf sb
= STRBUF_INIT
;
1586 strbuf_addbuf(&sb
, &dir
->basebuf
);
1587 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1588 pl
->src
= strbuf_detach(&sb
, NULL
);
1589 add_patterns(pl
->src
, pl
->src
, stk
->baselen
, pl
, istate
,
1591 untracked
? &oid_stat
: NULL
);
1594 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1595 * will first be called in valid_cached_dir() then maybe many
1596 * times more in last_matching_pattern(). When the cache is
1597 * used, last_matching_pattern() will not be called and
1598 * reading .gitignore content will be a waste.
1600 * So when it's called by valid_cached_dir() and we can get
1601 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1602 * modified on work tree), we could delay reading the
1603 * .gitignore content until we absolutely need it in
1604 * last_matching_pattern(). Be careful about ignore rule
1605 * order, though, if you do that.
1608 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1609 invalidate_gitignore(dir
->untracked
, untracked
);
1610 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1612 dir
->exclude_stack
= stk
;
1613 current
= stk
->baselen
;
1615 strbuf_setlen(&dir
->basebuf
, baselen
);
1619 * Loads the exclude lists for the directory containing pathname, then
1620 * scans all exclude lists to determine whether pathname is excluded.
1621 * Returns the exclude_list element which matched, or NULL for
1624 struct path_pattern
*last_matching_pattern(struct dir_struct
*dir
,
1625 struct index_state
*istate
,
1626 const char *pathname
,
1629 int pathlen
= strlen(pathname
);
1630 const char *basename
= strrchr(pathname
, '/');
1631 basename
= (basename
) ? basename
+1 : pathname
;
1633 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1636 return dir
->pattern
;
1638 return last_matching_pattern_from_lists(dir
, istate
, pathname
, pathlen
,
1643 * Loads the exclude lists for the directory containing pathname, then
1644 * scans all exclude lists to determine whether pathname is excluded.
1645 * Returns 1 if true, otherwise 0.
1647 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1648 const char *pathname
, int *dtype_p
)
1650 struct path_pattern
*pattern
=
1651 last_matching_pattern(dir
, istate
, pathname
, dtype_p
);
1653 return pattern
->flags
& PATTERN_FLAG_NEGATIVE
? 0 : 1;
1657 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1659 struct dir_entry
*ent
;
1661 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1666 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1667 struct index_state
*istate
,
1668 const char *pathname
, int len
)
1670 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1673 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->alloc
);
1674 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1677 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1678 struct index_state
*istate
,
1679 const char *pathname
, int len
)
1681 if (!index_name_is_other(istate
, pathname
, len
))
1684 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->ignored_alloc
);
1685 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1689 index_nonexistent
= 0,
1695 * Do not use the alphabetically sorted index to look up
1696 * the directory name; instead, use the case insensitive
1699 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1700 const char *dirname
, int len
)
1702 struct cache_entry
*ce
;
1704 if (index_dir_exists(istate
, dirname
, len
))
1705 return index_directory
;
1707 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1708 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1709 return index_gitdir
;
1711 return index_nonexistent
;
1715 * The index sorts alphabetically by entry name, which
1716 * means that a gitlink sorts as '\0' at the end, while
1717 * a directory (which is defined not as an entry, but as
1718 * the files it contains) will sort with the '/' at the
1721 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1722 const char *dirname
, int len
)
1727 return directory_exists_in_index_icase(istate
, dirname
, len
);
1729 pos
= index_name_pos(istate
, dirname
, len
);
1732 while (pos
< istate
->cache_nr
) {
1733 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1734 unsigned char endchar
;
1736 if (strncmp(ce
->name
, dirname
, len
))
1738 endchar
= ce
->name
[len
];
1742 return index_directory
;
1743 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1744 return index_gitdir
;
1746 return index_nonexistent
;
1750 * When we find a directory when traversing the filesystem, we
1751 * have three distinct cases:
1754 * - see it as a directory
1757 * and which one we choose depends on a combination of existing
1758 * git index contents and the flags passed into the directory
1759 * traversal routine.
1761 * Case 1: If we *already* have entries in the index under that
1762 * directory name, we always recurse into the directory to see
1765 * Case 2: If we *already* have that directory name as a gitlink,
1766 * we always continue to see it as a gitlink, regardless of whether
1767 * there is an actual git directory there or not (it might not
1768 * be checked out as a subproject!)
1770 * Case 3: if we didn't have it in the index previously, we
1771 * have a few sub-cases:
1773 * (a) if DIR_SHOW_OTHER_DIRECTORIES flag is set, we show it as
1774 * just a directory, unless DIR_HIDE_EMPTY_DIRECTORIES is
1775 * also true, in which case we need to check if it contains any
1776 * untracked and / or ignored files.
1777 * (b) if it looks like a git directory and we don't have the
1778 * DIR_NO_GITLINKS flag, then we treat it as a gitlink, and
1779 * show it as a directory.
1780 * (c) otherwise, we recurse into it.
1782 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1783 struct index_state
*istate
,
1784 struct untracked_cache_dir
*untracked
,
1785 const char *dirname
, int len
, int baselen
, int excluded
,
1786 const struct pathspec
*pathspec
)
1789 * WARNING: From this function, you can return path_recurse or you
1790 * can call read_directory_recursive() (or neither), but
1791 * you CAN'T DO BOTH.
1793 enum path_treatment state
;
1794 int matches_how
= 0;
1795 int nested_repo
= 0, check_only
, stop_early
;
1796 int old_ignored_nr
, old_untracked_nr
;
1797 /* The "len-1" is to strip the final '/' */
1798 enum exist_status status
= directory_exists_in_index(istate
, dirname
, len
-1);
1800 if (status
== index_directory
)
1801 return path_recurse
;
1802 if (status
== index_gitdir
)
1804 if (status
!= index_nonexistent
)
1805 BUG("Unhandled value for directory_exists_in_index: %d\n", status
);
1808 * We don't want to descend into paths that don't match the necessary
1809 * patterns. Clearly, if we don't have a pathspec, then we can't check
1810 * for matching patterns. Also, if (excluded) then we know we matched
1811 * the exclusion patterns so as an optimization we can skip checking
1812 * for matching patterns.
1814 if (pathspec
&& !excluded
) {
1815 matches_how
= match_pathspec_with_flags(istate
, pathspec
,
1819 DO_MATCH_LEADING_PATHSPEC
);
1825 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1826 !(dir
->flags
& DIR_NO_GITLINKS
)) {
1827 struct strbuf sb
= STRBUF_INIT
;
1828 strbuf_addstr(&sb
, dirname
);
1829 nested_repo
= is_nonbare_repository_dir(&sb
);
1830 strbuf_release(&sb
);
1833 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1834 (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
))
1836 return excluded
? path_excluded
: path_untracked
;
1839 if (!(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)) {
1841 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1842 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1845 * This is an excluded directory and we are
1846 * showing ignored paths that match an exclude
1847 * pattern. (e.g. show directory as ignored
1848 * only if it matches an exclude pattern).
1849 * This path will either be 'path_excluded`
1850 * (if we are showing empty directories or if
1851 * the directory is not empty), or will be
1852 * 'path_none' (empty directory, and we are
1853 * not showing empty directories).
1855 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1856 return path_excluded
;
1858 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1859 untracked
, 1, 1, pathspec
) == path_excluded
)
1860 return path_excluded
;
1864 return path_recurse
;
1867 assert(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
);
1870 * If we have a pathspec which could match something _below_ this
1871 * directory (e.g. when checking 'subdir/' having a pathspec like
1872 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1875 if (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
)
1876 return path_recurse
;
1878 /* Special cases for where this directory is excluded/ignored */
1881 * If DIR_SHOW_OTHER_DIRECTORIES is set and we're not
1882 * hiding empty directories, there is no need to
1883 * recurse into an ignored directory.
1885 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1886 return path_excluded
;
1889 * Even if we are hiding empty directories, we can still avoid
1890 * recursing into ignored directories for DIR_SHOW_IGNORED_TOO
1891 * if DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1893 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1894 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
1895 return path_excluded
;
1899 * Other than the path_recurse case above, we only need to
1900 * recurse into untracked directories if any of the following
1902 * - DIR_SHOW_IGNORED (because then we need to determine if
1903 * there are ignored entries below)
1904 * - DIR_SHOW_IGNORED_TOO (same as above)
1905 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
1906 * the directory is empty)
1909 !(dir
->flags
& (DIR_SHOW_IGNORED
|
1910 DIR_SHOW_IGNORED_TOO
|
1911 DIR_HIDE_EMPTY_DIRECTORIES
))) {
1912 return path_untracked
;
1916 * Even if we don't want to know all the paths under an untracked or
1917 * ignored directory, we may still need to go into the directory to
1918 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
1919 * an empty directory should be path_none instead of path_excluded or
1922 check_only
= ((dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
) &&
1923 !(dir
->flags
& DIR_SHOW_IGNORED_TOO
));
1926 * However, there's another optimization possible as a subset of
1927 * check_only, based on the cases we have to consider:
1928 * A) Directory matches no exclude patterns:
1929 * * Directory is empty => path_none
1930 * * Directory has an untracked file under it => path_untracked
1931 * * Directory has only ignored files under it => path_excluded
1932 * B) Directory matches an exclude pattern:
1933 * * Directory is empty => path_none
1934 * * Directory has an untracked file under it => path_excluded
1935 * * Directory has only ignored files under it => path_excluded
1936 * In case A, we can exit as soon as we've found an untracked
1937 * file but otherwise have to walk all files. In case B, though,
1938 * we can stop at the first file we find under the directory.
1940 stop_early
= check_only
&& excluded
;
1943 * If /every/ file within an untracked directory is ignored, then
1944 * we want to treat the directory as ignored (for e.g. status
1945 * --porcelain), without listing the individual ignored files
1946 * underneath. To do so, we'll save the current ignored_nr, and
1947 * pop all the ones added after it if it turns out the entire
1948 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
1949 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
1950 * untracked paths so will need to pop all those off the last
1951 * after we traverse.
1953 old_ignored_nr
= dir
->ignored_nr
;
1954 old_untracked_nr
= dir
->nr
;
1956 /* Actually recurse into dirname now, we'll fixup the state later. */
1957 untracked
= lookup_untracked(dir
->untracked
, untracked
,
1958 dirname
+ baselen
, len
- baselen
);
1959 state
= read_directory_recursive(dir
, istate
, dirname
, len
, untracked
,
1960 check_only
, stop_early
, pathspec
);
1962 /* There are a variety of reasons we may need to fixup the state... */
1963 if (state
== path_excluded
) {
1964 /* state == path_excluded implies all paths under
1965 * dirname were ignored...
1967 * if running e.g. `git status --porcelain --ignored=matching`,
1968 * then we want to see the subpaths that are ignored.
1970 * if running e.g. just `git status --porcelain`, then
1971 * we just want the directory itself to be listed as ignored
1972 * and not the individual paths underneath.
1974 int want_ignored_subpaths
=
1975 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1976 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
));
1978 if (want_ignored_subpaths
) {
1980 * with --ignored=matching, we want the subpaths
1981 * INSTEAD of the directory itself.
1986 for (i
= old_ignored_nr
+ 1; i
<dir
->ignored_nr
; ++i
)
1987 FREE_AND_NULL(dir
->ignored
[i
]);
1988 dir
->ignored_nr
= old_ignored_nr
;
1993 * We may need to ignore some of the untracked paths we found while
1994 * traversing subdirectories.
1996 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1997 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
1999 for (i
= old_untracked_nr
+ 1; i
<dir
->nr
; ++i
)
2000 FREE_AND_NULL(dir
->entries
[i
]);
2001 dir
->nr
= old_untracked_nr
;
2005 * If there is nothing under the current directory and we are not
2006 * hiding empty directories, then we need to report on the
2007 * untracked or ignored status of the directory itself.
2009 if (state
== path_none
&& !(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
2010 state
= excluded
? path_excluded
: path_untracked
;
2016 * This is an inexact early pruning of any recursive directory
2017 * reading - if the path cannot possibly be in the pathspec,
2018 * return true, and we'll skip it early.
2020 static int simplify_away(const char *path
, int pathlen
,
2021 const struct pathspec
*pathspec
)
2025 if (!pathspec
|| !pathspec
->nr
)
2028 GUARD_PATHSPEC(pathspec
,
2037 for (i
= 0; i
< pathspec
->nr
; i
++) {
2038 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2039 int len
= item
->nowildcard_len
;
2043 if (!ps_strncmp(item
, item
->match
, path
, len
))
2051 * This function tells us whether an excluded path matches a
2052 * list of "interesting" pathspecs. That is, whether a path matched
2053 * by any of the pathspecs could possibly be ignored by excluding
2054 * the specified path. This can happen if:
2056 * 1. the path is mentioned explicitly in the pathspec
2058 * 2. the path is a directory prefix of some element in the
2061 static int exclude_matches_pathspec(const char *path
, int pathlen
,
2062 const struct pathspec
*pathspec
)
2066 if (!pathspec
|| !pathspec
->nr
)
2069 GUARD_PATHSPEC(pathspec
,
2077 for (i
= 0; i
< pathspec
->nr
; i
++) {
2078 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2079 int len
= item
->nowildcard_len
;
2081 if (len
== pathlen
&&
2082 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2084 if (len
> pathlen
&&
2085 item
->match
[pathlen
] == '/' &&
2086 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2092 static int get_index_dtype(struct index_state
*istate
,
2093 const char *path
, int len
)
2096 const struct cache_entry
*ce
;
2098 ce
= index_file_exists(istate
, path
, len
, 0);
2100 if (!ce_uptodate(ce
))
2102 if (S_ISGITLINK(ce
->ce_mode
))
2105 * Nobody actually cares about the
2106 * difference between DT_LNK and DT_REG
2111 /* Try to look it up as a directory */
2112 pos
= index_name_pos(istate
, path
, len
);
2116 while (pos
< istate
->cache_nr
) {
2117 ce
= istate
->cache
[pos
++];
2118 if (strncmp(ce
->name
, path
, len
))
2120 if (ce
->name
[len
] > '/')
2122 if (ce
->name
[len
] < '/')
2124 if (!ce_uptodate(ce
))
2125 break; /* continue? */
2131 static int resolve_dtype(int dtype
, struct index_state
*istate
,
2132 const char *path
, int len
)
2136 if (dtype
!= DT_UNKNOWN
)
2138 dtype
= get_index_dtype(istate
, path
, len
);
2139 if (dtype
!= DT_UNKNOWN
)
2141 if (lstat(path
, &st
))
2143 if (S_ISREG(st
.st_mode
))
2145 if (S_ISDIR(st
.st_mode
))
2147 if (S_ISLNK(st
.st_mode
))
2152 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
2153 struct cached_dir
*cdir
,
2154 struct index_state
*istate
,
2155 struct strbuf
*path
,
2157 const struct pathspec
*pathspec
)
2160 * WARNING: From this function, you can return path_recurse or you
2161 * can call read_directory_recursive() (or neither), but
2162 * you CAN'T DO BOTH.
2164 strbuf_setlen(path
, baselen
);
2166 strbuf_addstr(path
, cdir
->file
);
2167 return path_untracked
;
2169 strbuf_addstr(path
, cdir
->ucd
->name
);
2170 /* treat_one_path() does this before it calls treat_directory() */
2171 strbuf_complete(path
, '/');
2172 if (cdir
->ucd
->check_only
)
2174 * check_only is set as a result of treat_directory() getting
2175 * to its bottom. Verify again the same set of directories
2176 * with check_only set.
2178 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
2179 cdir
->ucd
, 1, 0, pathspec
);
2181 * We get path_recurse in the first run when
2182 * directory_exists_in_index() returns index_nonexistent. We
2183 * are sure that new changes in the index does not impact the
2184 * outcome. Return now.
2186 return path_recurse
;
2189 static enum path_treatment
treat_path(struct dir_struct
*dir
,
2190 struct untracked_cache_dir
*untracked
,
2191 struct cached_dir
*cdir
,
2192 struct index_state
*istate
,
2193 struct strbuf
*path
,
2195 const struct pathspec
*pathspec
)
2197 int has_path_in_index
, dtype
, excluded
;
2200 return treat_path_fast(dir
, cdir
, istate
, path
,
2202 if (is_dot_or_dotdot(cdir
->d_name
) || !fspathcmp(cdir
->d_name
, ".git"))
2204 strbuf_setlen(path
, baselen
);
2205 strbuf_addstr(path
, cdir
->d_name
);
2206 if (simplify_away(path
->buf
, path
->len
, pathspec
))
2209 dtype
= resolve_dtype(cdir
->d_type
, istate
, path
->buf
, path
->len
);
2211 /* Always exclude indexed files */
2212 has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
,
2214 if (dtype
!= DT_DIR
&& has_path_in_index
)
2218 * When we are looking at a directory P in the working tree,
2219 * there are three cases:
2221 * (1) P exists in the index. Everything inside the directory P in
2222 * the working tree needs to go when P is checked out from the
2225 * (2) P does not exist in the index, but there is P/Q in the index.
2226 * We know P will stay a directory when we check out the contents
2227 * of the index, but we do not know yet if there is a directory
2228 * P/Q in the working tree to be killed, so we need to recurse.
2230 * (3) P does not exist in the index, and there is no P/Q in the index
2231 * to require P to be a directory, either. Only in this case, we
2232 * know that everything inside P will not be killed without
2235 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
2236 (dtype
== DT_DIR
) &&
2237 !has_path_in_index
&&
2238 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
2241 excluded
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
2244 * Excluded? If we don't explicitly want to show
2245 * ignored files, ignore it
2247 if (excluded
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
2248 return path_excluded
;
2255 * WARNING: Do not ignore/amend the return value from
2256 * treat_directory(), and especially do not change it to return
2257 * path_recurse as that can cause exponential slowdown.
2258 * Instead, modify treat_directory() to return the right value.
2260 strbuf_addch(path
, '/');
2261 return treat_directory(dir
, istate
, untracked
,
2262 path
->buf
, path
->len
,
2263 baselen
, excluded
, pathspec
);
2267 !match_pathspec(istate
, pathspec
, path
->buf
, path
->len
,
2268 0 /* prefix */, NULL
/* seen */,
2272 return path_excluded
;
2273 return path_untracked
;
2277 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
2281 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
2282 dir
->untracked_alloc
);
2283 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
2286 static int valid_cached_dir(struct dir_struct
*dir
,
2287 struct untracked_cache_dir
*untracked
,
2288 struct index_state
*istate
,
2289 struct strbuf
*path
,
2298 * With fsmonitor, we can trust the untracked cache's valid field.
2300 refresh_fsmonitor(istate
);
2301 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
2302 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
2303 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
2306 if (!untracked
->valid
||
2307 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
2308 fill_stat_data(&untracked
->stat_data
, &st
);
2313 if (untracked
->check_only
!= !!check_only
)
2317 * prep_exclude will be called eventually on this directory,
2318 * but it's called much later in last_matching_pattern(). We
2319 * need it now to determine the validity of the cache for this
2320 * path. The next calls will be nearly no-op, the way
2321 * prep_exclude() is designed.
2323 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
2324 strbuf_addch(path
, '/');
2325 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2326 strbuf_setlen(path
, path
->len
- 1);
2328 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2330 /* hopefully prep_exclude() haven't invalidated this entry... */
2331 return untracked
->valid
;
2334 static int open_cached_dir(struct cached_dir
*cdir
,
2335 struct dir_struct
*dir
,
2336 struct untracked_cache_dir
*untracked
,
2337 struct index_state
*istate
,
2338 struct strbuf
*path
,
2343 memset(cdir
, 0, sizeof(*cdir
));
2344 cdir
->untracked
= untracked
;
2345 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
2347 c_path
= path
->len
? path
->buf
: ".";
2348 cdir
->fdir
= opendir(c_path
);
2350 warning_errno(_("could not open directory '%s'"), c_path
);
2351 if (dir
->untracked
) {
2352 invalidate_directory(dir
->untracked
, untracked
);
2353 dir
->untracked
->dir_opened
++;
2360 static int read_cached_dir(struct cached_dir
*cdir
)
2365 de
= readdir_skip_dot_and_dotdot(cdir
->fdir
);
2367 cdir
->d_name
= NULL
;
2368 cdir
->d_type
= DT_UNKNOWN
;
2371 cdir
->d_name
= de
->d_name
;
2372 cdir
->d_type
= DTYPE(de
);
2375 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
2376 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
2386 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
2387 struct untracked_cache_dir
*d
= cdir
->untracked
;
2388 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
2394 static void close_cached_dir(struct cached_dir
*cdir
)
2397 closedir(cdir
->fdir
);
2399 * We have gone through this directory and found no untracked
2400 * entries. Mark it valid.
2402 if (cdir
->untracked
) {
2403 cdir
->untracked
->valid
= 1;
2404 cdir
->untracked
->recurse
= 1;
2408 static void add_path_to_appropriate_result_list(struct dir_struct
*dir
,
2409 struct untracked_cache_dir
*untracked
,
2410 struct cached_dir
*cdir
,
2411 struct index_state
*istate
,
2412 struct strbuf
*path
,
2414 const struct pathspec
*pathspec
,
2415 enum path_treatment state
)
2417 /* add the path to the appropriate result list */
2420 if (dir
->flags
& DIR_SHOW_IGNORED
)
2421 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2422 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2423 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
2424 exclude_matches_pathspec(path
->buf
, path
->len
,
2426 dir_add_ignored(dir
, istate
, path
->buf
, path
->len
);
2429 case path_untracked
:
2430 if (dir
->flags
& DIR_SHOW_IGNORED
)
2432 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2434 add_untracked(untracked
, path
->buf
+ baselen
);
2443 * Read a directory tree. We currently ignore anything but
2444 * directories, regular files and symlinks. That's because git
2445 * doesn't handle them at all yet. Maybe that will change some
2448 * Also, we ignore the name ".git" (even if it is not a directory).
2449 * That likely will not change.
2451 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2452 * to signal that a file was found. This is the least significant value that
2453 * indicates that a file was encountered that does not depend on the order of
2454 * whether an untracked or excluded path was encountered first.
2456 * Returns the most significant path_treatment value encountered in the scan.
2457 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2458 * significant path_treatment value that will be returned.
2461 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
2462 struct index_state
*istate
, const char *base
, int baselen
,
2463 struct untracked_cache_dir
*untracked
, int check_only
,
2464 int stop_at_first_file
, const struct pathspec
*pathspec
)
2467 * WARNING: Do NOT recurse unless path_recurse is returned from
2468 * treat_path(). Recursing on any other return value
2469 * can result in exponential slowdown.
2471 struct cached_dir cdir
;
2472 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
2473 struct strbuf path
= STRBUF_INIT
;
2475 strbuf_add(&path
, base
, baselen
);
2477 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
2479 dir
->visited_directories
++;
2482 untracked
->check_only
= !!check_only
;
2484 while (!read_cached_dir(&cdir
)) {
2485 /* check how the file or directory should be treated */
2486 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
2488 dir
->visited_paths
++;
2490 if (state
> dir_state
)
2493 /* recurse into subdir if instructed by treat_path */
2494 if (state
== path_recurse
) {
2495 struct untracked_cache_dir
*ud
;
2496 ud
= lookup_untracked(dir
->untracked
, untracked
,
2498 path
.len
- baselen
);
2500 read_directory_recursive(dir
, istate
, path
.buf
,
2502 check_only
, stop_at_first_file
, pathspec
);
2503 if (subdir_state
> dir_state
)
2504 dir_state
= subdir_state
;
2507 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2508 0 /* prefix */, NULL
,
2509 0 /* do NOT special case dirs */))
2514 if (stop_at_first_file
) {
2516 * If stopping at first file, then
2517 * signal that a file was found by
2518 * returning `path_excluded`. This is
2519 * to return a consistent value
2520 * regardless of whether an ignored or
2521 * excluded file happened to be
2524 * In current usage, the
2525 * `stop_at_first_file` is passed when
2526 * an ancestor directory has matched
2527 * an exclude pattern, so any found
2528 * files will be excluded.
2530 if (dir_state
>= path_excluded
) {
2531 dir_state
= path_excluded
;
2536 /* abort early if maximum state has been reached */
2537 if (dir_state
== path_untracked
) {
2539 add_untracked(untracked
, path
.buf
+ baselen
);
2542 /* skip the add_path_to_appropriate_result_list() */
2546 add_path_to_appropriate_result_list(dir
, untracked
, &cdir
,
2547 istate
, &path
, baselen
,
2550 close_cached_dir(&cdir
);
2552 strbuf_release(&path
);
2557 int cmp_dir_entry(const void *p1
, const void *p2
)
2559 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2560 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2562 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2565 /* check if *out lexically strictly contains *in */
2566 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2568 return (out
->len
< in
->len
) &&
2569 (out
->name
[out
->len
- 1] == '/') &&
2570 !memcmp(out
->name
, in
->name
, out
->len
);
2573 static int treat_leading_path(struct dir_struct
*dir
,
2574 struct index_state
*istate
,
2575 const char *path
, int len
,
2576 const struct pathspec
*pathspec
)
2578 struct strbuf sb
= STRBUF_INIT
;
2579 struct strbuf subdir
= STRBUF_INIT
;
2580 int prevlen
, baselen
;
2582 struct cached_dir cdir
;
2583 enum path_treatment state
= path_none
;
2586 * For each directory component of path, we are going to check whether
2587 * that path is relevant given the pathspec. For example, if path is
2589 * then we will ask treat_path() whether we should go into foo, then
2590 * whether we should go into bar, then whether baz is relevant.
2591 * Checking each is important because e.g. if path is
2593 * then we need to check .git to know we shouldn't traverse it.
2594 * If the return from treat_path() is:
2595 * * path_none, for any path, we return false.
2596 * * path_recurse, for all path components, we return true
2597 * * <anything else> for some intermediate component, we make sure
2598 * to add that path to the relevant list but return false
2599 * signifying that we shouldn't recurse into it.
2602 while (len
&& path
[len
- 1] == '/')
2607 memset(&cdir
, 0, sizeof(cdir
));
2608 cdir
.d_type
= DT_DIR
;
2612 prevlen
= baselen
+ !!baselen
;
2613 cp
= path
+ prevlen
;
2614 cp
= memchr(cp
, '/', path
+ len
- cp
);
2618 baselen
= cp
- path
;
2620 strbuf_add(&sb
, path
, baselen
);
2621 if (!is_directory(sb
.buf
))
2624 strbuf_add(&sb
, path
, prevlen
);
2625 strbuf_reset(&subdir
);
2626 strbuf_add(&subdir
, path
+prevlen
, baselen
-prevlen
);
2627 cdir
.d_name
= subdir
.buf
;
2628 state
= treat_path(dir
, NULL
, &cdir
, istate
, &sb
, prevlen
, pathspec
);
2630 if (state
!= path_recurse
)
2631 break; /* do not recurse into it */
2633 break; /* finished checking */
2635 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
, istate
,
2636 &sb
, baselen
, pathspec
,
2639 strbuf_release(&subdir
);
2640 strbuf_release(&sb
);
2641 return state
== path_recurse
;
2644 static const char *get_ident_string(void)
2646 static struct strbuf sb
= STRBUF_INIT
;
2651 if (uname(&uts
) < 0)
2652 die_errno(_("failed to get kernel name and information"));
2653 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2658 static int ident_in_untracked(const struct untracked_cache
*uc
)
2661 * Previous git versions may have saved many NUL separated
2662 * strings in the "ident" field, but it is insane to manage
2663 * many locations, so just take care of the first one.
2666 return !strcmp(uc
->ident
.buf
, get_ident_string());
2669 static void set_untracked_ident(struct untracked_cache
*uc
)
2671 strbuf_reset(&uc
->ident
);
2672 strbuf_addstr(&uc
->ident
, get_ident_string());
2675 * This strbuf used to contain a list of NUL separated
2676 * strings, so save NUL too for backward compatibility.
2678 strbuf_addch(&uc
->ident
, 0);
2681 static void new_untracked_cache(struct index_state
*istate
)
2683 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2684 strbuf_init(&uc
->ident
, 100);
2685 uc
->exclude_per_dir
= ".gitignore";
2686 /* should be the same flags used by git-status */
2687 uc
->dir_flags
= DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2688 set_untracked_ident(uc
);
2689 istate
->untracked
= uc
;
2690 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2693 void add_untracked_cache(struct index_state
*istate
)
2695 if (!istate
->untracked
) {
2696 new_untracked_cache(istate
);
2698 if (!ident_in_untracked(istate
->untracked
)) {
2699 free_untracked_cache(istate
->untracked
);
2700 new_untracked_cache(istate
);
2705 void remove_untracked_cache(struct index_state
*istate
)
2707 if (istate
->untracked
) {
2708 free_untracked_cache(istate
->untracked
);
2709 istate
->untracked
= NULL
;
2710 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2714 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2716 const struct pathspec
*pathspec
)
2718 struct untracked_cache_dir
*root
;
2719 static int untracked_cache_disabled
= -1;
2721 if (!dir
->untracked
)
2723 if (untracked_cache_disabled
< 0)
2724 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2725 if (untracked_cache_disabled
)
2729 * We only support $GIT_DIR/info/exclude and core.excludesfile
2730 * as the global ignore rule files. Any other additions
2731 * (e.g. from command line) invalidate the cache. This
2732 * condition also catches running setup_standard_excludes()
2733 * before setting dir->untracked!
2735 if (dir
->unmanaged_exclude_files
)
2739 * Optimize for the main use case only: whole-tree git
2740 * status. More work involved in treat_leading_path() if we
2741 * use cache on just a subset of the worktree. pathspec
2742 * support could make the matter even worse.
2744 if (base_len
|| (pathspec
&& pathspec
->nr
))
2747 /* Different set of flags may produce different results */
2748 if (dir
->flags
!= dir
->untracked
->dir_flags
||
2750 * See treat_directory(), case index_nonexistent. Without
2751 * this flag, we may need to also cache .git file content
2752 * for the resolve_gitlink_ref() call, which we don't.
2754 !(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
) ||
2755 /* We don't support collecting ignore files */
2756 (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2757 DIR_COLLECT_IGNORED
)))
2761 * If we use .gitignore in the cache and now you change it to
2762 * .gitexclude, everything will go wrong.
2764 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2765 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2769 * EXC_CMDL is not considered in the cache. If people set it,
2772 if (dir
->exclude_list_group
[EXC_CMDL
].nr
)
2775 if (!ident_in_untracked(dir
->untracked
)) {
2776 warning(_("untracked cache is disabled on this system or location"));
2780 if (!dir
->untracked
->root
)
2781 FLEX_ALLOC_STR(dir
->untracked
->root
, name
, "");
2783 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2784 root
= dir
->untracked
->root
;
2785 if (!oideq(&dir
->ss_info_exclude
.oid
,
2786 &dir
->untracked
->ss_info_exclude
.oid
)) {
2787 invalidate_gitignore(dir
->untracked
, root
);
2788 dir
->untracked
->ss_info_exclude
= dir
->ss_info_exclude
;
2790 if (!oideq(&dir
->ss_excludes_file
.oid
,
2791 &dir
->untracked
->ss_excludes_file
.oid
)) {
2792 invalidate_gitignore(dir
->untracked
, root
);
2793 dir
->untracked
->ss_excludes_file
= dir
->ss_excludes_file
;
2796 /* Make sure this directory is not dropped out at saving phase */
2801 static void emit_traversal_statistics(struct dir_struct
*dir
,
2802 struct repository
*repo
,
2806 if (!trace2_is_enabled())
2810 trace2_data_string("read_directory", repo
, "path", "");
2812 struct strbuf tmp
= STRBUF_INIT
;
2813 strbuf_add(&tmp
, path
, path_len
);
2814 trace2_data_string("read_directory", repo
, "path", tmp
.buf
);
2815 strbuf_release(&tmp
);
2818 trace2_data_intmax("read_directory", repo
,
2819 "directories-visited", dir
->visited_directories
);
2820 trace2_data_intmax("read_directory", repo
,
2821 "paths-visited", dir
->visited_paths
);
2823 if (!dir
->untracked
)
2825 trace2_data_intmax("read_directory", repo
,
2826 "node-creation", dir
->untracked
->dir_created
);
2827 trace2_data_intmax("read_directory", repo
,
2828 "gitignore-invalidation",
2829 dir
->untracked
->gitignore_invalidated
);
2830 trace2_data_intmax("read_directory", repo
,
2831 "directory-invalidation",
2832 dir
->untracked
->dir_invalidated
);
2833 trace2_data_intmax("read_directory", repo
,
2834 "opendir", dir
->untracked
->dir_opened
);
2837 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
2838 const char *path
, int len
, const struct pathspec
*pathspec
)
2840 struct untracked_cache_dir
*untracked
;
2842 trace2_region_enter("dir", "read_directory", istate
->repo
);
2843 dir
->visited_paths
= 0;
2844 dir
->visited_directories
= 0;
2846 if (has_symlink_leading_path(path
, len
)) {
2847 trace2_region_leave("dir", "read_directory", istate
->repo
);
2851 untracked
= validate_untracked_cache(dir
, len
, pathspec
);
2854 * make sure untracked cache code path is disabled,
2855 * e.g. prep_exclude()
2857 dir
->untracked
= NULL
;
2858 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
2859 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
2860 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
2861 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
2863 emit_traversal_statistics(dir
, istate
->repo
, path
, len
);
2865 trace2_region_leave("dir", "read_directory", istate
->repo
);
2866 if (dir
->untracked
) {
2867 static int force_untracked_cache
= -1;
2869 if (force_untracked_cache
< 0)
2870 force_untracked_cache
=
2871 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2872 if (force_untracked_cache
&&
2873 dir
->untracked
== istate
->untracked
&&
2874 (dir
->untracked
->dir_opened
||
2875 dir
->untracked
->gitignore_invalidated
||
2876 dir
->untracked
->dir_invalidated
))
2877 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2878 if (dir
->untracked
!= istate
->untracked
) {
2879 FREE_AND_NULL(dir
->untracked
);
2886 int file_exists(const char *f
)
2889 return lstat(f
, &sb
) == 0;
2892 int repo_file_exists(struct repository
*repo
, const char *path
)
2894 if (repo
!= the_repository
)
2895 BUG("do not know how to check file existence in arbitrary repo");
2897 return file_exists(path
);
2900 static int cmp_icase(char a
, char b
)
2905 return toupper(a
) - toupper(b
);
2910 * Given two normalized paths (a trailing slash is ok), if subdir is
2911 * outside dir, return -1. Otherwise return the offset in subdir that
2912 * can be used as relative path to dir.
2914 int dir_inside_of(const char *subdir
, const char *dir
)
2918 assert(dir
&& subdir
&& *dir
&& *subdir
);
2920 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
2926 /* hel[p]/me vs hel[l]/yeah */
2927 if (*dir
&& *subdir
)
2931 return !*dir
? offset
: -1; /* same dir */
2933 /* foo/[b]ar vs foo/[] */
2934 if (is_dir_sep(dir
[-1]))
2935 return is_dir_sep(subdir
[-1]) ? offset
: -1;
2937 /* foo[/]bar vs foo[] */
2938 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
2941 int is_inside_dir(const char *dir
)
2950 rc
= (dir_inside_of(cwd
, dir
) >= 0);
2955 int is_empty_dir(const char *path
)
2957 DIR *dir
= opendir(path
);
2964 e
= readdir_skip_dot_and_dotdot(dir
);
2972 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
2976 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
2977 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
2978 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
2979 struct object_id submodule_head
;
2981 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
2982 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
2983 /* Do not descend and nuke a nested git work tree. */
2989 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
2990 dir
= opendir(path
->buf
);
2992 if (errno
== ENOENT
)
2993 return keep_toplevel
? -1 : 0;
2994 else if (errno
== EACCES
&& !keep_toplevel
)
2996 * An empty dir could be removable even if it
2999 return rmdir(path
->buf
);
3003 strbuf_complete(path
, '/');
3006 while ((e
= readdir_skip_dot_and_dotdot(dir
)) != NULL
) {
3009 strbuf_setlen(path
, len
);
3010 strbuf_addstr(path
, e
->d_name
);
3011 if (lstat(path
->buf
, &st
)) {
3012 if (errno
== ENOENT
)
3014 * file disappeared, which is what we
3019 } else if (S_ISDIR(st
.st_mode
)) {
3020 if (!remove_dir_recurse(path
, flag
, &kept_down
))
3021 continue; /* happy */
3022 } else if (!only_empty
&&
3023 (!unlink(path
->buf
) || errno
== ENOENT
)) {
3024 continue; /* happy, too */
3027 /* path too long, stat fails, or non-directory still exists */
3033 strbuf_setlen(path
, original_len
);
3034 if (!ret
&& !keep_toplevel
&& !kept_down
)
3035 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
3038 * report the uplevel that it is not an error that we
3039 * did not rmdir() our directory.
3045 int remove_dir_recursively(struct strbuf
*path
, int flag
)
3047 return remove_dir_recurse(path
, flag
, NULL
);
3050 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
3052 void setup_standard_excludes(struct dir_struct
*dir
)
3054 dir
->exclude_per_dir
= ".gitignore";
3056 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
3058 excludes_file
= xdg_config_home("ignore");
3059 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
3060 add_patterns_from_file_1(dir
, excludes_file
,
3061 dir
->untracked
? &dir
->ss_excludes_file
: NULL
);
3063 /* per repository user preference */
3064 if (startup_info
->have_repository
) {
3065 const char *path
= git_path_info_exclude();
3066 if (!access_or_warn(path
, R_OK
, 0))
3067 add_patterns_from_file_1(dir
, path
,
3068 dir
->untracked
? &dir
->ss_info_exclude
: NULL
);
3072 char *get_sparse_checkout_filename(void)
3074 return git_pathdup("info/sparse-checkout");
3077 int get_sparse_checkout_patterns(struct pattern_list
*pl
)
3080 char *sparse_filename
= get_sparse_checkout_filename();
3082 pl
->use_cone_patterns
= core_sparse_checkout_cone
;
3083 res
= add_patterns_from_file_to_list(sparse_filename
, "", 0, pl
, NULL
, 0);
3085 free(sparse_filename
);
3089 int remove_path(const char *name
)
3093 if (unlink(name
) && !is_missing_file_error(errno
))
3096 slash
= strrchr(name
, '/');
3098 char *dirs
= xstrdup(name
);
3099 slash
= dirs
+ (slash
- name
);
3102 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
3109 * Frees memory within dir which was allocated, and resets fields for further
3110 * use. Does not free dir itself.
3112 void dir_clear(struct dir_struct
*dir
)
3115 struct exclude_list_group
*group
;
3116 struct pattern_list
*pl
;
3117 struct exclude_stack
*stk
;
3119 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
3120 group
= &dir
->exclude_list_group
[i
];
3121 for (j
= 0; j
< group
->nr
; j
++) {
3124 free((char *)pl
->src
);
3125 clear_pattern_list(pl
);
3130 for (i
= 0; i
< dir
->ignored_nr
; i
++)
3131 free(dir
->ignored
[i
]);
3132 for (i
= 0; i
< dir
->nr
; i
++)
3133 free(dir
->entries
[i
]);
3137 stk
= dir
->exclude_stack
;
3139 struct exclude_stack
*prev
= stk
->prev
;
3143 strbuf_release(&dir
->basebuf
);
3148 struct ondisk_untracked_cache
{
3149 struct stat_data info_exclude_stat
;
3150 struct stat_data excludes_file_stat
;
3154 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3157 int index
; /* number of written untracked_cache_dir */
3158 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
3159 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
3160 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
3162 struct strbuf sb_stat
;
3163 struct strbuf sb_sha1
;
3166 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
3168 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
3169 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
3170 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
3171 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
3172 to
->sd_dev
= htonl(from
->sd_dev
);
3173 to
->sd_ino
= htonl(from
->sd_ino
);
3174 to
->sd_uid
= htonl(from
->sd_uid
);
3175 to
->sd_gid
= htonl(from
->sd_gid
);
3176 to
->sd_size
= htonl(from
->sd_size
);
3179 static void write_one_dir(struct untracked_cache_dir
*untracked
,
3180 struct write_data
*wd
)
3182 struct stat_data stat_data
;
3183 struct strbuf
*out
= &wd
->out
;
3184 unsigned char intbuf
[16];
3185 unsigned int intlen
, value
;
3186 int i
= wd
->index
++;
3189 * untracked_nr should be reset whenever valid is clear, but
3192 if (!untracked
->valid
) {
3193 untracked
->untracked_nr
= 0;
3194 untracked
->check_only
= 0;
3197 if (untracked
->check_only
)
3198 ewah_set(wd
->check_only
, i
);
3199 if (untracked
->valid
) {
3200 ewah_set(wd
->valid
, i
);
3201 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
3202 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
3204 if (!is_null_oid(&untracked
->exclude_oid
)) {
3205 ewah_set(wd
->sha1_valid
, i
);
3206 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
3207 the_hash_algo
->rawsz
);
3210 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
3211 strbuf_add(out
, intbuf
, intlen
);
3213 /* skip non-recurse directories */
3214 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
3215 if (untracked
->dirs
[i
]->recurse
)
3217 intlen
= encode_varint(value
, intbuf
);
3218 strbuf_add(out
, intbuf
, intlen
);
3220 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
3222 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
3223 strbuf_add(out
, untracked
->untracked
[i
],
3224 strlen(untracked
->untracked
[i
]) + 1);
3226 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
3227 if (untracked
->dirs
[i
]->recurse
)
3228 write_one_dir(untracked
->dirs
[i
], wd
);
3231 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
3233 struct ondisk_untracked_cache
*ouc
;
3234 struct write_data wd
;
3235 unsigned char varbuf
[16];
3237 const unsigned hashsz
= the_hash_algo
->rawsz
;
3239 CALLOC_ARRAY(ouc
, 1);
3240 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
3241 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
3242 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
3244 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
3245 strbuf_add(out
, varbuf
, varint_len
);
3246 strbuf_addbuf(out
, &untracked
->ident
);
3248 strbuf_add(out
, ouc
, sizeof(*ouc
));
3249 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
3250 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
3251 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
3254 if (!untracked
->root
) {
3255 varint_len
= encode_varint(0, varbuf
);
3256 strbuf_add(out
, varbuf
, varint_len
);
3261 wd
.check_only
= ewah_new();
3262 wd
.valid
= ewah_new();
3263 wd
.sha1_valid
= ewah_new();
3264 strbuf_init(&wd
.out
, 1024);
3265 strbuf_init(&wd
.sb_stat
, 1024);
3266 strbuf_init(&wd
.sb_sha1
, 1024);
3267 write_one_dir(untracked
->root
, &wd
);
3269 varint_len
= encode_varint(wd
.index
, varbuf
);
3270 strbuf_add(out
, varbuf
, varint_len
);
3271 strbuf_addbuf(out
, &wd
.out
);
3272 ewah_serialize_strbuf(wd
.valid
, out
);
3273 ewah_serialize_strbuf(wd
.check_only
, out
);
3274 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
3275 strbuf_addbuf(out
, &wd
.sb_stat
);
3276 strbuf_addbuf(out
, &wd
.sb_sha1
);
3277 strbuf_addch(out
, '\0'); /* safe guard for string lists */
3279 ewah_free(wd
.valid
);
3280 ewah_free(wd
.check_only
);
3281 ewah_free(wd
.sha1_valid
);
3282 strbuf_release(&wd
.out
);
3283 strbuf_release(&wd
.sb_stat
);
3284 strbuf_release(&wd
.sb_sha1
);
3287 static void free_untracked(struct untracked_cache_dir
*ucd
)
3292 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
3293 free_untracked(ucd
->dirs
[i
]);
3294 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
3295 free(ucd
->untracked
[i
]);
3296 free(ucd
->untracked
);
3301 void free_untracked_cache(struct untracked_cache
*uc
)
3304 free_untracked(uc
->root
);
3310 struct untracked_cache_dir
**ucd
;
3311 struct ewah_bitmap
*check_only
;
3312 struct ewah_bitmap
*valid
;
3313 struct ewah_bitmap
*sha1_valid
;
3314 const unsigned char *data
;
3315 const unsigned char *end
;
3318 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
3320 memcpy(to
, data
, sizeof(*to
));
3321 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
3322 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
3323 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
3324 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
3325 to
->sd_dev
= ntohl(to
->sd_dev
);
3326 to
->sd_ino
= ntohl(to
->sd_ino
);
3327 to
->sd_uid
= ntohl(to
->sd_uid
);
3328 to
->sd_gid
= ntohl(to
->sd_gid
);
3329 to
->sd_size
= ntohl(to
->sd_size
);
3332 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
3333 struct read_data
*rd
)
3335 struct untracked_cache_dir ud
, *untracked
;
3336 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
3337 const unsigned char *eos
;
3341 memset(&ud
, 0, sizeof(ud
));
3343 value
= decode_varint(&data
);
3347 ud
.untracked_alloc
= value
;
3348 ud
.untracked_nr
= value
;
3349 if (ud
.untracked_nr
)
3350 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
3352 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
3355 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
3357 eos
= memchr(data
, '\0', end
- data
);
3358 if (!eos
|| eos
== end
)
3361 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
3362 memcpy(untracked
, &ud
, sizeof(ud
));
3363 memcpy(untracked
->name
, data
, eos
- data
+ 1);
3366 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
3367 eos
= memchr(data
, '\0', end
- data
);
3368 if (!eos
|| eos
== end
)
3370 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
3374 rd
->ucd
[rd
->index
++] = untracked
;
3377 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
3378 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
3384 static void set_check_only(size_t pos
, void *cb
)
3386 struct read_data
*rd
= cb
;
3387 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3391 static void read_stat(size_t pos
, void *cb
)
3393 struct read_data
*rd
= cb
;
3394 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3395 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
3396 rd
->data
= rd
->end
+ 1;
3399 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
3400 rd
->data
+= sizeof(struct stat_data
);
3404 static void read_oid(size_t pos
, void *cb
)
3406 struct read_data
*rd
= cb
;
3407 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3408 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
3409 rd
->data
= rd
->end
+ 1;
3412 oidread(&ud
->exclude_oid
, rd
->data
);
3413 rd
->data
+= the_hash_algo
->rawsz
;
3416 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
3417 const unsigned char *sha1
)
3419 stat_data_from_disk(&oid_stat
->stat
, data
);
3420 oidread(&oid_stat
->oid
, sha1
);
3421 oid_stat
->valid
= 1;
3424 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
3426 struct untracked_cache
*uc
;
3427 struct read_data rd
;
3428 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
3432 const char *exclude_per_dir
;
3433 const unsigned hashsz
= the_hash_algo
->rawsz
;
3434 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
3435 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
3437 if (sz
<= 1 || end
[-1] != '\0')
3441 ident_len
= decode_varint(&next
);
3442 if (next
+ ident_len
> end
)
3444 ident
= (const char *)next
;
3447 if (next
+ exclude_per_dir_offset
+ 1 > end
)
3450 CALLOC_ARRAY(uc
, 1);
3451 strbuf_init(&uc
->ident
, ident_len
);
3452 strbuf_add(&uc
->ident
, ident
, ident_len
);
3453 load_oid_stat(&uc
->ss_info_exclude
,
3454 next
+ ouc_offset(info_exclude_stat
),
3456 load_oid_stat(&uc
->ss_excludes_file
,
3457 next
+ ouc_offset(excludes_file_stat
),
3458 next
+ offset
+ hashsz
);
3459 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
3460 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
3461 uc
->exclude_per_dir
= xstrdup(exclude_per_dir
);
3462 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3463 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
3467 len
= decode_varint(&next
);
3468 if (next
> end
|| len
== 0)
3471 rd
.valid
= ewah_new();
3472 rd
.check_only
= ewah_new();
3473 rd
.sha1_valid
= ewah_new();
3477 ALLOC_ARRAY(rd
.ucd
, len
);
3479 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
3483 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
3488 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
3493 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
3497 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
3498 rd
.data
= next
+ len
;
3499 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
3500 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
3505 ewah_free(rd
.valid
);
3506 ewah_free(rd
.check_only
);
3507 ewah_free(rd
.sha1_valid
);
3510 free_untracked_cache(uc
);
3516 static void invalidate_one_directory(struct untracked_cache
*uc
,
3517 struct untracked_cache_dir
*ucd
)
3519 uc
->dir_invalidated
++;
3521 ucd
->untracked_nr
= 0;
3525 * Normally when an entry is added or removed from a directory,
3526 * invalidating that directory is enough. No need to touch its
3527 * ancestors. When a directory is shown as "foo/bar/" in git-status
3528 * however, deleting or adding an entry may have cascading effect.
3530 * Say the "foo/bar/file" has become untracked, we need to tell the
3531 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3532 * directory any more (because "bar" is managed by foo as an untracked
3535 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3536 * was the last untracked entry in the entire "foo", we should show
3537 * "foo/" instead. Which means we have to invalidate past "bar" up to
3540 * This function traverses all directories from root to leaf. If there
3541 * is a chance of one of the above cases happening, we invalidate back
3542 * to root. Otherwise we just invalidate the leaf. There may be a more
3543 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3544 * detect these cases and avoid unnecessary invalidation, for example,
3545 * checking for the untracked entry named "bar/" in "foo", but for now
3546 * stick to something safe and simple.
3548 static int invalidate_one_component(struct untracked_cache
*uc
,
3549 struct untracked_cache_dir
*dir
,
3550 const char *path
, int len
)
3552 const char *rest
= strchr(path
, '/');
3555 int component_len
= rest
- path
;
3556 struct untracked_cache_dir
*d
=
3557 lookup_untracked(uc
, dir
, path
, component_len
);
3559 invalidate_one_component(uc
, d
, rest
+ 1,
3560 len
- (component_len
+ 1));
3562 invalidate_one_directory(uc
, dir
);
3566 invalidate_one_directory(uc
, dir
);
3567 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3570 void untracked_cache_invalidate_path(struct index_state
*istate
,
3571 const char *path
, int safe_path
)
3573 if (!istate
->untracked
|| !istate
->untracked
->root
)
3575 if (!safe_path
&& !verify_path(path
, 0))
3577 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3578 path
, strlen(path
));
3581 void untracked_cache_remove_from_index(struct index_state
*istate
,
3584 untracked_cache_invalidate_path(istate
, path
, 1);
3587 void untracked_cache_add_to_index(struct index_state
*istate
,
3590 untracked_cache_invalidate_path(istate
, path
, 1);
3593 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3594 const char *sub_gitdir
)
3597 struct repository subrepo
;
3598 struct strbuf sub_wt
= STRBUF_INIT
;
3599 struct strbuf sub_gd
= STRBUF_INIT
;
3601 const struct submodule
*sub
;
3603 /* If the submodule has no working tree, we can ignore it. */
3604 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3607 if (repo_read_index(&subrepo
) < 0)
3608 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3610 /* TODO: audit for interaction with sparse-index. */
3611 ensure_full_index(subrepo
.index
);
3612 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3613 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3615 if (!S_ISGITLINK(ce
->ce_mode
))
3618 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3619 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3621 * Skip entries with the same name in different stages
3622 * to make sure an entry is returned only once.
3626 sub
= submodule_from_path(&subrepo
, null_oid(), ce
->name
);
3627 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3628 /* .gitmodules broken or inactive sub */
3631 strbuf_reset(&sub_wt
);
3632 strbuf_reset(&sub_gd
);
3633 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3634 strbuf_addf(&sub_gd
, "%s/modules/%s", sub_gitdir
, sub
->name
);
3636 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3638 strbuf_release(&sub_wt
);
3639 strbuf_release(&sub_gd
);
3640 repo_clear(&subrepo
);
3643 void connect_work_tree_and_git_dir(const char *work_tree_
,
3644 const char *git_dir_
,
3645 int recurse_into_nested
)
3647 struct strbuf gitfile_sb
= STRBUF_INIT
;
3648 struct strbuf cfg_sb
= STRBUF_INIT
;
3649 struct strbuf rel_path
= STRBUF_INIT
;
3650 char *git_dir
, *work_tree
;
3652 /* Prepare .git file */
3653 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3654 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3655 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3657 /* Prepare config file */
3658 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3659 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3660 die(_("could not create directories for %s"), cfg_sb
.buf
);
3662 git_dir
= real_pathdup(git_dir_
, 1);
3663 work_tree
= real_pathdup(work_tree_
, 1);
3665 /* Write .git file */
3666 write_file(gitfile_sb
.buf
, "gitdir: %s",
3667 relative_path(git_dir
, work_tree
, &rel_path
));
3668 /* Update core.worktree setting */
3669 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3670 relative_path(work_tree
, git_dir
, &rel_path
));
3672 strbuf_release(&gitfile_sb
);
3673 strbuf_release(&cfg_sb
);
3674 strbuf_release(&rel_path
);
3676 if (recurse_into_nested
)
3677 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
3684 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3686 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
3688 if (rename(old_git_dir
, new_git_dir
) < 0)
3689 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3690 old_git_dir
, new_git_dir
);
3692 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);