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 int count_slashes(const char *s
)
71 int fspathcmp(const char *a
, const char *b
)
73 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
76 int fspathncmp(const char *a
, const char *b
, size_t count
)
78 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
81 int git_fnmatch(const struct pathspec_item
*item
,
82 const char *pattern
, const char *string
,
86 if (ps_strncmp(item
, pattern
, string
, prefix
))
91 if (item
->flags
& PATHSPEC_ONESTAR
) {
92 int pattern_len
= strlen(++pattern
);
93 int string_len
= strlen(string
);
94 return string_len
< pattern_len
||
95 ps_strcmp(item
, pattern
,
96 string
+ string_len
- pattern_len
);
98 if (item
->magic
& PATHSPEC_GLOB
)
99 return wildmatch(pattern
, string
,
101 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0));
103 /* wildmatch has not learned no FNM_PATHNAME mode yet */
104 return wildmatch(pattern
, string
,
105 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0);
108 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
109 const char *string
, int stringlen
,
113 struct strbuf pat_buf
= STRBUF_INIT
;
114 struct strbuf str_buf
= STRBUF_INIT
;
115 const char *use_pat
= pattern
;
116 const char *use_str
= string
;
118 if (pattern
[patternlen
]) {
119 strbuf_add(&pat_buf
, pattern
, patternlen
);
120 use_pat
= pat_buf
.buf
;
122 if (string
[stringlen
]) {
123 strbuf_add(&str_buf
, string
, stringlen
);
124 use_str
= str_buf
.buf
;
128 flags
|= WM_CASEFOLD
;
129 match_status
= wildmatch(use_pat
, use_str
, flags
);
131 strbuf_release(&pat_buf
);
132 strbuf_release(&str_buf
);
137 static size_t common_prefix_len(const struct pathspec
*pathspec
)
143 * ":(icase)path" is treated as a pathspec full of
144 * wildcard. In other words, only prefix is considered common
145 * prefix. If the pathspec is abc/foo abc/bar, running in
146 * subdir xyz, the common prefix is still xyz, not xyz/abc as
149 GUARD_PATHSPEC(pathspec
,
158 for (n
= 0; n
< pathspec
->nr
; n
++) {
159 size_t i
= 0, len
= 0, item_len
;
160 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
162 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
163 item_len
= pathspec
->items
[n
].prefix
;
165 item_len
= pathspec
->items
[n
].nowildcard_len
;
166 while (i
< item_len
&& (n
== 0 || i
< max
)) {
167 char c
= pathspec
->items
[n
].match
[i
];
168 if (c
!= pathspec
->items
[0].match
[i
])
174 if (n
== 0 || len
< max
) {
184 * Returns a copy of the longest leading path common among all
187 char *common_prefix(const struct pathspec
*pathspec
)
189 unsigned long len
= common_prefix_len(pathspec
);
191 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
194 int fill_directory(struct dir_struct
*dir
,
195 struct index_state
*istate
,
196 const struct pathspec
*pathspec
)
201 unsigned exclusive_flags
= DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
;
202 if ((dir
->flags
& exclusive_flags
) == exclusive_flags
)
203 BUG("DIR_SHOW_IGNORED and DIR_SHOW_IGNORED_TOO are exclusive");
206 * Calculate common prefix for the pathspec, and
207 * use that to optimize the directory walk
209 prefix_len
= common_prefix_len(pathspec
);
210 prefix
= prefix_len
? pathspec
->items
[0].match
: "";
212 /* Read the directory and prune it */
213 read_directory(dir
, istate
, prefix
, prefix_len
, pathspec
);
218 int within_depth(const char *name
, int namelen
,
219 int depth
, int max_depth
)
221 const char *cp
= name
, *cpe
= name
+ namelen
;
227 if (depth
> max_depth
)
234 * Read the contents of the blob with the given OID into a buffer.
235 * Append a trailing LF to the end if the last line doesn't have one.
238 * -1 when the OID is invalid or unknown or does not refer to a blob.
239 * 0 when the blob is empty.
240 * 1 along with { data, size } of the (possibly augmented) buffer
243 * Optionally updates the given oid_stat with the given OID (when valid).
245 static int do_read_blob(const struct object_id
*oid
, struct oid_stat
*oid_stat
,
246 size_t *size_out
, char **data_out
)
248 enum object_type type
;
255 data
= read_object_file(oid
, &type
, &sz
);
256 if (!data
|| type
!= OBJ_BLOB
) {
262 memset(&oid_stat
->stat
, 0, sizeof(oid_stat
->stat
));
263 oidcpy(&oid_stat
->oid
, oid
);
271 if (data
[sz
- 1] != '\n') {
272 data
= xrealloc(data
, st_add(sz
, 1));
276 *size_out
= xsize_t(sz
);
282 #define DO_MATCH_EXCLUDE (1<<0)
283 #define DO_MATCH_DIRECTORY (1<<1)
284 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
287 * Does the given pathspec match the given name? A match is found if
289 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
290 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
291 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
292 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
294 * Return value tells which case it was (1-4), or 0 when there is no match.
296 * It may be instructive to look at a small table of concrete examples
297 * to understand the differences between 1, 2, and 4:
300 * | a/b | a/b/ | a/b/c
301 * ------+-----------+-----------+------------
302 * a/b | EXACT | EXACT[1] | LEADING[2]
303 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
304 * a/b/c | RECURSIVE | RECURSIVE | EXACT
306 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
307 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
309 static int match_pathspec_item(const struct index_state
*istate
,
310 const struct pathspec_item
*item
, int prefix
,
311 const char *name
, int namelen
, unsigned flags
)
313 /* name/namelen has prefix cut off by caller */
314 const char *match
= item
->match
+ prefix
;
315 int matchlen
= item
->len
- prefix
;
318 * The normal call pattern is:
319 * 1. prefix = common_prefix_len(ps);
320 * 2. prune something, or fill_directory
321 * 3. match_pathspec()
323 * 'prefix' at #1 may be shorter than the command's prefix and
324 * it's ok for #2 to match extra files. Those extras will be
327 * Suppose the pathspec is 'foo' and '../bar' running from
328 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
329 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
330 * user does not want XYZ/foo, only the "foo" part should be
331 * case-insensitive. We need to filter out XYZ/foo here. In
332 * other words, we do not trust the caller on comparing the
333 * prefix part when :(icase) is involved. We do exact
334 * comparison ourselves.
336 * Normally the caller (common_prefix_len() in fact) does
337 * _exact_ matching on name[-prefix+1..-1] and we do not need
338 * to check that part. Be defensive and check it anyway, in
339 * case common_prefix_len is changed, or a new caller is
340 * introduced that does not use common_prefix_len.
342 * If the penalty turns out too high when prefix is really
343 * long, maybe change it to
344 * strncmp(match, name, item->prefix - prefix)
346 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
347 strncmp(item
->match
, name
- prefix
, item
->prefix
))
350 if (item
->attr_match_nr
&&
351 !match_pathspec_attrs(istate
, name
, namelen
, item
))
354 /* If the match was just the prefix, we matched */
356 return MATCHED_RECURSIVELY
;
358 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
359 if (matchlen
== namelen
)
360 return MATCHED_EXACTLY
;
362 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
363 return MATCHED_RECURSIVELY
;
364 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
365 match
[matchlen
- 1] == '/' &&
366 namelen
== matchlen
- 1 &&
367 !ps_strncmp(item
, match
, name
, namelen
))
368 return MATCHED_EXACTLY
;
370 if (item
->nowildcard_len
< item
->len
&&
371 !git_fnmatch(item
, match
, name
,
372 item
->nowildcard_len
- prefix
))
373 return MATCHED_FNMATCH
;
375 /* Perform checks to see if "name" is a leading string of the pathspec */
376 if ( (flags
& DO_MATCH_LEADING_PATHSPEC
) &&
377 !(flags
& DO_MATCH_EXCLUDE
)) {
378 /* name is a literal prefix of the pathspec */
379 int offset
= name
[namelen
-1] == '/' ? 1 : 0;
380 if ((namelen
< matchlen
) &&
381 (match
[namelen
-offset
] == '/') &&
382 !ps_strncmp(item
, match
, name
, namelen
))
383 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
385 /* name doesn't match up to the first wild character */
386 if (item
->nowildcard_len
< item
->len
&&
387 ps_strncmp(item
, match
, name
,
388 item
->nowildcard_len
- prefix
))
392 * name has no wildcard, and it didn't match as a leading
393 * pathspec so return.
395 if (item
->nowildcard_len
== item
->len
)
399 * Here is where we would perform a wildmatch to check if
400 * "name" can be matched as a directory (or a prefix) against
401 * the pathspec. Since wildmatch doesn't have this capability
402 * at the present we have to punt and say that it is a match,
403 * potentially returning a false positive
404 * The submodules themselves will be able to perform more
405 * accurate matching to determine if the pathspec matches.
407 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
414 * do_match_pathspec() is meant to ONLY be called by
415 * match_pathspec_with_flags(); calling it directly risks pathspecs
416 * like ':!unwanted_path' being ignored.
418 * Given a name and a list of pathspecs, returns the nature of the
419 * closest (i.e. most specific) match of the name to any of the
422 * The caller typically calls this multiple times with the same
423 * pathspec and seen[] array but with different name/namelen
424 * (e.g. entries from the index) and is interested in seeing if and
425 * how each pathspec matches all the names it calls this function
426 * with. A mark is left in the seen[] array for each pathspec element
427 * indicating the closest type of match that element achieved, so if
428 * seen[n] remains zero after multiple invocations, that means the nth
429 * pathspec did not match any names, which could indicate that the
430 * user mistyped the nth pathspec.
432 static int do_match_pathspec(const struct index_state
*istate
,
433 const struct pathspec
*ps
,
434 const char *name
, int namelen
,
435 int prefix
, char *seen
,
438 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
450 if (!ps
->recursive
||
451 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
453 return MATCHED_RECURSIVELY
;
455 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
456 return MATCHED_EXACTLY
;
464 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
467 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
468 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
471 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
474 * Make exclude patterns optional and never report
475 * "pathspec ':(exclude)foo' matches no files"
477 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
478 seen
[i
] = MATCHED_FNMATCH
;
479 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
482 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
483 ps
->max_depth
!= -1 &&
484 how
&& how
!= MATCHED_FNMATCH
) {
485 int len
= ps
->items
[i
].len
;
486 if (name
[len
] == '/')
488 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
489 how
= MATCHED_EXACTLY
;
496 if (seen
&& seen
[i
] < how
)
503 static int match_pathspec_with_flags(const struct index_state
*istate
,
504 const struct pathspec
*ps
,
505 const char *name
, int namelen
,
506 int prefix
, char *seen
, unsigned flags
)
508 int positive
, negative
;
509 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
510 prefix
, seen
, flags
);
511 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
513 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
515 flags
| DO_MATCH_EXCLUDE
);
516 return negative
? 0 : positive
;
519 int match_pathspec(const struct index_state
*istate
,
520 const struct pathspec
*ps
,
521 const char *name
, int namelen
,
522 int prefix
, char *seen
, int is_dir
)
524 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
525 return match_pathspec_with_flags(istate
, ps
, name
, namelen
,
526 prefix
, seen
, flags
);
530 * Check if a submodule is a superset of the pathspec
532 int submodule_path_match(const struct index_state
*istate
,
533 const struct pathspec
*ps
,
534 const char *submodule_name
,
537 int matched
= match_pathspec_with_flags(istate
, ps
, submodule_name
,
538 strlen(submodule_name
),
541 DO_MATCH_LEADING_PATHSPEC
);
545 int report_path_error(const char *ps_matched
,
546 const struct pathspec
*pathspec
)
549 * Make sure all pathspec matched; otherwise it is an error.
552 for (num
= 0; num
< pathspec
->nr
; num
++) {
553 int other
, found_dup
;
558 * The caller might have fed identical pathspec
559 * twice. Do not barf on such a mistake.
560 * FIXME: parse_pathspec should have eliminated
561 * duplicate pathspec.
563 for (found_dup
= other
= 0;
564 !found_dup
&& other
< pathspec
->nr
;
566 if (other
== num
|| !ps_matched
[other
])
568 if (!strcmp(pathspec
->items
[other
].original
,
569 pathspec
->items
[num
].original
))
571 * Ok, we have a match already.
578 error(_("pathspec '%s' did not match any file(s) known to git"),
579 pathspec
->items
[num
].original
);
586 * Return the length of the "simple" part of a path match limiter.
588 int simple_length(const char *match
)
593 unsigned char c
= *match
++;
595 if (c
== '\0' || is_glob_special(c
))
600 int no_wildcard(const char *string
)
602 return string
[simple_length(string
)] == '\0';
605 void parse_path_pattern(const char **pattern
,
610 const char *p
= *pattern
;
615 *flags
|= PATTERN_FLAG_NEGATIVE
;
619 if (len
&& p
[len
- 1] == '/') {
621 *flags
|= PATTERN_FLAG_MUSTBEDIR
;
623 for (i
= 0; i
< len
; i
++) {
628 *flags
|= PATTERN_FLAG_NODIR
;
629 *nowildcardlen
= simple_length(p
);
631 * we should have excluded the trailing slash from 'p' too,
632 * but that's one more allocation. Instead just make sure
633 * nowildcardlen does not exceed real patternlen
635 if (*nowildcardlen
> len
)
636 *nowildcardlen
= len
;
637 if (*p
== '*' && no_wildcard(p
+ 1))
638 *flags
|= PATTERN_FLAG_ENDSWITH
;
643 int pl_hashmap_cmp(const void *unused_cmp_data
,
644 const struct hashmap_entry
*a
,
645 const struct hashmap_entry
*b
,
648 const struct pattern_entry
*ee1
=
649 container_of(a
, struct pattern_entry
, ent
);
650 const struct pattern_entry
*ee2
=
651 container_of(b
, struct pattern_entry
, ent
);
653 size_t min_len
= ee1
->patternlen
<= ee2
->patternlen
658 return strncasecmp(ee1
->pattern
, ee2
->pattern
, min_len
);
659 return strncmp(ee1
->pattern
, ee2
->pattern
, min_len
);
662 static char *dup_and_filter_pattern(const char *pattern
)
666 char *result
= xstrdup(pattern
);
672 /* skip escape characters (once) */
692 static void add_pattern_to_hashsets(struct pattern_list
*pl
, struct path_pattern
*given
)
694 struct pattern_entry
*translated
;
697 const char *prev
, *cur
, *next
;
699 if (!pl
->use_cone_patterns
)
702 if (given
->flags
& PATTERN_FLAG_NEGATIVE
&&
703 given
->flags
& PATTERN_FLAG_MUSTBEDIR
&&
704 !strcmp(given
->pattern
, "/*")) {
709 if (!given
->flags
&& !strcmp(given
->pattern
, "/*")) {
714 if (given
->patternlen
< 2 ||
715 *given
->pattern
== '*' ||
716 strstr(given
->pattern
, "**")) {
717 /* Not a cone pattern. */
718 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
722 prev
= given
->pattern
;
723 cur
= given
->pattern
+ 1;
724 next
= given
->pattern
+ 2;
727 /* Watch for glob characters '*', '\', '[', '?' */
728 if (!is_glob_special(*cur
))
731 /* But only if *prev != '\\' */
735 /* But allow the initial '\' */
737 is_glob_special(*next
))
740 /* But a trailing '/' then '*' is fine */
746 /* Not a cone pattern. */
747 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
756 if (given
->patternlen
> 2 &&
757 !strcmp(given
->pattern
+ given
->patternlen
- 2, "/*")) {
758 if (!(given
->flags
& PATTERN_FLAG_NEGATIVE
)) {
759 /* Not a cone pattern. */
760 warning(_("unrecognized pattern: '%s'"), given
->pattern
);
764 truncated
= dup_and_filter_pattern(given
->pattern
);
766 translated
= xmalloc(sizeof(struct pattern_entry
));
767 translated
->pattern
= truncated
;
768 translated
->patternlen
= given
->patternlen
- 2;
769 hashmap_entry_init(&translated
->ent
,
771 strihash(translated
->pattern
) :
772 strhash(translated
->pattern
));
774 if (!hashmap_get_entry(&pl
->recursive_hashmap
,
775 translated
, ent
, NULL
)) {
776 /* We did not see the "parent" included */
777 warning(_("unrecognized negative pattern: '%s'"),
784 hashmap_add(&pl
->parent_hashmap
, &translated
->ent
);
785 hashmap_remove(&pl
->recursive_hashmap
, &translated
->ent
, &data
);
790 if (given
->flags
& PATTERN_FLAG_NEGATIVE
) {
791 warning(_("unrecognized negative pattern: '%s'"),
796 translated
= xmalloc(sizeof(struct pattern_entry
));
798 translated
->pattern
= dup_and_filter_pattern(given
->pattern
);
799 translated
->patternlen
= given
->patternlen
;
800 hashmap_entry_init(&translated
->ent
,
802 strihash(translated
->pattern
) :
803 strhash(translated
->pattern
));
805 hashmap_add(&pl
->recursive_hashmap
, &translated
->ent
);
807 if (hashmap_get_entry(&pl
->parent_hashmap
, translated
, ent
, NULL
)) {
808 /* we already included this at the parent level */
809 warning(_("your sparse-checkout file may have issues: pattern '%s' is repeated"),
811 hashmap_remove(&pl
->parent_hashmap
, &translated
->ent
, &data
);
819 warning(_("disabling cone pattern matching"));
820 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
821 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
822 pl
->use_cone_patterns
= 0;
825 static int hashmap_contains_path(struct hashmap
*map
,
826 struct strbuf
*pattern
)
828 struct pattern_entry p
;
830 /* Check straight mapping */
831 p
.pattern
= pattern
->buf
;
832 p
.patternlen
= pattern
->len
;
833 hashmap_entry_init(&p
.ent
,
835 strihash(p
.pattern
) :
837 return !!hashmap_get_entry(map
, &p
, ent
, NULL
);
840 int hashmap_contains_parent(struct hashmap
*map
,
842 struct strbuf
*buffer
)
846 strbuf_setlen(buffer
, 0);
849 strbuf_addch(buffer
, '/');
851 strbuf_addstr(buffer
, path
);
853 slash_pos
= strrchr(buffer
->buf
, '/');
855 while (slash_pos
> buffer
->buf
) {
856 strbuf_setlen(buffer
, slash_pos
- buffer
->buf
);
858 if (hashmap_contains_path(map
, buffer
))
861 slash_pos
= strrchr(buffer
->buf
, '/');
867 void add_pattern(const char *string
, const char *base
,
868 int baselen
, struct pattern_list
*pl
, int srcpos
)
870 struct path_pattern
*pattern
;
875 parse_path_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
876 if (flags
& PATTERN_FLAG_MUSTBEDIR
) {
877 FLEXPTR_ALLOC_MEM(pattern
, pattern
, string
, patternlen
);
879 pattern
= xmalloc(sizeof(*pattern
));
880 pattern
->pattern
= string
;
882 pattern
->patternlen
= patternlen
;
883 pattern
->nowildcardlen
= nowildcardlen
;
884 pattern
->base
= base
;
885 pattern
->baselen
= baselen
;
886 pattern
->flags
= flags
;
887 pattern
->srcpos
= srcpos
;
888 ALLOC_GROW(pl
->patterns
, pl
->nr
+ 1, pl
->alloc
);
889 pl
->patterns
[pl
->nr
++] = pattern
;
892 add_pattern_to_hashsets(pl
, pattern
);
895 static int read_skip_worktree_file_from_index(const struct index_state
*istate
,
897 size_t *size_out
, char **data_out
,
898 struct oid_stat
*oid_stat
)
903 pos
= index_name_pos(istate
, path
, len
);
906 if (!ce_skip_worktree(istate
->cache
[pos
]))
909 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
913 * Frees memory within pl which was allocated for exclude patterns and
914 * the file buffer. Does not free pl itself.
916 void clear_pattern_list(struct pattern_list
*pl
)
920 for (i
= 0; i
< pl
->nr
; i
++)
921 free(pl
->patterns
[i
]);
924 hashmap_clear_and_free(&pl
->recursive_hashmap
, struct pattern_entry
, ent
);
925 hashmap_clear_and_free(&pl
->parent_hashmap
, struct pattern_entry
, ent
);
927 memset(pl
, 0, sizeof(*pl
));
930 static void trim_trailing_spaces(char *buf
)
932 char *p
, *last_space
= NULL
;
934 for (p
= buf
; *p
; p
++)
954 * Given a subdirectory name and "dir" of the current directory,
955 * search the subdir in "dir" and return it, or create a new one if it
956 * does not exist in "dir".
958 * If "name" has the trailing slash, it'll be excluded in the search.
960 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
961 struct untracked_cache_dir
*dir
,
962 const char *name
, int len
)
965 struct untracked_cache_dir
*d
;
968 if (len
&& name
[len
- 1] == '/')
972 while (last
> first
) {
973 int cmp
, next
= first
+ ((last
- first
) >> 1);
975 cmp
= strncmp(name
, d
->name
, len
);
976 if (!cmp
&& strlen(d
->name
) > len
)
988 FLEX_ALLOC_MEM(d
, name
, name
, len
);
990 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
991 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
992 dir
->dirs_nr
- first
);
994 dir
->dirs
[first
] = d
;
998 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
1002 dir
->untracked_nr
= 0;
1003 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1004 do_invalidate_gitignore(dir
->dirs
[i
]);
1007 static void invalidate_gitignore(struct untracked_cache
*uc
,
1008 struct untracked_cache_dir
*dir
)
1010 uc
->gitignore_invalidated
++;
1011 do_invalidate_gitignore(dir
);
1014 static void invalidate_directory(struct untracked_cache
*uc
,
1015 struct untracked_cache_dir
*dir
)
1020 * Invalidation increment here is just roughly correct. If
1021 * untracked_nr or any of dirs[].recurse is non-zero, we
1022 * should increment dir_invalidated too. But that's more
1026 uc
->dir_invalidated
++;
1029 dir
->untracked_nr
= 0;
1030 for (i
= 0; i
< dir
->dirs_nr
; i
++)
1031 dir
->dirs
[i
]->recurse
= 0;
1034 static int add_patterns_from_buffer(char *buf
, size_t size
,
1035 const char *base
, int baselen
,
1036 struct pattern_list
*pl
);
1039 * Given a file with name "fname", read it (either from disk, or from
1040 * an index if 'istate' is non-null), parse it and store the
1041 * exclude rules in "pl".
1043 * If "oid_stat" is not NULL, compute oid of the exclude file and fill
1044 * stat data from disk (only valid if add_patterns returns zero). If
1045 * oid_stat.valid is non-zero, "oid_stat" must contain good value as input.
1047 static int add_patterns(const char *fname
, const char *base
, int baselen
,
1048 struct pattern_list
*pl
, struct index_state
*istate
,
1049 struct oid_stat
*oid_stat
)
1057 fd
= open(fname
, O_RDONLY
);
1058 if (fd
< 0 || fstat(fd
, &st
) < 0) {
1060 warn_on_fopen_errors(fname
);
1065 r
= read_skip_worktree_file_from_index(istate
, fname
,
1071 size
= xsize_t(st
.st_size
);
1074 fill_stat_data(&oid_stat
->stat
, &st
);
1075 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
1076 oid_stat
->valid
= 1;
1081 buf
= xmallocz(size
);
1082 if (read_in_full(fd
, buf
, size
) != size
) {
1091 if (oid_stat
->valid
&&
1092 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
1093 ; /* no content change, oid_stat->oid still good */
1095 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
1096 !ce_stage(istate
->cache
[pos
]) &&
1097 ce_uptodate(istate
->cache
[pos
]) &&
1098 !would_convert_to_git(istate
, fname
))
1099 oidcpy(&oid_stat
->oid
,
1100 &istate
->cache
[pos
]->oid
);
1102 hash_object_file(the_hash_algo
, buf
, size
,
1103 "blob", &oid_stat
->oid
);
1104 fill_stat_data(&oid_stat
->stat
, &st
);
1105 oid_stat
->valid
= 1;
1109 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1113 static int add_patterns_from_buffer(char *buf
, size_t size
,
1114 const char *base
, int baselen
,
1115 struct pattern_list
*pl
)
1120 hashmap_init(&pl
->recursive_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1121 hashmap_init(&pl
->parent_hashmap
, pl_hashmap_cmp
, NULL
, 0);
1125 if (skip_utf8_bom(&buf
, size
))
1126 size
-= buf
- pl
->filebuf
;
1130 for (i
= 0; i
< size
; i
++) {
1131 if (buf
[i
] == '\n') {
1132 if (entry
!= buf
+ i
&& entry
[0] != '#') {
1133 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
1134 trim_trailing_spaces(entry
);
1135 add_pattern(entry
, base
, baselen
, pl
, lineno
);
1138 entry
= buf
+ i
+ 1;
1144 int add_patterns_from_file_to_list(const char *fname
, const char *base
,
1145 int baselen
, struct pattern_list
*pl
,
1146 struct index_state
*istate
)
1148 return add_patterns(fname
, base
, baselen
, pl
, istate
, NULL
);
1151 int add_patterns_from_blob_to_list(
1152 struct object_id
*oid
,
1153 const char *base
, int baselen
,
1154 struct pattern_list
*pl
)
1160 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
1164 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
1168 struct pattern_list
*add_pattern_list(struct dir_struct
*dir
,
1169 int group_type
, const char *src
)
1171 struct pattern_list
*pl
;
1172 struct exclude_list_group
*group
;
1174 group
= &dir
->exclude_list_group
[group_type
];
1175 ALLOC_GROW(group
->pl
, group
->nr
+ 1, group
->alloc
);
1176 pl
= &group
->pl
[group
->nr
++];
1177 memset(pl
, 0, sizeof(*pl
));
1183 * Used to set up core.excludesfile and .git/info/exclude lists.
1185 static void add_patterns_from_file_1(struct dir_struct
*dir
, const char *fname
,
1186 struct oid_stat
*oid_stat
)
1188 struct pattern_list
*pl
;
1190 * catch setup_standard_excludes() that's called before
1191 * dir->untracked is assigned. That function behaves
1192 * differently when dir->untracked is non-NULL.
1194 if (!dir
->untracked
)
1195 dir
->unmanaged_exclude_files
++;
1196 pl
= add_pattern_list(dir
, EXC_FILE
, fname
);
1197 if (add_patterns(fname
, "", 0, pl
, NULL
, oid_stat
) < 0)
1198 die(_("cannot use %s as an exclude file"), fname
);
1201 void add_patterns_from_file(struct dir_struct
*dir
, const char *fname
)
1203 dir
->unmanaged_exclude_files
++; /* see validate_untracked_cache() */
1204 add_patterns_from_file_1(dir
, fname
, NULL
);
1207 int match_basename(const char *basename
, int basenamelen
,
1208 const char *pattern
, int prefix
, int patternlen
,
1211 if (prefix
== patternlen
) {
1212 if (patternlen
== basenamelen
&&
1213 !fspathncmp(pattern
, basename
, basenamelen
))
1215 } else if (flags
& PATTERN_FLAG_ENDSWITH
) {
1216 /* "*literal" matching against "fooliteral" */
1217 if (patternlen
- 1 <= basenamelen
&&
1218 !fspathncmp(pattern
+ 1,
1219 basename
+ basenamelen
- (patternlen
- 1),
1223 if (fnmatch_icase_mem(pattern
, patternlen
,
1224 basename
, basenamelen
,
1231 int match_pathname(const char *pathname
, int pathlen
,
1232 const char *base
, int baselen
,
1233 const char *pattern
, int prefix
, int patternlen
,
1240 * match with FNM_PATHNAME; the pattern has base implicitly
1243 if (*pattern
== '/') {
1250 * baselen does not count the trailing slash. base[] may or
1251 * may not end with a trailing slash though.
1253 if (pathlen
< baselen
+ 1 ||
1254 (baselen
&& pathname
[baselen
] != '/') ||
1255 fspathncmp(pathname
, base
, baselen
))
1258 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
1259 name
= pathname
+ pathlen
- namelen
;
1263 * if the non-wildcard part is longer than the
1264 * remaining pathname, surely it cannot match.
1266 if (prefix
> namelen
)
1269 if (fspathncmp(pattern
, name
, prefix
))
1272 patternlen
-= prefix
;
1277 * If the whole pattern did not have a wildcard,
1278 * then our prefix match is all we need; we
1279 * do not need to call fnmatch at all.
1281 if (!patternlen
&& !namelen
)
1285 return fnmatch_icase_mem(pattern
, patternlen
,
1291 * Scan the given exclude list in reverse to see whether pathname
1292 * should be ignored. The first match (i.e. the last on the list), if
1293 * any, determines the fate. Returns the exclude_list element which
1294 * matched, or NULL for undecided.
1296 static struct path_pattern
*last_matching_pattern_from_list(const char *pathname
,
1298 const char *basename
,
1300 struct pattern_list
*pl
,
1301 struct index_state
*istate
)
1303 struct path_pattern
*res
= NULL
; /* undecided */
1307 return NULL
; /* undefined */
1309 for (i
= pl
->nr
- 1; 0 <= i
; i
--) {
1310 struct path_pattern
*pattern
= pl
->patterns
[i
];
1311 const char *exclude
= pattern
->pattern
;
1312 int prefix
= pattern
->nowildcardlen
;
1314 if (pattern
->flags
& PATTERN_FLAG_MUSTBEDIR
) {
1315 *dtype
= resolve_dtype(*dtype
, istate
, pathname
, pathlen
);
1316 if (*dtype
!= DT_DIR
)
1320 if (pattern
->flags
& PATTERN_FLAG_NODIR
) {
1321 if (match_basename(basename
,
1322 pathlen
- (basename
- pathname
),
1323 exclude
, prefix
, pattern
->patternlen
,
1331 assert(pattern
->baselen
== 0 ||
1332 pattern
->base
[pattern
->baselen
- 1] == '/');
1333 if (match_pathname(pathname
, pathlen
,
1335 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1336 exclude
, prefix
, pattern
->patternlen
,
1346 * Scan the list of patterns to determine if the ordered list
1347 * of patterns matches on 'pathname'.
1349 * Return 1 for a match, 0 for not matched and -1 for undecided.
1351 enum pattern_match_result
path_matches_pattern_list(
1352 const char *pathname
, int pathlen
,
1353 const char *basename
, int *dtype
,
1354 struct pattern_list
*pl
,
1355 struct index_state
*istate
)
1357 struct path_pattern
*pattern
;
1358 struct strbuf parent_pathname
= STRBUF_INIT
;
1359 int result
= NOT_MATCHED
;
1360 const char *slash_pos
;
1362 if (!pl
->use_cone_patterns
) {
1363 pattern
= last_matching_pattern_from_list(pathname
, pathlen
, basename
,
1366 if (pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1378 strbuf_addch(&parent_pathname
, '/');
1379 strbuf_add(&parent_pathname
, pathname
, pathlen
);
1381 if (hashmap_contains_path(&pl
->recursive_hashmap
,
1382 &parent_pathname
)) {
1383 result
= MATCHED_RECURSIVE
;
1387 slash_pos
= strrchr(parent_pathname
.buf
, '/');
1389 if (slash_pos
== parent_pathname
.buf
) {
1390 /* include every file in root */
1395 strbuf_setlen(&parent_pathname
, slash_pos
- parent_pathname
.buf
);
1397 if (hashmap_contains_path(&pl
->parent_hashmap
, &parent_pathname
)) {
1402 if (hashmap_contains_parent(&pl
->recursive_hashmap
,
1405 result
= MATCHED_RECURSIVE
;
1408 strbuf_release(&parent_pathname
);
1412 static struct path_pattern
*last_matching_pattern_from_lists(
1413 struct dir_struct
*dir
, struct index_state
*istate
,
1414 const char *pathname
, int pathlen
,
1415 const char *basename
, int *dtype_p
)
1418 struct exclude_list_group
*group
;
1419 struct path_pattern
*pattern
;
1420 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1421 group
= &dir
->exclude_list_group
[i
];
1422 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1423 pattern
= last_matching_pattern_from_list(
1424 pathname
, pathlen
, basename
, dtype_p
,
1425 &group
->pl
[j
], istate
);
1434 * Loads the per-directory exclude list for the substring of base
1435 * which has a char length of baselen.
1437 static void prep_exclude(struct dir_struct
*dir
,
1438 struct index_state
*istate
,
1439 const char *base
, int baselen
)
1441 struct exclude_list_group
*group
;
1442 struct pattern_list
*pl
;
1443 struct exclude_stack
*stk
= NULL
;
1444 struct untracked_cache_dir
*untracked
;
1447 group
= &dir
->exclude_list_group
[EXC_DIRS
];
1450 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1451 * which originate from directories not in the prefix of the
1452 * path being checked.
1454 while ((stk
= dir
->exclude_stack
) != NULL
) {
1455 if (stk
->baselen
<= baselen
&&
1456 !strncmp(dir
->basebuf
.buf
, base
, stk
->baselen
))
1458 pl
= &group
->pl
[dir
->exclude_stack
->exclude_ix
];
1459 dir
->exclude_stack
= stk
->prev
;
1460 dir
->pattern
= NULL
;
1461 free((char *)pl
->src
); /* see strbuf_detach() below */
1462 clear_pattern_list(pl
);
1467 /* Skip traversing into sub directories if the parent is excluded */
1472 * Lazy initialization. All call sites currently just
1473 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1474 * them seems lots of work for little benefit.
1476 if (!dir
->basebuf
.buf
)
1477 strbuf_init(&dir
->basebuf
, PATH_MAX
);
1479 /* Read from the parent directories and push them down. */
1480 current
= stk
? stk
->baselen
: -1;
1481 strbuf_setlen(&dir
->basebuf
, current
< 0 ? 0 : current
);
1483 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1487 while (current
< baselen
) {
1489 struct oid_stat oid_stat
;
1491 CALLOC_ARRAY(stk
, 1);
1496 cp
= strchr(base
+ current
+ 1, '/');
1498 die("oops in prep_exclude");
1501 lookup_untracked(dir
->untracked
, untracked
,
1503 cp
- base
- current
);
1505 stk
->prev
= dir
->exclude_stack
;
1506 stk
->baselen
= cp
- base
;
1507 stk
->exclude_ix
= group
->nr
;
1508 stk
->ucd
= untracked
;
1509 pl
= add_pattern_list(dir
, EXC_DIRS
, NULL
);
1510 strbuf_add(&dir
->basebuf
, base
+ current
, stk
->baselen
- current
);
1511 assert(stk
->baselen
== dir
->basebuf
.len
);
1513 /* Abort if the directory is excluded */
1516 dir
->basebuf
.buf
[stk
->baselen
- 1] = 0;
1517 dir
->pattern
= last_matching_pattern_from_lists(dir
,
1519 dir
->basebuf
.buf
, stk
->baselen
- 1,
1520 dir
->basebuf
.buf
+ current
, &dt
);
1521 dir
->basebuf
.buf
[stk
->baselen
- 1] = '/';
1523 dir
->pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1524 dir
->pattern
= NULL
;
1526 dir
->exclude_stack
= stk
;
1531 /* Try to read per-directory file */
1532 oidclr(&oid_stat
.oid
);
1534 if (dir
->exclude_per_dir
&&
1536 * If we know that no files have been added in
1537 * this directory (i.e. valid_cached_dir() has
1538 * been executed and set untracked->valid) ..
1540 (!untracked
|| !untracked
->valid
||
1542 * .. and .gitignore does not exist before
1543 * (i.e. null exclude_oid). Then we can skip
1544 * loading .gitignore, which would result in
1547 !is_null_oid(&untracked
->exclude_oid
))) {
1549 * dir->basebuf gets reused by the traversal, but we
1550 * need fname to remain unchanged to ensure the src
1551 * member of each struct path_pattern correctly
1552 * back-references its source file. Other invocations
1553 * of add_pattern_list provide stable strings, so we
1554 * strbuf_detach() and free() here in the caller.
1556 struct strbuf sb
= STRBUF_INIT
;
1557 strbuf_addbuf(&sb
, &dir
->basebuf
);
1558 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1559 pl
->src
= strbuf_detach(&sb
, NULL
);
1560 add_patterns(pl
->src
, pl
->src
, stk
->baselen
, pl
, istate
,
1561 untracked
? &oid_stat
: NULL
);
1564 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1565 * will first be called in valid_cached_dir() then maybe many
1566 * times more in last_matching_pattern(). When the cache is
1567 * used, last_matching_pattern() will not be called and
1568 * reading .gitignore content will be a waste.
1570 * So when it's called by valid_cached_dir() and we can get
1571 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1572 * modified on work tree), we could delay reading the
1573 * .gitignore content until we absolutely need it in
1574 * last_matching_pattern(). Be careful about ignore rule
1575 * order, though, if you do that.
1578 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1579 invalidate_gitignore(dir
->untracked
, untracked
);
1580 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1582 dir
->exclude_stack
= stk
;
1583 current
= stk
->baselen
;
1585 strbuf_setlen(&dir
->basebuf
, baselen
);
1589 * Loads the exclude lists for the directory containing pathname, then
1590 * scans all exclude lists to determine whether pathname is excluded.
1591 * Returns the exclude_list element which matched, or NULL for
1594 struct path_pattern
*last_matching_pattern(struct dir_struct
*dir
,
1595 struct index_state
*istate
,
1596 const char *pathname
,
1599 int pathlen
= strlen(pathname
);
1600 const char *basename
= strrchr(pathname
, '/');
1601 basename
= (basename
) ? basename
+1 : pathname
;
1603 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1606 return dir
->pattern
;
1608 return last_matching_pattern_from_lists(dir
, istate
, pathname
, pathlen
,
1613 * Loads the exclude lists for the directory containing pathname, then
1614 * scans all exclude lists to determine whether pathname is excluded.
1615 * Returns 1 if true, otherwise 0.
1617 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1618 const char *pathname
, int *dtype_p
)
1620 struct path_pattern
*pattern
=
1621 last_matching_pattern(dir
, istate
, pathname
, dtype_p
);
1623 return pattern
->flags
& PATTERN_FLAG_NEGATIVE
? 0 : 1;
1627 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1629 struct dir_entry
*ent
;
1631 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1636 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1637 struct index_state
*istate
,
1638 const char *pathname
, int len
)
1640 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1643 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->alloc
);
1644 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1647 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1648 struct index_state
*istate
,
1649 const char *pathname
, int len
)
1651 if (!index_name_is_other(istate
, pathname
, len
))
1654 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->ignored_alloc
);
1655 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1659 index_nonexistent
= 0,
1665 * Do not use the alphabetically sorted index to look up
1666 * the directory name; instead, use the case insensitive
1669 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1670 const char *dirname
, int len
)
1672 struct cache_entry
*ce
;
1674 if (index_dir_exists(istate
, dirname
, len
))
1675 return index_directory
;
1677 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1678 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1679 return index_gitdir
;
1681 return index_nonexistent
;
1685 * The index sorts alphabetically by entry name, which
1686 * means that a gitlink sorts as '\0' at the end, while
1687 * a directory (which is defined not as an entry, but as
1688 * the files it contains) will sort with the '/' at the
1691 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1692 const char *dirname
, int len
)
1697 return directory_exists_in_index_icase(istate
, dirname
, len
);
1699 pos
= index_name_pos(istate
, dirname
, len
);
1702 while (pos
< istate
->cache_nr
) {
1703 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1704 unsigned char endchar
;
1706 if (strncmp(ce
->name
, dirname
, len
))
1708 endchar
= ce
->name
[len
];
1712 return index_directory
;
1713 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1714 return index_gitdir
;
1716 return index_nonexistent
;
1720 * When we find a directory when traversing the filesystem, we
1721 * have three distinct cases:
1724 * - see it as a directory
1727 * and which one we choose depends on a combination of existing
1728 * git index contents and the flags passed into the directory
1729 * traversal routine.
1731 * Case 1: If we *already* have entries in the index under that
1732 * directory name, we always recurse into the directory to see
1735 * Case 2: If we *already* have that directory name as a gitlink,
1736 * we always continue to see it as a gitlink, regardless of whether
1737 * there is an actual git directory there or not (it might not
1738 * be checked out as a subproject!)
1740 * Case 3: if we didn't have it in the index previously, we
1741 * have a few sub-cases:
1743 * (a) if "show_other_directories" is true, we show it as
1744 * just a directory, unless "hide_empty_directories" is
1745 * also true, in which case we need to check if it contains any
1746 * untracked and / or ignored files.
1747 * (b) if it looks like a git directory, and we don't have
1748 * 'no_gitlinks' set we treat it as a gitlink, and show it
1750 * (c) otherwise, we recurse into it.
1752 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1753 struct index_state
*istate
,
1754 struct untracked_cache_dir
*untracked
,
1755 const char *dirname
, int len
, int baselen
, int excluded
,
1756 const struct pathspec
*pathspec
)
1759 * WARNING: From this function, you can return path_recurse or you
1760 * can call read_directory_recursive() (or neither), but
1761 * you CAN'T DO BOTH.
1763 enum path_treatment state
;
1764 int matches_how
= 0;
1765 int nested_repo
= 0, check_only
, stop_early
;
1766 int old_ignored_nr
, old_untracked_nr
;
1767 /* The "len-1" is to strip the final '/' */
1768 enum exist_status status
= directory_exists_in_index(istate
, dirname
, len
-1);
1770 if (status
== index_directory
)
1771 return path_recurse
;
1772 if (status
== index_gitdir
)
1774 if (status
!= index_nonexistent
)
1775 BUG("Unhandled value for directory_exists_in_index: %d\n", status
);
1778 * We don't want to descend into paths that don't match the necessary
1779 * patterns. Clearly, if we don't have a pathspec, then we can't check
1780 * for matching patterns. Also, if (excluded) then we know we matched
1781 * the exclusion patterns so as an optimization we can skip checking
1782 * for matching patterns.
1784 if (pathspec
&& !excluded
) {
1785 matches_how
= match_pathspec_with_flags(istate
, pathspec
,
1789 DO_MATCH_LEADING_PATHSPEC
);
1795 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1796 !(dir
->flags
& DIR_NO_GITLINKS
)) {
1797 struct strbuf sb
= STRBUF_INIT
;
1798 strbuf_addstr(&sb
, dirname
);
1799 nested_repo
= is_nonbare_repository_dir(&sb
);
1800 strbuf_release(&sb
);
1803 if ((dir
->flags
& DIR_SKIP_NESTED_GIT
) ||
1804 (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
))
1806 return excluded
? path_excluded
: path_untracked
;
1809 if (!(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)) {
1811 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1812 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1815 * This is an excluded directory and we are
1816 * showing ignored paths that match an exclude
1817 * pattern. (e.g. show directory as ignored
1818 * only if it matches an exclude pattern).
1819 * This path will either be 'path_excluded`
1820 * (if we are showing empty directories or if
1821 * the directory is not empty), or will be
1822 * 'path_none' (empty directory, and we are
1823 * not showing empty directories).
1825 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1826 return path_excluded
;
1828 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1829 untracked
, 1, 1, pathspec
) == path_excluded
)
1830 return path_excluded
;
1834 return path_recurse
;
1837 /* This is the "show_other_directories" case */
1840 * If we have a pathspec which could match something _below_ this
1841 * directory (e.g. when checking 'subdir/' having a pathspec like
1842 * 'subdir/some/deep/path/file' or 'subdir/widget-*.c'), then we
1845 if (matches_how
== MATCHED_RECURSIVELY_LEADING_PATHSPEC
)
1846 return path_recurse
;
1849 * Other than the path_recurse case immediately above, we only need
1850 * to recurse into untracked/ignored directories if either of the
1851 * following bits is set:
1852 * - DIR_SHOW_IGNORED_TOO (because then we need to determine if
1853 * there are ignored entries below)
1854 * - DIR_HIDE_EMPTY_DIRECTORIES (because we have to determine if
1855 * the directory is empty)
1857 if (!(dir
->flags
& (DIR_SHOW_IGNORED_TOO
| DIR_HIDE_EMPTY_DIRECTORIES
)))
1858 return excluded
? path_excluded
: path_untracked
;
1861 * ...and even if DIR_SHOW_IGNORED_TOO is set, we can still avoid
1862 * recursing into ignored directories if the path is excluded and
1863 * DIR_SHOW_IGNORED_TOO_MODE_MATCHING is also set.
1866 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1867 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
1868 return path_excluded
;
1871 * Even if we don't want to know all the paths under an untracked or
1872 * ignored directory, we may still need to go into the directory to
1873 * determine if it is empty (because with DIR_HIDE_EMPTY_DIRECTORIES,
1874 * an empty directory should be path_none instead of path_excluded or
1877 check_only
= ((dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
) &&
1878 !(dir
->flags
& DIR_SHOW_IGNORED_TOO
));
1881 * However, there's another optimization possible as a subset of
1882 * check_only, based on the cases we have to consider:
1883 * A) Directory matches no exclude patterns:
1884 * * Directory is empty => path_none
1885 * * Directory has an untracked file under it => path_untracked
1886 * * Directory has only ignored files under it => path_excluded
1887 * B) Directory matches an exclude pattern:
1888 * * Directory is empty => path_none
1889 * * Directory has an untracked file under it => path_excluded
1890 * * Directory has only ignored files under it => path_excluded
1891 * In case A, we can exit as soon as we've found an untracked
1892 * file but otherwise have to walk all files. In case B, though,
1893 * we can stop at the first file we find under the directory.
1895 stop_early
= check_only
&& excluded
;
1898 * If /every/ file within an untracked directory is ignored, then
1899 * we want to treat the directory as ignored (for e.g. status
1900 * --porcelain), without listing the individual ignored files
1901 * underneath. To do so, we'll save the current ignored_nr, and
1902 * pop all the ones added after it if it turns out the entire
1903 * directory is ignored. Also, when DIR_SHOW_IGNORED_TOO and
1904 * !DIR_KEEP_UNTRACKED_CONTENTS then we don't want to show
1905 * untracked paths so will need to pop all those off the last
1906 * after we traverse.
1908 old_ignored_nr
= dir
->ignored_nr
;
1909 old_untracked_nr
= dir
->nr
;
1911 /* Actually recurse into dirname now, we'll fixup the state later. */
1912 untracked
= lookup_untracked(dir
->untracked
, untracked
,
1913 dirname
+ baselen
, len
- baselen
);
1914 state
= read_directory_recursive(dir
, istate
, dirname
, len
, untracked
,
1915 check_only
, stop_early
, pathspec
);
1917 /* There are a variety of reasons we may need to fixup the state... */
1918 if (state
== path_excluded
) {
1919 /* state == path_excluded implies all paths under
1920 * dirname were ignored...
1922 * if running e.g. `git status --porcelain --ignored=matching`,
1923 * then we want to see the subpaths that are ignored.
1925 * if running e.g. just `git status --porcelain`, then
1926 * we just want the directory itself to be listed as ignored
1927 * and not the individual paths underneath.
1929 int want_ignored_subpaths
=
1930 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1931 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
));
1933 if (want_ignored_subpaths
) {
1935 * with --ignored=matching, we want the subpaths
1936 * INSTEAD of the directory itself.
1941 for (i
= old_ignored_nr
+ 1; i
<dir
->ignored_nr
; ++i
)
1942 FREE_AND_NULL(dir
->ignored
[i
]);
1943 dir
->ignored_nr
= old_ignored_nr
;
1948 * We may need to ignore some of the untracked paths we found while
1949 * traversing subdirectories.
1951 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1952 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
1954 for (i
= old_untracked_nr
+ 1; i
<dir
->nr
; ++i
)
1955 FREE_AND_NULL(dir
->entries
[i
]);
1956 dir
->nr
= old_untracked_nr
;
1960 * If there is nothing under the current directory and we are not
1961 * hiding empty directories, then we need to report on the
1962 * untracked or ignored status of the directory itself.
1964 if (state
== path_none
&& !(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1965 state
= excluded
? path_excluded
: path_untracked
;
1971 * This is an inexact early pruning of any recursive directory
1972 * reading - if the path cannot possibly be in the pathspec,
1973 * return true, and we'll skip it early.
1975 static int simplify_away(const char *path
, int pathlen
,
1976 const struct pathspec
*pathspec
)
1980 if (!pathspec
|| !pathspec
->nr
)
1983 GUARD_PATHSPEC(pathspec
,
1992 for (i
= 0; i
< pathspec
->nr
; i
++) {
1993 const struct pathspec_item
*item
= &pathspec
->items
[i
];
1994 int len
= item
->nowildcard_len
;
1998 if (!ps_strncmp(item
, item
->match
, path
, len
))
2006 * This function tells us whether an excluded path matches a
2007 * list of "interesting" pathspecs. That is, whether a path matched
2008 * by any of the pathspecs could possibly be ignored by excluding
2009 * the specified path. This can happen if:
2011 * 1. the path is mentioned explicitly in the pathspec
2013 * 2. the path is a directory prefix of some element in the
2016 static int exclude_matches_pathspec(const char *path
, int pathlen
,
2017 const struct pathspec
*pathspec
)
2021 if (!pathspec
|| !pathspec
->nr
)
2024 GUARD_PATHSPEC(pathspec
,
2032 for (i
= 0; i
< pathspec
->nr
; i
++) {
2033 const struct pathspec_item
*item
= &pathspec
->items
[i
];
2034 int len
= item
->nowildcard_len
;
2036 if (len
== pathlen
&&
2037 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2039 if (len
> pathlen
&&
2040 item
->match
[pathlen
] == '/' &&
2041 !ps_strncmp(item
, item
->match
, path
, pathlen
))
2047 static int get_index_dtype(struct index_state
*istate
,
2048 const char *path
, int len
)
2051 const struct cache_entry
*ce
;
2053 ce
= index_file_exists(istate
, path
, len
, 0);
2055 if (!ce_uptodate(ce
))
2057 if (S_ISGITLINK(ce
->ce_mode
))
2060 * Nobody actually cares about the
2061 * difference between DT_LNK and DT_REG
2066 /* Try to look it up as a directory */
2067 pos
= index_name_pos(istate
, path
, len
);
2071 while (pos
< istate
->cache_nr
) {
2072 ce
= istate
->cache
[pos
++];
2073 if (strncmp(ce
->name
, path
, len
))
2075 if (ce
->name
[len
] > '/')
2077 if (ce
->name
[len
] < '/')
2079 if (!ce_uptodate(ce
))
2080 break; /* continue? */
2086 static int resolve_dtype(int dtype
, struct index_state
*istate
,
2087 const char *path
, int len
)
2091 if (dtype
!= DT_UNKNOWN
)
2093 dtype
= get_index_dtype(istate
, path
, len
);
2094 if (dtype
!= DT_UNKNOWN
)
2096 if (lstat(path
, &st
))
2098 if (S_ISREG(st
.st_mode
))
2100 if (S_ISDIR(st
.st_mode
))
2102 if (S_ISLNK(st
.st_mode
))
2107 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
2108 struct cached_dir
*cdir
,
2109 struct index_state
*istate
,
2110 struct strbuf
*path
,
2112 const struct pathspec
*pathspec
)
2115 * WARNING: From this function, you can return path_recurse or you
2116 * can call read_directory_recursive() (or neither), but
2117 * you CAN'T DO BOTH.
2119 strbuf_setlen(path
, baselen
);
2121 strbuf_addstr(path
, cdir
->file
);
2122 return path_untracked
;
2124 strbuf_addstr(path
, cdir
->ucd
->name
);
2125 /* treat_one_path() does this before it calls treat_directory() */
2126 strbuf_complete(path
, '/');
2127 if (cdir
->ucd
->check_only
)
2129 * check_only is set as a result of treat_directory() getting
2130 * to its bottom. Verify again the same set of directories
2131 * with check_only set.
2133 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
2134 cdir
->ucd
, 1, 0, pathspec
);
2136 * We get path_recurse in the first run when
2137 * directory_exists_in_index() returns index_nonexistent. We
2138 * are sure that new changes in the index does not impact the
2139 * outcome. Return now.
2141 return path_recurse
;
2144 static enum path_treatment
treat_path(struct dir_struct
*dir
,
2145 struct untracked_cache_dir
*untracked
,
2146 struct cached_dir
*cdir
,
2147 struct index_state
*istate
,
2148 struct strbuf
*path
,
2150 const struct pathspec
*pathspec
)
2152 int has_path_in_index
, dtype
, excluded
;
2155 return treat_path_fast(dir
, cdir
, istate
, path
,
2157 if (is_dot_or_dotdot(cdir
->d_name
) || !fspathcmp(cdir
->d_name
, ".git"))
2159 strbuf_setlen(path
, baselen
);
2160 strbuf_addstr(path
, cdir
->d_name
);
2161 if (simplify_away(path
->buf
, path
->len
, pathspec
))
2164 dtype
= resolve_dtype(cdir
->d_type
, istate
, path
->buf
, path
->len
);
2166 /* Always exclude indexed files */
2167 has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
,
2169 if (dtype
!= DT_DIR
&& has_path_in_index
)
2173 * When we are looking at a directory P in the working tree,
2174 * there are three cases:
2176 * (1) P exists in the index. Everything inside the directory P in
2177 * the working tree needs to go when P is checked out from the
2180 * (2) P does not exist in the index, but there is P/Q in the index.
2181 * We know P will stay a directory when we check out the contents
2182 * of the index, but we do not know yet if there is a directory
2183 * P/Q in the working tree to be killed, so we need to recurse.
2185 * (3) P does not exist in the index, and there is no P/Q in the index
2186 * to require P to be a directory, either. Only in this case, we
2187 * know that everything inside P will not be killed without
2190 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
2191 (dtype
== DT_DIR
) &&
2192 !has_path_in_index
&&
2193 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
2196 excluded
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
2199 * Excluded? If we don't explicitly want to show
2200 * ignored files, ignore it
2202 if (excluded
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
2203 return path_excluded
;
2210 * WARNING: Do not ignore/amend the return value from
2211 * treat_directory(), and especially do not change it to return
2212 * path_recurse as that can cause exponential slowdown.
2213 * Instead, modify treat_directory() to return the right value.
2215 strbuf_addch(path
, '/');
2216 return treat_directory(dir
, istate
, untracked
,
2217 path
->buf
, path
->len
,
2218 baselen
, excluded
, pathspec
);
2222 !match_pathspec(istate
, pathspec
, path
->buf
, path
->len
,
2223 0 /* prefix */, NULL
/* seen */,
2227 return path_excluded
;
2228 return path_untracked
;
2232 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
2236 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
2237 dir
->untracked_alloc
);
2238 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
2241 static int valid_cached_dir(struct dir_struct
*dir
,
2242 struct untracked_cache_dir
*untracked
,
2243 struct index_state
*istate
,
2244 struct strbuf
*path
,
2253 * With fsmonitor, we can trust the untracked cache's valid field.
2255 refresh_fsmonitor(istate
);
2256 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
2257 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
2258 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
2261 if (!untracked
->valid
||
2262 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
2263 fill_stat_data(&untracked
->stat_data
, &st
);
2268 if (untracked
->check_only
!= !!check_only
)
2272 * prep_exclude will be called eventually on this directory,
2273 * but it's called much later in last_matching_pattern(). We
2274 * need it now to determine the validity of the cache for this
2275 * path. The next calls will be nearly no-op, the way
2276 * prep_exclude() is designed.
2278 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
2279 strbuf_addch(path
, '/');
2280 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2281 strbuf_setlen(path
, path
->len
- 1);
2283 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
2285 /* hopefully prep_exclude() haven't invalidated this entry... */
2286 return untracked
->valid
;
2289 static int open_cached_dir(struct cached_dir
*cdir
,
2290 struct dir_struct
*dir
,
2291 struct untracked_cache_dir
*untracked
,
2292 struct index_state
*istate
,
2293 struct strbuf
*path
,
2298 memset(cdir
, 0, sizeof(*cdir
));
2299 cdir
->untracked
= untracked
;
2300 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
2302 c_path
= path
->len
? path
->buf
: ".";
2303 cdir
->fdir
= opendir(c_path
);
2305 warning_errno(_("could not open directory '%s'"), c_path
);
2306 if (dir
->untracked
) {
2307 invalidate_directory(dir
->untracked
, untracked
);
2308 dir
->untracked
->dir_opened
++;
2315 static int read_cached_dir(struct cached_dir
*cdir
)
2320 de
= readdir(cdir
->fdir
);
2322 cdir
->d_name
= NULL
;
2323 cdir
->d_type
= DT_UNKNOWN
;
2326 cdir
->d_name
= de
->d_name
;
2327 cdir
->d_type
= DTYPE(de
);
2330 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
2331 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
2341 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
2342 struct untracked_cache_dir
*d
= cdir
->untracked
;
2343 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
2349 static void close_cached_dir(struct cached_dir
*cdir
)
2352 closedir(cdir
->fdir
);
2354 * We have gone through this directory and found no untracked
2355 * entries. Mark it valid.
2357 if (cdir
->untracked
) {
2358 cdir
->untracked
->valid
= 1;
2359 cdir
->untracked
->recurse
= 1;
2363 static void add_path_to_appropriate_result_list(struct dir_struct
*dir
,
2364 struct untracked_cache_dir
*untracked
,
2365 struct cached_dir
*cdir
,
2366 struct index_state
*istate
,
2367 struct strbuf
*path
,
2369 const struct pathspec
*pathspec
,
2370 enum path_treatment state
)
2372 /* add the path to the appropriate result list */
2375 if (dir
->flags
& DIR_SHOW_IGNORED
)
2376 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2377 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2378 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
2379 exclude_matches_pathspec(path
->buf
, path
->len
,
2381 dir_add_ignored(dir
, istate
, path
->buf
, path
->len
);
2384 case path_untracked
:
2385 if (dir
->flags
& DIR_SHOW_IGNORED
)
2387 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
2389 add_untracked(untracked
, path
->buf
+ baselen
);
2398 * Read a directory tree. We currently ignore anything but
2399 * directories, regular files and symlinks. That's because git
2400 * doesn't handle them at all yet. Maybe that will change some
2403 * Also, we ignore the name ".git" (even if it is not a directory).
2404 * That likely will not change.
2406 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
2407 * to signal that a file was found. This is the least significant value that
2408 * indicates that a file was encountered that does not depend on the order of
2409 * whether an untracked or excluded path was encountered first.
2411 * Returns the most significant path_treatment value encountered in the scan.
2412 * If 'stop_at_first_file' is specified, `path_excluded` is the most
2413 * significant path_treatment value that will be returned.
2416 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
2417 struct index_state
*istate
, const char *base
, int baselen
,
2418 struct untracked_cache_dir
*untracked
, int check_only
,
2419 int stop_at_first_file
, const struct pathspec
*pathspec
)
2422 * WARNING: Do NOT recurse unless path_recurse is returned from
2423 * treat_path(). Recursing on any other return value
2424 * can result in exponential slowdown.
2426 struct cached_dir cdir
;
2427 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
2428 struct strbuf path
= STRBUF_INIT
;
2430 strbuf_add(&path
, base
, baselen
);
2432 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
2436 untracked
->check_only
= !!check_only
;
2438 while (!read_cached_dir(&cdir
)) {
2439 /* check how the file or directory should be treated */
2440 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
2443 if (state
> dir_state
)
2446 /* recurse into subdir if instructed by treat_path */
2447 if (state
== path_recurse
) {
2448 struct untracked_cache_dir
*ud
;
2449 ud
= lookup_untracked(dir
->untracked
, untracked
,
2451 path
.len
- baselen
);
2453 read_directory_recursive(dir
, istate
, path
.buf
,
2455 check_only
, stop_at_first_file
, pathspec
);
2456 if (subdir_state
> dir_state
)
2457 dir_state
= subdir_state
;
2460 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2461 0 /* prefix */, NULL
,
2462 0 /* do NOT special case dirs */))
2467 if (stop_at_first_file
) {
2469 * If stopping at first file, then
2470 * signal that a file was found by
2471 * returning `path_excluded`. This is
2472 * to return a consistent value
2473 * regardless of whether an ignored or
2474 * excluded file happened to be
2477 * In current usage, the
2478 * `stop_at_first_file` is passed when
2479 * an ancestor directory has matched
2480 * an exclude pattern, so any found
2481 * files will be excluded.
2483 if (dir_state
>= path_excluded
) {
2484 dir_state
= path_excluded
;
2489 /* abort early if maximum state has been reached */
2490 if (dir_state
== path_untracked
) {
2492 add_untracked(untracked
, path
.buf
+ baselen
);
2495 /* skip the add_path_to_appropriate_result_list() */
2499 add_path_to_appropriate_result_list(dir
, untracked
, &cdir
,
2500 istate
, &path
, baselen
,
2503 close_cached_dir(&cdir
);
2505 strbuf_release(&path
);
2510 int cmp_dir_entry(const void *p1
, const void *p2
)
2512 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2513 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2515 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2518 /* check if *out lexically strictly contains *in */
2519 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2521 return (out
->len
< in
->len
) &&
2522 (out
->name
[out
->len
- 1] == '/') &&
2523 !memcmp(out
->name
, in
->name
, out
->len
);
2526 static int treat_leading_path(struct dir_struct
*dir
,
2527 struct index_state
*istate
,
2528 const char *path
, int len
,
2529 const struct pathspec
*pathspec
)
2531 struct strbuf sb
= STRBUF_INIT
;
2532 struct strbuf subdir
= STRBUF_INIT
;
2533 int prevlen
, baselen
;
2535 struct cached_dir cdir
;
2536 enum path_treatment state
= path_none
;
2539 * For each directory component of path, we are going to check whether
2540 * that path is relevant given the pathspec. For example, if path is
2542 * then we will ask treat_path() whether we should go into foo, then
2543 * whether we should go into bar, then whether baz is relevant.
2544 * Checking each is important because e.g. if path is
2546 * then we need to check .git to know we shouldn't traverse it.
2547 * If the return from treat_path() is:
2548 * * path_none, for any path, we return false.
2549 * * path_recurse, for all path components, we return true
2550 * * <anything else> for some intermediate component, we make sure
2551 * to add that path to the relevant list but return false
2552 * signifying that we shouldn't recurse into it.
2555 while (len
&& path
[len
- 1] == '/')
2560 memset(&cdir
, 0, sizeof(cdir
));
2561 cdir
.d_type
= DT_DIR
;
2565 prevlen
= baselen
+ !!baselen
;
2566 cp
= path
+ prevlen
;
2567 cp
= memchr(cp
, '/', path
+ len
- cp
);
2571 baselen
= cp
- path
;
2573 strbuf_add(&sb
, path
, baselen
);
2574 if (!is_directory(sb
.buf
))
2577 strbuf_add(&sb
, path
, prevlen
);
2578 strbuf_reset(&subdir
);
2579 strbuf_add(&subdir
, path
+prevlen
, baselen
-prevlen
);
2580 cdir
.d_name
= subdir
.buf
;
2581 state
= treat_path(dir
, NULL
, &cdir
, istate
, &sb
, prevlen
, pathspec
);
2583 if (state
!= path_recurse
)
2584 break; /* do not recurse into it */
2586 break; /* finished checking */
2588 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
, istate
,
2589 &sb
, baselen
, pathspec
,
2592 strbuf_release(&subdir
);
2593 strbuf_release(&sb
);
2594 return state
== path_recurse
;
2597 static const char *get_ident_string(void)
2599 static struct strbuf sb
= STRBUF_INIT
;
2604 if (uname(&uts
) < 0)
2605 die_errno(_("failed to get kernel name and information"));
2606 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2611 static int ident_in_untracked(const struct untracked_cache
*uc
)
2614 * Previous git versions may have saved many NUL separated
2615 * strings in the "ident" field, but it is insane to manage
2616 * many locations, so just take care of the first one.
2619 return !strcmp(uc
->ident
.buf
, get_ident_string());
2622 static void set_untracked_ident(struct untracked_cache
*uc
)
2624 strbuf_reset(&uc
->ident
);
2625 strbuf_addstr(&uc
->ident
, get_ident_string());
2628 * This strbuf used to contain a list of NUL separated
2629 * strings, so save NUL too for backward compatibility.
2631 strbuf_addch(&uc
->ident
, 0);
2634 static void new_untracked_cache(struct index_state
*istate
)
2636 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2637 strbuf_init(&uc
->ident
, 100);
2638 uc
->exclude_per_dir
= ".gitignore";
2639 /* should be the same flags used by git-status */
2640 uc
->dir_flags
= DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2641 set_untracked_ident(uc
);
2642 istate
->untracked
= uc
;
2643 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2646 void add_untracked_cache(struct index_state
*istate
)
2648 if (!istate
->untracked
) {
2649 new_untracked_cache(istate
);
2651 if (!ident_in_untracked(istate
->untracked
)) {
2652 free_untracked_cache(istate
->untracked
);
2653 new_untracked_cache(istate
);
2658 void remove_untracked_cache(struct index_state
*istate
)
2660 if (istate
->untracked
) {
2661 free_untracked_cache(istate
->untracked
);
2662 istate
->untracked
= NULL
;
2663 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2667 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2669 const struct pathspec
*pathspec
)
2671 struct untracked_cache_dir
*root
;
2672 static int untracked_cache_disabled
= -1;
2674 if (!dir
->untracked
)
2676 if (untracked_cache_disabled
< 0)
2677 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2678 if (untracked_cache_disabled
)
2682 * We only support $GIT_DIR/info/exclude and core.excludesfile
2683 * as the global ignore rule files. Any other additions
2684 * (e.g. from command line) invalidate the cache. This
2685 * condition also catches running setup_standard_excludes()
2686 * before setting dir->untracked!
2688 if (dir
->unmanaged_exclude_files
)
2692 * Optimize for the main use case only: whole-tree git
2693 * status. More work involved in treat_leading_path() if we
2694 * use cache on just a subset of the worktree. pathspec
2695 * support could make the matter even worse.
2697 if (base_len
|| (pathspec
&& pathspec
->nr
))
2700 /* Different set of flags may produce different results */
2701 if (dir
->flags
!= dir
->untracked
->dir_flags
||
2703 * See treat_directory(), case index_nonexistent. Without
2704 * this flag, we may need to also cache .git file content
2705 * for the resolve_gitlink_ref() call, which we don't.
2707 !(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
) ||
2708 /* We don't support collecting ignore files */
2709 (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2710 DIR_COLLECT_IGNORED
)))
2714 * If we use .gitignore in the cache and now you change it to
2715 * .gitexclude, everything will go wrong.
2717 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2718 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2722 * EXC_CMDL is not considered in the cache. If people set it,
2725 if (dir
->exclude_list_group
[EXC_CMDL
].nr
)
2728 if (!ident_in_untracked(dir
->untracked
)) {
2729 warning(_("untracked cache is disabled on this system or location"));
2733 if (!dir
->untracked
->root
)
2734 FLEX_ALLOC_STR(dir
->untracked
->root
, name
, "");
2736 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2737 root
= dir
->untracked
->root
;
2738 if (!oideq(&dir
->ss_info_exclude
.oid
,
2739 &dir
->untracked
->ss_info_exclude
.oid
)) {
2740 invalidate_gitignore(dir
->untracked
, root
);
2741 dir
->untracked
->ss_info_exclude
= dir
->ss_info_exclude
;
2743 if (!oideq(&dir
->ss_excludes_file
.oid
,
2744 &dir
->untracked
->ss_excludes_file
.oid
)) {
2745 invalidate_gitignore(dir
->untracked
, root
);
2746 dir
->untracked
->ss_excludes_file
= dir
->ss_excludes_file
;
2749 /* Make sure this directory is not dropped out at saving phase */
2754 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
2755 const char *path
, int len
, const struct pathspec
*pathspec
)
2757 struct untracked_cache_dir
*untracked
;
2759 trace_performance_enter();
2761 if (has_symlink_leading_path(path
, len
)) {
2762 trace_performance_leave("read directory %.*s", len
, path
);
2766 untracked
= validate_untracked_cache(dir
, len
, pathspec
);
2769 * make sure untracked cache code path is disabled,
2770 * e.g. prep_exclude()
2772 dir
->untracked
= NULL
;
2773 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
2774 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
2775 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
2776 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
2778 trace_performance_leave("read directory %.*s", len
, path
);
2779 if (dir
->untracked
) {
2780 static int force_untracked_cache
= -1;
2781 static struct trace_key trace_untracked_stats
= TRACE_KEY_INIT(UNTRACKED_STATS
);
2783 if (force_untracked_cache
< 0)
2784 force_untracked_cache
=
2785 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2786 trace_printf_key(&trace_untracked_stats
,
2787 "node creation: %u\n"
2788 "gitignore invalidation: %u\n"
2789 "directory invalidation: %u\n"
2791 dir
->untracked
->dir_created
,
2792 dir
->untracked
->gitignore_invalidated
,
2793 dir
->untracked
->dir_invalidated
,
2794 dir
->untracked
->dir_opened
);
2795 if (force_untracked_cache
&&
2796 dir
->untracked
== istate
->untracked
&&
2797 (dir
->untracked
->dir_opened
||
2798 dir
->untracked
->gitignore_invalidated
||
2799 dir
->untracked
->dir_invalidated
))
2800 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2801 if (dir
->untracked
!= istate
->untracked
) {
2802 FREE_AND_NULL(dir
->untracked
);
2808 int file_exists(const char *f
)
2811 return lstat(f
, &sb
) == 0;
2814 int repo_file_exists(struct repository
*repo
, const char *path
)
2816 if (repo
!= the_repository
)
2817 BUG("do not know how to check file existence in arbitrary repo");
2819 return file_exists(path
);
2822 static int cmp_icase(char a
, char b
)
2827 return toupper(a
) - toupper(b
);
2832 * Given two normalized paths (a trailing slash is ok), if subdir is
2833 * outside dir, return -1. Otherwise return the offset in subdir that
2834 * can be used as relative path to dir.
2836 int dir_inside_of(const char *subdir
, const char *dir
)
2840 assert(dir
&& subdir
&& *dir
&& *subdir
);
2842 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
2848 /* hel[p]/me vs hel[l]/yeah */
2849 if (*dir
&& *subdir
)
2853 return !*dir
? offset
: -1; /* same dir */
2855 /* foo/[b]ar vs foo/[] */
2856 if (is_dir_sep(dir
[-1]))
2857 return is_dir_sep(subdir
[-1]) ? offset
: -1;
2859 /* foo[/]bar vs foo[] */
2860 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
2863 int is_inside_dir(const char *dir
)
2872 rc
= (dir_inside_of(cwd
, dir
) >= 0);
2877 int is_empty_dir(const char *path
)
2879 DIR *dir
= opendir(path
);
2886 while ((e
= readdir(dir
)) != NULL
)
2887 if (!is_dot_or_dotdot(e
->d_name
)) {
2896 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
2900 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
2901 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
2902 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
2903 struct object_id submodule_head
;
2905 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
2906 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
2907 /* Do not descend and nuke a nested git work tree. */
2913 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
2914 dir
= opendir(path
->buf
);
2916 if (errno
== ENOENT
)
2917 return keep_toplevel
? -1 : 0;
2918 else if (errno
== EACCES
&& !keep_toplevel
)
2920 * An empty dir could be removable even if it
2923 return rmdir(path
->buf
);
2927 strbuf_complete(path
, '/');
2930 while ((e
= readdir(dir
)) != NULL
) {
2932 if (is_dot_or_dotdot(e
->d_name
))
2935 strbuf_setlen(path
, len
);
2936 strbuf_addstr(path
, e
->d_name
);
2937 if (lstat(path
->buf
, &st
)) {
2938 if (errno
== ENOENT
)
2940 * file disappeared, which is what we
2945 } else if (S_ISDIR(st
.st_mode
)) {
2946 if (!remove_dir_recurse(path
, flag
, &kept_down
))
2947 continue; /* happy */
2948 } else if (!only_empty
&&
2949 (!unlink(path
->buf
) || errno
== ENOENT
)) {
2950 continue; /* happy, too */
2953 /* path too long, stat fails, or non-directory still exists */
2959 strbuf_setlen(path
, original_len
);
2960 if (!ret
&& !keep_toplevel
&& !kept_down
)
2961 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
2964 * report the uplevel that it is not an error that we
2965 * did not rmdir() our directory.
2971 int remove_dir_recursively(struct strbuf
*path
, int flag
)
2973 return remove_dir_recurse(path
, flag
, NULL
);
2976 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
2978 void setup_standard_excludes(struct dir_struct
*dir
)
2980 dir
->exclude_per_dir
= ".gitignore";
2982 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2984 excludes_file
= xdg_config_home("ignore");
2985 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
2986 add_patterns_from_file_1(dir
, excludes_file
,
2987 dir
->untracked
? &dir
->ss_excludes_file
: NULL
);
2989 /* per repository user preference */
2990 if (startup_info
->have_repository
) {
2991 const char *path
= git_path_info_exclude();
2992 if (!access_or_warn(path
, R_OK
, 0))
2993 add_patterns_from_file_1(dir
, path
,
2994 dir
->untracked
? &dir
->ss_info_exclude
: NULL
);
2998 char *get_sparse_checkout_filename(void)
3000 return git_pathdup("info/sparse-checkout");
3003 int get_sparse_checkout_patterns(struct pattern_list
*pl
)
3006 char *sparse_filename
= get_sparse_checkout_filename();
3008 pl
->use_cone_patterns
= core_sparse_checkout_cone
;
3009 res
= add_patterns_from_file_to_list(sparse_filename
, "", 0, pl
, NULL
);
3011 free(sparse_filename
);
3015 int remove_path(const char *name
)
3019 if (unlink(name
) && !is_missing_file_error(errno
))
3022 slash
= strrchr(name
, '/');
3024 char *dirs
= xstrdup(name
);
3025 slash
= dirs
+ (slash
- name
);
3028 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
3035 * Frees memory within dir which was allocated, and resets fields for further
3036 * use. Does not free dir itself.
3038 void dir_clear(struct dir_struct
*dir
)
3041 struct exclude_list_group
*group
;
3042 struct pattern_list
*pl
;
3043 struct exclude_stack
*stk
;
3045 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
3046 group
= &dir
->exclude_list_group
[i
];
3047 for (j
= 0; j
< group
->nr
; j
++) {
3050 free((char *)pl
->src
);
3051 clear_pattern_list(pl
);
3056 for (i
= 0; i
< dir
->ignored_nr
; i
++)
3057 free(dir
->ignored
[i
]);
3058 for (i
= 0; i
< dir
->nr
; i
++)
3059 free(dir
->entries
[i
]);
3063 stk
= dir
->exclude_stack
;
3065 struct exclude_stack
*prev
= stk
->prev
;
3069 strbuf_release(&dir
->basebuf
);
3074 struct ondisk_untracked_cache
{
3075 struct stat_data info_exclude_stat
;
3076 struct stat_data excludes_file_stat
;
3080 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
3083 int index
; /* number of written untracked_cache_dir */
3084 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
3085 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
3086 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
3088 struct strbuf sb_stat
;
3089 struct strbuf sb_sha1
;
3092 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
3094 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
3095 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
3096 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
3097 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
3098 to
->sd_dev
= htonl(from
->sd_dev
);
3099 to
->sd_ino
= htonl(from
->sd_ino
);
3100 to
->sd_uid
= htonl(from
->sd_uid
);
3101 to
->sd_gid
= htonl(from
->sd_gid
);
3102 to
->sd_size
= htonl(from
->sd_size
);
3105 static void write_one_dir(struct untracked_cache_dir
*untracked
,
3106 struct write_data
*wd
)
3108 struct stat_data stat_data
;
3109 struct strbuf
*out
= &wd
->out
;
3110 unsigned char intbuf
[16];
3111 unsigned int intlen
, value
;
3112 int i
= wd
->index
++;
3115 * untracked_nr should be reset whenever valid is clear, but
3118 if (!untracked
->valid
) {
3119 untracked
->untracked_nr
= 0;
3120 untracked
->check_only
= 0;
3123 if (untracked
->check_only
)
3124 ewah_set(wd
->check_only
, i
);
3125 if (untracked
->valid
) {
3126 ewah_set(wd
->valid
, i
);
3127 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
3128 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
3130 if (!is_null_oid(&untracked
->exclude_oid
)) {
3131 ewah_set(wd
->sha1_valid
, i
);
3132 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
3133 the_hash_algo
->rawsz
);
3136 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
3137 strbuf_add(out
, intbuf
, intlen
);
3139 /* skip non-recurse directories */
3140 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
3141 if (untracked
->dirs
[i
]->recurse
)
3143 intlen
= encode_varint(value
, intbuf
);
3144 strbuf_add(out
, intbuf
, intlen
);
3146 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
3148 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
3149 strbuf_add(out
, untracked
->untracked
[i
],
3150 strlen(untracked
->untracked
[i
]) + 1);
3152 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
3153 if (untracked
->dirs
[i
]->recurse
)
3154 write_one_dir(untracked
->dirs
[i
], wd
);
3157 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
3159 struct ondisk_untracked_cache
*ouc
;
3160 struct write_data wd
;
3161 unsigned char varbuf
[16];
3163 const unsigned hashsz
= the_hash_algo
->rawsz
;
3165 CALLOC_ARRAY(ouc
, 1);
3166 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
3167 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
3168 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
3170 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
3171 strbuf_add(out
, varbuf
, varint_len
);
3172 strbuf_addbuf(out
, &untracked
->ident
);
3174 strbuf_add(out
, ouc
, sizeof(*ouc
));
3175 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
3176 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
3177 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
3180 if (!untracked
->root
) {
3181 varint_len
= encode_varint(0, varbuf
);
3182 strbuf_add(out
, varbuf
, varint_len
);
3187 wd
.check_only
= ewah_new();
3188 wd
.valid
= ewah_new();
3189 wd
.sha1_valid
= ewah_new();
3190 strbuf_init(&wd
.out
, 1024);
3191 strbuf_init(&wd
.sb_stat
, 1024);
3192 strbuf_init(&wd
.sb_sha1
, 1024);
3193 write_one_dir(untracked
->root
, &wd
);
3195 varint_len
= encode_varint(wd
.index
, varbuf
);
3196 strbuf_add(out
, varbuf
, varint_len
);
3197 strbuf_addbuf(out
, &wd
.out
);
3198 ewah_serialize_strbuf(wd
.valid
, out
);
3199 ewah_serialize_strbuf(wd
.check_only
, out
);
3200 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
3201 strbuf_addbuf(out
, &wd
.sb_stat
);
3202 strbuf_addbuf(out
, &wd
.sb_sha1
);
3203 strbuf_addch(out
, '\0'); /* safe guard for string lists */
3205 ewah_free(wd
.valid
);
3206 ewah_free(wd
.check_only
);
3207 ewah_free(wd
.sha1_valid
);
3208 strbuf_release(&wd
.out
);
3209 strbuf_release(&wd
.sb_stat
);
3210 strbuf_release(&wd
.sb_sha1
);
3213 static void free_untracked(struct untracked_cache_dir
*ucd
)
3218 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
3219 free_untracked(ucd
->dirs
[i
]);
3220 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
3221 free(ucd
->untracked
[i
]);
3222 free(ucd
->untracked
);
3227 void free_untracked_cache(struct untracked_cache
*uc
)
3230 free_untracked(uc
->root
);
3236 struct untracked_cache_dir
**ucd
;
3237 struct ewah_bitmap
*check_only
;
3238 struct ewah_bitmap
*valid
;
3239 struct ewah_bitmap
*sha1_valid
;
3240 const unsigned char *data
;
3241 const unsigned char *end
;
3244 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
3246 memcpy(to
, data
, sizeof(*to
));
3247 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
3248 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
3249 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
3250 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
3251 to
->sd_dev
= ntohl(to
->sd_dev
);
3252 to
->sd_ino
= ntohl(to
->sd_ino
);
3253 to
->sd_uid
= ntohl(to
->sd_uid
);
3254 to
->sd_gid
= ntohl(to
->sd_gid
);
3255 to
->sd_size
= ntohl(to
->sd_size
);
3258 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
3259 struct read_data
*rd
)
3261 struct untracked_cache_dir ud
, *untracked
;
3262 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
3263 const unsigned char *eos
;
3267 memset(&ud
, 0, sizeof(ud
));
3269 value
= decode_varint(&data
);
3273 ud
.untracked_alloc
= value
;
3274 ud
.untracked_nr
= value
;
3275 if (ud
.untracked_nr
)
3276 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
3278 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
3281 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
3283 eos
= memchr(data
, '\0', end
- data
);
3284 if (!eos
|| eos
== end
)
3287 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
3288 memcpy(untracked
, &ud
, sizeof(ud
));
3289 memcpy(untracked
->name
, data
, eos
- data
+ 1);
3292 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
3293 eos
= memchr(data
, '\0', end
- data
);
3294 if (!eos
|| eos
== end
)
3296 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
3300 rd
->ucd
[rd
->index
++] = untracked
;
3303 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
3304 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
3310 static void set_check_only(size_t pos
, void *cb
)
3312 struct read_data
*rd
= cb
;
3313 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3317 static void read_stat(size_t pos
, void *cb
)
3319 struct read_data
*rd
= cb
;
3320 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3321 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
3322 rd
->data
= rd
->end
+ 1;
3325 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
3326 rd
->data
+= sizeof(struct stat_data
);
3330 static void read_oid(size_t pos
, void *cb
)
3332 struct read_data
*rd
= cb
;
3333 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
3334 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
3335 rd
->data
= rd
->end
+ 1;
3338 hashcpy(ud
->exclude_oid
.hash
, rd
->data
);
3339 rd
->data
+= the_hash_algo
->rawsz
;
3342 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
3343 const unsigned char *sha1
)
3345 stat_data_from_disk(&oid_stat
->stat
, data
);
3346 hashcpy(oid_stat
->oid
.hash
, sha1
);
3347 oid_stat
->valid
= 1;
3350 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
3352 struct untracked_cache
*uc
;
3353 struct read_data rd
;
3354 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
3358 const char *exclude_per_dir
;
3359 const unsigned hashsz
= the_hash_algo
->rawsz
;
3360 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
3361 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
3363 if (sz
<= 1 || end
[-1] != '\0')
3367 ident_len
= decode_varint(&next
);
3368 if (next
+ ident_len
> end
)
3370 ident
= (const char *)next
;
3373 if (next
+ exclude_per_dir_offset
+ 1 > end
)
3376 CALLOC_ARRAY(uc
, 1);
3377 strbuf_init(&uc
->ident
, ident_len
);
3378 strbuf_add(&uc
->ident
, ident
, ident_len
);
3379 load_oid_stat(&uc
->ss_info_exclude
,
3380 next
+ ouc_offset(info_exclude_stat
),
3382 load_oid_stat(&uc
->ss_excludes_file
,
3383 next
+ ouc_offset(excludes_file_stat
),
3384 next
+ offset
+ hashsz
);
3385 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
3386 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
3387 uc
->exclude_per_dir
= xstrdup(exclude_per_dir
);
3388 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3389 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
3393 len
= decode_varint(&next
);
3394 if (next
> end
|| len
== 0)
3397 rd
.valid
= ewah_new();
3398 rd
.check_only
= ewah_new();
3399 rd
.sha1_valid
= ewah_new();
3403 ALLOC_ARRAY(rd
.ucd
, len
);
3405 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
3409 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
3414 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
3419 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
3423 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
3424 rd
.data
= next
+ len
;
3425 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
3426 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
3431 ewah_free(rd
.valid
);
3432 ewah_free(rd
.check_only
);
3433 ewah_free(rd
.sha1_valid
);
3436 free_untracked_cache(uc
);
3442 static void invalidate_one_directory(struct untracked_cache
*uc
,
3443 struct untracked_cache_dir
*ucd
)
3445 uc
->dir_invalidated
++;
3447 ucd
->untracked_nr
= 0;
3451 * Normally when an entry is added or removed from a directory,
3452 * invalidating that directory is enough. No need to touch its
3453 * ancestors. When a directory is shown as "foo/bar/" in git-status
3454 * however, deleting or adding an entry may have cascading effect.
3456 * Say the "foo/bar/file" has become untracked, we need to tell the
3457 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3458 * directory any more (because "bar" is managed by foo as an untracked
3461 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3462 * was the last untracked entry in the entire "foo", we should show
3463 * "foo/" instead. Which means we have to invalidate past "bar" up to
3466 * This function traverses all directories from root to leaf. If there
3467 * is a chance of one of the above cases happening, we invalidate back
3468 * to root. Otherwise we just invalidate the leaf. There may be a more
3469 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3470 * detect these cases and avoid unnecessary invalidation, for example,
3471 * checking for the untracked entry named "bar/" in "foo", but for now
3472 * stick to something safe and simple.
3474 static int invalidate_one_component(struct untracked_cache
*uc
,
3475 struct untracked_cache_dir
*dir
,
3476 const char *path
, int len
)
3478 const char *rest
= strchr(path
, '/');
3481 int component_len
= rest
- path
;
3482 struct untracked_cache_dir
*d
=
3483 lookup_untracked(uc
, dir
, path
, component_len
);
3485 invalidate_one_component(uc
, d
, rest
+ 1,
3486 len
- (component_len
+ 1));
3488 invalidate_one_directory(uc
, dir
);
3492 invalidate_one_directory(uc
, dir
);
3493 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3496 void untracked_cache_invalidate_path(struct index_state
*istate
,
3497 const char *path
, int safe_path
)
3499 if (!istate
->untracked
|| !istate
->untracked
->root
)
3501 if (!safe_path
&& !verify_path(path
, 0))
3503 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3504 path
, strlen(path
));
3507 void untracked_cache_remove_from_index(struct index_state
*istate
,
3510 untracked_cache_invalidate_path(istate
, path
, 1);
3513 void untracked_cache_add_to_index(struct index_state
*istate
,
3516 untracked_cache_invalidate_path(istate
, path
, 1);
3519 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3520 const char *sub_gitdir
)
3523 struct repository subrepo
;
3524 struct strbuf sub_wt
= STRBUF_INIT
;
3525 struct strbuf sub_gd
= STRBUF_INIT
;
3527 const struct submodule
*sub
;
3529 /* If the submodule has no working tree, we can ignore it. */
3530 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3533 if (repo_read_index(&subrepo
) < 0)
3534 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3536 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3537 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3539 if (!S_ISGITLINK(ce
->ce_mode
))
3542 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3543 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3545 * Skip entries with the same name in different stages
3546 * to make sure an entry is returned only once.
3550 sub
= submodule_from_path(&subrepo
, &null_oid
, ce
->name
);
3551 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3552 /* .gitmodules broken or inactive sub */
3555 strbuf_reset(&sub_wt
);
3556 strbuf_reset(&sub_gd
);
3557 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3558 strbuf_addf(&sub_gd
, "%s/modules/%s", sub_gitdir
, sub
->name
);
3560 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3562 strbuf_release(&sub_wt
);
3563 strbuf_release(&sub_gd
);
3564 repo_clear(&subrepo
);
3567 void connect_work_tree_and_git_dir(const char *work_tree_
,
3568 const char *git_dir_
,
3569 int recurse_into_nested
)
3571 struct strbuf gitfile_sb
= STRBUF_INIT
;
3572 struct strbuf cfg_sb
= STRBUF_INIT
;
3573 struct strbuf rel_path
= STRBUF_INIT
;
3574 char *git_dir
, *work_tree
;
3576 /* Prepare .git file */
3577 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3578 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3579 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3581 /* Prepare config file */
3582 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3583 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3584 die(_("could not create directories for %s"), cfg_sb
.buf
);
3586 git_dir
= real_pathdup(git_dir_
, 1);
3587 work_tree
= real_pathdup(work_tree_
, 1);
3589 /* Write .git file */
3590 write_file(gitfile_sb
.buf
, "gitdir: %s",
3591 relative_path(git_dir
, work_tree
, &rel_path
));
3592 /* Update core.worktree setting */
3593 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3594 relative_path(work_tree
, git_dir
, &rel_path
));
3596 strbuf_release(&gitfile_sb
);
3597 strbuf_release(&cfg_sb
);
3598 strbuf_release(&rel_path
);
3600 if (recurse_into_nested
)
3601 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
3608 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3610 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
3612 if (rename(old_git_dir
, new_git_dir
) < 0)
3613 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3614 old_git_dir
, new_git_dir
);
3616 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);