2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
5 * See Documentation/technical/api-directory-listing.txt
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
13 #include "object-store.h"
16 #include "wildmatch.h"
20 #include "ewah/ewok.h"
21 #include "fsmonitor.h"
22 #include "submodule-config.h"
25 * Tells read_directory_recursive how a file or directory should be treated.
26 * Values are ordered by significance, e.g. if a directory contains both
27 * excluded and untracked files, it is listed as untracked because
28 * path_untracked > path_excluded.
38 * Support data structure for our opendir/readdir/closedir wrappers
42 struct untracked_cache_dir
*untracked
;
48 struct untracked_cache_dir
*ucd
;
51 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
52 struct index_state
*istate
, const char *path
, int len
,
53 struct untracked_cache_dir
*untracked
,
54 int check_only
, int stop_at_first_file
, const struct pathspec
*pathspec
);
55 static int get_dtype(struct dirent
*de
, struct index_state
*istate
,
56 const char *path
, int len
);
58 int count_slashes(const char *s
)
67 int fspathcmp(const char *a
, const char *b
)
69 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
72 int fspathncmp(const char *a
, const char *b
, size_t count
)
74 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
77 int git_fnmatch(const struct pathspec_item
*item
,
78 const char *pattern
, const char *string
,
82 if (ps_strncmp(item
, pattern
, string
, prefix
))
87 if (item
->flags
& PATHSPEC_ONESTAR
) {
88 int pattern_len
= strlen(++pattern
);
89 int string_len
= strlen(string
);
90 return string_len
< pattern_len
||
91 ps_strcmp(item
, pattern
,
92 string
+ string_len
- pattern_len
);
94 if (item
->magic
& PATHSPEC_GLOB
)
95 return wildmatch(pattern
, string
,
97 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0));
99 /* wildmatch has not learned no FNM_PATHNAME mode yet */
100 return wildmatch(pattern
, string
,
101 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0);
104 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
105 const char *string
, int stringlen
,
109 struct strbuf pat_buf
= STRBUF_INIT
;
110 struct strbuf str_buf
= STRBUF_INIT
;
111 const char *use_pat
= pattern
;
112 const char *use_str
= string
;
114 if (pattern
[patternlen
]) {
115 strbuf_add(&pat_buf
, pattern
, patternlen
);
116 use_pat
= pat_buf
.buf
;
118 if (string
[stringlen
]) {
119 strbuf_add(&str_buf
, string
, stringlen
);
120 use_str
= str_buf
.buf
;
124 flags
|= WM_CASEFOLD
;
125 match_status
= wildmatch(use_pat
, use_str
, flags
);
127 strbuf_release(&pat_buf
);
128 strbuf_release(&str_buf
);
133 static size_t common_prefix_len(const struct pathspec
*pathspec
)
139 * ":(icase)path" is treated as a pathspec full of
140 * wildcard. In other words, only prefix is considered common
141 * prefix. If the pathspec is abc/foo abc/bar, running in
142 * subdir xyz, the common prefix is still xyz, not xyz/abc as
145 GUARD_PATHSPEC(pathspec
,
154 for (n
= 0; n
< pathspec
->nr
; n
++) {
155 size_t i
= 0, len
= 0, item_len
;
156 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
158 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
159 item_len
= pathspec
->items
[n
].prefix
;
161 item_len
= pathspec
->items
[n
].nowildcard_len
;
162 while (i
< item_len
&& (n
== 0 || i
< max
)) {
163 char c
= pathspec
->items
[n
].match
[i
];
164 if (c
!= pathspec
->items
[0].match
[i
])
170 if (n
== 0 || len
< max
) {
180 * Returns a copy of the longest leading path common among all
183 char *common_prefix(const struct pathspec
*pathspec
)
185 unsigned long len
= common_prefix_len(pathspec
);
187 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
190 int fill_directory(struct dir_struct
*dir
,
191 struct index_state
*istate
,
192 const struct pathspec
*pathspec
)
198 * Calculate common prefix for the pathspec, and
199 * use that to optimize the directory walk
201 prefix_len
= common_prefix_len(pathspec
);
202 prefix
= prefix_len
? pathspec
->items
[0].match
: "";
204 /* Read the directory and prune it */
205 read_directory(dir
, istate
, prefix
, prefix_len
, pathspec
);
210 int within_depth(const char *name
, int namelen
,
211 int depth
, int max_depth
)
213 const char *cp
= name
, *cpe
= name
+ namelen
;
219 if (depth
> max_depth
)
226 * Read the contents of the blob with the given OID into a buffer.
227 * Append a trailing LF to the end if the last line doesn't have one.
230 * -1 when the OID is invalid or unknown or does not refer to a blob.
231 * 0 when the blob is empty.
232 * 1 along with { data, size } of the (possibly augmented) buffer
235 * Optionally updates the given oid_stat with the given OID (when valid).
237 static int do_read_blob(const struct object_id
*oid
, struct oid_stat
*oid_stat
,
238 size_t *size_out
, char **data_out
)
240 enum object_type type
;
247 data
= read_object_file(oid
, &type
, &sz
);
248 if (!data
|| type
!= OBJ_BLOB
) {
254 memset(&oid_stat
->stat
, 0, sizeof(oid_stat
->stat
));
255 oidcpy(&oid_stat
->oid
, oid
);
263 if (data
[sz
- 1] != '\n') {
264 data
= xrealloc(data
, st_add(sz
, 1));
268 *size_out
= xsize_t(sz
);
274 #define DO_MATCH_EXCLUDE (1<<0)
275 #define DO_MATCH_DIRECTORY (1<<1)
276 #define DO_MATCH_LEADING_PATHSPEC (1<<2)
279 * Does the given pathspec match the given name? A match is found if
281 * (1) the pathspec string is leading directory of 'name' ("RECURSIVELY"), or
282 * (2) the pathspec string has a leading part matching 'name' ("LEADING"), or
283 * (3) the pathspec string is a wildcard and matches 'name' ("WILDCARD"), or
284 * (4) the pathspec string is exactly the same as 'name' ("EXACT").
286 * Return value tells which case it was (1-4), or 0 when there is no match.
288 * It may be instructive to look at a small table of concrete examples
289 * to understand the differences between 1, 2, and 4:
292 * | a/b | a/b/ | a/b/c
293 * ------+-----------+-----------+------------
294 * a/b | EXACT | EXACT[1] | LEADING[2]
295 * Names a/b/ | RECURSIVE | EXACT | LEADING[2]
296 * a/b/c | RECURSIVE | RECURSIVE | EXACT
298 * [1] Only if DO_MATCH_DIRECTORY is passed; otherwise, this is NOT a match.
299 * [2] Only if DO_MATCH_LEADING_PATHSPEC is passed; otherwise, not a match.
301 static int match_pathspec_item(const struct index_state
*istate
,
302 const struct pathspec_item
*item
, int prefix
,
303 const char *name
, int namelen
, unsigned flags
)
305 /* name/namelen has prefix cut off by caller */
306 const char *match
= item
->match
+ prefix
;
307 int matchlen
= item
->len
- prefix
;
310 * The normal call pattern is:
311 * 1. prefix = common_prefix_len(ps);
312 * 2. prune something, or fill_directory
313 * 3. match_pathspec()
315 * 'prefix' at #1 may be shorter than the command's prefix and
316 * it's ok for #2 to match extra files. Those extras will be
319 * Suppose the pathspec is 'foo' and '../bar' running from
320 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
321 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
322 * user does not want XYZ/foo, only the "foo" part should be
323 * case-insensitive. We need to filter out XYZ/foo here. In
324 * other words, we do not trust the caller on comparing the
325 * prefix part when :(icase) is involved. We do exact
326 * comparison ourselves.
328 * Normally the caller (common_prefix_len() in fact) does
329 * _exact_ matching on name[-prefix+1..-1] and we do not need
330 * to check that part. Be defensive and check it anyway, in
331 * case common_prefix_len is changed, or a new caller is
332 * introduced that does not use common_prefix_len.
334 * If the penalty turns out too high when prefix is really
335 * long, maybe change it to
336 * strncmp(match, name, item->prefix - prefix)
338 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
339 strncmp(item
->match
, name
- prefix
, item
->prefix
))
342 if (item
->attr_match_nr
&&
343 !match_pathspec_attrs(istate
, name
, namelen
, item
))
346 /* If the match was just the prefix, we matched */
348 return MATCHED_RECURSIVELY
;
350 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
351 if (matchlen
== namelen
)
352 return MATCHED_EXACTLY
;
354 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
355 return MATCHED_RECURSIVELY
;
356 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
357 match
[matchlen
- 1] == '/' &&
358 namelen
== matchlen
- 1 &&
359 !ps_strncmp(item
, match
, name
, namelen
))
360 return MATCHED_EXACTLY
;
362 if (item
->nowildcard_len
< item
->len
&&
363 !git_fnmatch(item
, match
, name
,
364 item
->nowildcard_len
- prefix
))
365 return MATCHED_FNMATCH
;
367 /* Perform checks to see if "name" is a leading string of the pathspec */
368 if (flags
& DO_MATCH_LEADING_PATHSPEC
) {
369 /* name is a literal prefix of the pathspec */
370 int offset
= name
[namelen
-1] == '/' ? 1 : 0;
371 if ((namelen
< matchlen
) &&
372 (match
[namelen
-offset
] == '/') &&
373 !ps_strncmp(item
, match
, name
, namelen
))
374 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
376 /* name doesn't match up to the first wild character */
377 if (item
->nowildcard_len
< item
->len
&&
378 ps_strncmp(item
, match
, name
,
379 item
->nowildcard_len
- prefix
))
383 * name has no wildcard, and it didn't match as a leading
384 * pathspec so return.
386 if (item
->nowildcard_len
== item
->len
)
390 * Here is where we would perform a wildmatch to check if
391 * "name" can be matched as a directory (or a prefix) against
392 * the pathspec. Since wildmatch doesn't have this capability
393 * at the present we have to punt and say that it is a match,
394 * potentially returning a false positive
395 * The submodules themselves will be able to perform more
396 * accurate matching to determine if the pathspec matches.
398 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
405 * Given a name and a list of pathspecs, returns the nature of the
406 * closest (i.e. most specific) match of the name to any of the
409 * The caller typically calls this multiple times with the same
410 * pathspec and seen[] array but with different name/namelen
411 * (e.g. entries from the index) and is interested in seeing if and
412 * how each pathspec matches all the names it calls this function
413 * with. A mark is left in the seen[] array for each pathspec element
414 * indicating the closest type of match that element achieved, so if
415 * seen[n] remains zero after multiple invocations, that means the nth
416 * pathspec did not match any names, which could indicate that the
417 * user mistyped the nth pathspec.
419 static int do_match_pathspec(const struct index_state
*istate
,
420 const struct pathspec
*ps
,
421 const char *name
, int namelen
,
422 int prefix
, char *seen
,
425 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
437 if (!ps
->recursive
||
438 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
440 return MATCHED_RECURSIVELY
;
442 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
443 return MATCHED_EXACTLY
;
451 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
454 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
455 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
458 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
461 * Make exclude patterns optional and never report
462 * "pathspec ':(exclude)foo' matches no files"
464 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
465 seen
[i
] = MATCHED_FNMATCH
;
466 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
469 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
470 ps
->max_depth
!= -1 &&
471 how
&& how
!= MATCHED_FNMATCH
) {
472 int len
= ps
->items
[i
].len
;
473 if (name
[len
] == '/')
475 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
476 how
= MATCHED_EXACTLY
;
483 if (seen
&& seen
[i
] < how
)
490 int match_pathspec(const struct index_state
*istate
,
491 const struct pathspec
*ps
,
492 const char *name
, int namelen
,
493 int prefix
, char *seen
, int is_dir
)
495 int positive
, negative
;
496 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
497 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
498 prefix
, seen
, flags
);
499 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
501 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
503 flags
| DO_MATCH_EXCLUDE
);
504 return negative
? 0 : positive
;
508 * Check if a submodule is a superset of the pathspec
510 int submodule_path_match(const struct index_state
*istate
,
511 const struct pathspec
*ps
,
512 const char *submodule_name
,
515 int matched
= do_match_pathspec(istate
, ps
, submodule_name
,
516 strlen(submodule_name
),
519 DO_MATCH_LEADING_PATHSPEC
);
523 int report_path_error(const char *ps_matched
,
524 const struct pathspec
*pathspec
)
527 * Make sure all pathspec matched; otherwise it is an error.
530 for (num
= 0; num
< pathspec
->nr
; num
++) {
531 int other
, found_dup
;
536 * The caller might have fed identical pathspec
537 * twice. Do not barf on such a mistake.
538 * FIXME: parse_pathspec should have eliminated
539 * duplicate pathspec.
541 for (found_dup
= other
= 0;
542 !found_dup
&& other
< pathspec
->nr
;
544 if (other
== num
|| !ps_matched
[other
])
546 if (!strcmp(pathspec
->items
[other
].original
,
547 pathspec
->items
[num
].original
))
549 * Ok, we have a match already.
556 error(_("pathspec '%s' did not match any file(s) known to git"),
557 pathspec
->items
[num
].original
);
564 * Return the length of the "simple" part of a path match limiter.
566 int simple_length(const char *match
)
571 unsigned char c
= *match
++;
573 if (c
== '\0' || is_glob_special(c
))
578 int no_wildcard(const char *string
)
580 return string
[simple_length(string
)] == '\0';
583 void parse_path_pattern(const char **pattern
,
588 const char *p
= *pattern
;
593 *flags
|= PATTERN_FLAG_NEGATIVE
;
597 if (len
&& p
[len
- 1] == '/') {
599 *flags
|= PATTERN_FLAG_MUSTBEDIR
;
601 for (i
= 0; i
< len
; i
++) {
606 *flags
|= PATTERN_FLAG_NODIR
;
607 *nowildcardlen
= simple_length(p
);
609 * we should have excluded the trailing slash from 'p' too,
610 * but that's one more allocation. Instead just make sure
611 * nowildcardlen does not exceed real patternlen
613 if (*nowildcardlen
> len
)
614 *nowildcardlen
= len
;
615 if (*p
== '*' && no_wildcard(p
+ 1))
616 *flags
|= PATTERN_FLAG_ENDSWITH
;
621 void add_pattern(const char *string
, const char *base
,
622 int baselen
, struct pattern_list
*pl
, int srcpos
)
624 struct path_pattern
*pattern
;
629 parse_path_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
630 if (flags
& PATTERN_FLAG_MUSTBEDIR
) {
631 FLEXPTR_ALLOC_MEM(pattern
, pattern
, string
, patternlen
);
633 pattern
= xmalloc(sizeof(*pattern
));
634 pattern
->pattern
= string
;
636 pattern
->patternlen
= patternlen
;
637 pattern
->nowildcardlen
= nowildcardlen
;
638 pattern
->base
= base
;
639 pattern
->baselen
= baselen
;
640 pattern
->flags
= flags
;
641 pattern
->srcpos
= srcpos
;
642 ALLOC_GROW(pl
->patterns
, pl
->nr
+ 1, pl
->alloc
);
643 pl
->patterns
[pl
->nr
++] = pattern
;
647 static int read_skip_worktree_file_from_index(const struct index_state
*istate
,
649 size_t *size_out
, char **data_out
,
650 struct oid_stat
*oid_stat
)
655 pos
= index_name_pos(istate
, path
, len
);
658 if (!ce_skip_worktree(istate
->cache
[pos
]))
661 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
665 * Frees memory within pl which was allocated for exclude patterns and
666 * the file buffer. Does not free pl itself.
668 void clear_pattern_list(struct pattern_list
*pl
)
672 for (i
= 0; i
< pl
->nr
; i
++)
673 free(pl
->patterns
[i
]);
677 memset(pl
, 0, sizeof(*pl
));
680 static void trim_trailing_spaces(char *buf
)
682 char *p
, *last_space
= NULL
;
684 for (p
= buf
; *p
; p
++)
704 * Given a subdirectory name and "dir" of the current directory,
705 * search the subdir in "dir" and return it, or create a new one if it
706 * does not exist in "dir".
708 * If "name" has the trailing slash, it'll be excluded in the search.
710 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
711 struct untracked_cache_dir
*dir
,
712 const char *name
, int len
)
715 struct untracked_cache_dir
*d
;
718 if (len
&& name
[len
- 1] == '/')
722 while (last
> first
) {
723 int cmp
, next
= first
+ ((last
- first
) >> 1);
725 cmp
= strncmp(name
, d
->name
, len
);
726 if (!cmp
&& strlen(d
->name
) > len
)
738 FLEX_ALLOC_MEM(d
, name
, name
, len
);
740 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
741 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
742 dir
->dirs_nr
- first
);
744 dir
->dirs
[first
] = d
;
748 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
752 dir
->untracked_nr
= 0;
753 for (i
= 0; i
< dir
->dirs_nr
; i
++)
754 do_invalidate_gitignore(dir
->dirs
[i
]);
757 static void invalidate_gitignore(struct untracked_cache
*uc
,
758 struct untracked_cache_dir
*dir
)
760 uc
->gitignore_invalidated
++;
761 do_invalidate_gitignore(dir
);
764 static void invalidate_directory(struct untracked_cache
*uc
,
765 struct untracked_cache_dir
*dir
)
770 * Invalidation increment here is just roughly correct. If
771 * untracked_nr or any of dirs[].recurse is non-zero, we
772 * should increment dir_invalidated too. But that's more
776 uc
->dir_invalidated
++;
779 dir
->untracked_nr
= 0;
780 for (i
= 0; i
< dir
->dirs_nr
; i
++)
781 dir
->dirs
[i
]->recurse
= 0;
784 static int add_patterns_from_buffer(char *buf
, size_t size
,
785 const char *base
, int baselen
,
786 struct pattern_list
*pl
);
789 * Given a file with name "fname", read it (either from disk, or from
790 * an index if 'istate' is non-null), parse it and store the
791 * exclude rules in "pl".
793 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
794 * stat data from disk (only valid if add_patterns returns zero). If
795 * ss_valid is non-zero, "ss" must contain good value as input.
797 static int add_patterns(const char *fname
, const char *base
, int baselen
,
798 struct pattern_list
*pl
, struct index_state
*istate
,
799 struct oid_stat
*oid_stat
)
807 fd
= open(fname
, O_RDONLY
);
808 if (fd
< 0 || fstat(fd
, &st
) < 0) {
810 warn_on_fopen_errors(fname
);
815 r
= read_skip_worktree_file_from_index(istate
, fname
,
821 size
= xsize_t(st
.st_size
);
824 fill_stat_data(&oid_stat
->stat
, &st
);
825 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
831 buf
= xmallocz(size
);
832 if (read_in_full(fd
, buf
, size
) != size
) {
841 if (oid_stat
->valid
&&
842 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
843 ; /* no content change, ss->sha1 still good */
845 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
846 !ce_stage(istate
->cache
[pos
]) &&
847 ce_uptodate(istate
->cache
[pos
]) &&
848 !would_convert_to_git(istate
, fname
))
849 oidcpy(&oid_stat
->oid
,
850 &istate
->cache
[pos
]->oid
);
852 hash_object_file(buf
, size
, "blob",
854 fill_stat_data(&oid_stat
->stat
, &st
);
859 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
863 static int add_patterns_from_buffer(char *buf
, size_t size
,
864 const char *base
, int baselen
,
865 struct pattern_list
*pl
)
872 if (skip_utf8_bom(&buf
, size
))
873 size
-= buf
- pl
->filebuf
;
877 for (i
= 0; i
< size
; i
++) {
878 if (buf
[i
] == '\n') {
879 if (entry
!= buf
+ i
&& entry
[0] != '#') {
880 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
881 trim_trailing_spaces(entry
);
882 add_pattern(entry
, base
, baselen
, pl
, lineno
);
891 int add_patterns_from_file_to_list(const char *fname
, const char *base
,
892 int baselen
, struct pattern_list
*pl
,
893 struct index_state
*istate
)
895 return add_patterns(fname
, base
, baselen
, pl
, istate
, NULL
);
898 int add_patterns_from_blob_to_list(
899 struct object_id
*oid
,
900 const char *base
, int baselen
,
901 struct pattern_list
*pl
)
907 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
911 add_patterns_from_buffer(buf
, size
, base
, baselen
, pl
);
915 struct pattern_list
*add_pattern_list(struct dir_struct
*dir
,
916 int group_type
, const char *src
)
918 struct pattern_list
*pl
;
919 struct exclude_list_group
*group
;
921 group
= &dir
->exclude_list_group
[group_type
];
922 ALLOC_GROW(group
->pl
, group
->nr
+ 1, group
->alloc
);
923 pl
= &group
->pl
[group
->nr
++];
924 memset(pl
, 0, sizeof(*pl
));
930 * Used to set up core.excludesfile and .git/info/exclude lists.
932 static void add_patterns_from_file_1(struct dir_struct
*dir
, const char *fname
,
933 struct oid_stat
*oid_stat
)
935 struct pattern_list
*pl
;
937 * catch setup_standard_excludes() that's called before
938 * dir->untracked is assigned. That function behaves
939 * differently when dir->untracked is non-NULL.
942 dir
->unmanaged_exclude_files
++;
943 pl
= add_pattern_list(dir
, EXC_FILE
, fname
);
944 if (add_patterns(fname
, "", 0, pl
, NULL
, oid_stat
) < 0)
945 die(_("cannot use %s as an exclude file"), fname
);
948 void add_patterns_from_file(struct dir_struct
*dir
, const char *fname
)
950 dir
->unmanaged_exclude_files
++; /* see validate_untracked_cache() */
951 add_patterns_from_file_1(dir
, fname
, NULL
);
954 int match_basename(const char *basename
, int basenamelen
,
955 const char *pattern
, int prefix
, int patternlen
,
958 if (prefix
== patternlen
) {
959 if (patternlen
== basenamelen
&&
960 !fspathncmp(pattern
, basename
, basenamelen
))
962 } else if (flags
& PATTERN_FLAG_ENDSWITH
) {
963 /* "*literal" matching against "fooliteral" */
964 if (patternlen
- 1 <= basenamelen
&&
965 !fspathncmp(pattern
+ 1,
966 basename
+ basenamelen
- (patternlen
- 1),
970 if (fnmatch_icase_mem(pattern
, patternlen
,
971 basename
, basenamelen
,
978 int match_pathname(const char *pathname
, int pathlen
,
979 const char *base
, int baselen
,
980 const char *pattern
, int prefix
, int patternlen
,
987 * match with FNM_PATHNAME; the pattern has base implicitly
990 if (*pattern
== '/') {
997 * baselen does not count the trailing slash. base[] may or
998 * may not end with a trailing slash though.
1000 if (pathlen
< baselen
+ 1 ||
1001 (baselen
&& pathname
[baselen
] != '/') ||
1002 fspathncmp(pathname
, base
, baselen
))
1005 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
1006 name
= pathname
+ pathlen
- namelen
;
1010 * if the non-wildcard part is longer than the
1011 * remaining pathname, surely it cannot match.
1013 if (prefix
> namelen
)
1016 if (fspathncmp(pattern
, name
, prefix
))
1019 patternlen
-= prefix
;
1024 * If the whole pattern did not have a wildcard,
1025 * then our prefix match is all we need; we
1026 * do not need to call fnmatch at all.
1028 if (!patternlen
&& !namelen
)
1032 return fnmatch_icase_mem(pattern
, patternlen
,
1038 * Scan the given exclude list in reverse to see whether pathname
1039 * should be ignored. The first match (i.e. the last on the list), if
1040 * any, determines the fate. Returns the exclude_list element which
1041 * matched, or NULL for undecided.
1043 static struct path_pattern
*last_matching_pattern_from_list(const char *pathname
,
1045 const char *basename
,
1047 struct pattern_list
*pl
,
1048 struct index_state
*istate
)
1050 struct path_pattern
*res
= NULL
; /* undecided */
1054 return NULL
; /* undefined */
1056 for (i
= pl
->nr
- 1; 0 <= i
; i
--) {
1057 struct path_pattern
*pattern
= pl
->patterns
[i
];
1058 const char *exclude
= pattern
->pattern
;
1059 int prefix
= pattern
->nowildcardlen
;
1061 if (pattern
->flags
& PATTERN_FLAG_MUSTBEDIR
) {
1062 if (*dtype
== DT_UNKNOWN
)
1063 *dtype
= get_dtype(NULL
, istate
, pathname
, pathlen
);
1064 if (*dtype
!= DT_DIR
)
1068 if (pattern
->flags
& PATTERN_FLAG_NODIR
) {
1069 if (match_basename(basename
,
1070 pathlen
- (basename
- pathname
),
1071 exclude
, prefix
, pattern
->patternlen
,
1079 assert(pattern
->baselen
== 0 ||
1080 pattern
->base
[pattern
->baselen
- 1] == '/');
1081 if (match_pathname(pathname
, pathlen
,
1083 pattern
->baselen
? pattern
->baselen
- 1 : 0,
1084 exclude
, prefix
, pattern
->patternlen
,
1094 * Scan the list of patterns to determine if the ordered list
1095 * of patterns matches on 'pathname'.
1097 * Return 1 for a match, 0 for not matched and -1 for undecided.
1099 enum pattern_match_result
path_matches_pattern_list(
1100 const char *pathname
, int pathlen
,
1101 const char *basename
, int *dtype
,
1102 struct pattern_list
*pl
,
1103 struct index_state
*istate
)
1105 struct path_pattern
*pattern
;
1106 pattern
= last_matching_pattern_from_list(pathname
, pathlen
, basename
,
1109 if (pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1118 static struct path_pattern
*last_matching_pattern_from_lists(
1119 struct dir_struct
*dir
, struct index_state
*istate
,
1120 const char *pathname
, int pathlen
,
1121 const char *basename
, int *dtype_p
)
1124 struct exclude_list_group
*group
;
1125 struct path_pattern
*pattern
;
1126 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1127 group
= &dir
->exclude_list_group
[i
];
1128 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1129 pattern
= last_matching_pattern_from_list(
1130 pathname
, pathlen
, basename
, dtype_p
,
1131 &group
->pl
[j
], istate
);
1140 * Loads the per-directory exclude list for the substring of base
1141 * which has a char length of baselen.
1143 static void prep_exclude(struct dir_struct
*dir
,
1144 struct index_state
*istate
,
1145 const char *base
, int baselen
)
1147 struct exclude_list_group
*group
;
1148 struct pattern_list
*pl
;
1149 struct exclude_stack
*stk
= NULL
;
1150 struct untracked_cache_dir
*untracked
;
1153 group
= &dir
->exclude_list_group
[EXC_DIRS
];
1156 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1157 * which originate from directories not in the prefix of the
1158 * path being checked.
1160 while ((stk
= dir
->exclude_stack
) != NULL
) {
1161 if (stk
->baselen
<= baselen
&&
1162 !strncmp(dir
->basebuf
.buf
, base
, stk
->baselen
))
1164 pl
= &group
->pl
[dir
->exclude_stack
->exclude_ix
];
1165 dir
->exclude_stack
= stk
->prev
;
1166 dir
->pattern
= NULL
;
1167 free((char *)pl
->src
); /* see strbuf_detach() below */
1168 clear_pattern_list(pl
);
1173 /* Skip traversing into sub directories if the parent is excluded */
1178 * Lazy initialization. All call sites currently just
1179 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1180 * them seems lots of work for little benefit.
1182 if (!dir
->basebuf
.buf
)
1183 strbuf_init(&dir
->basebuf
, PATH_MAX
);
1185 /* Read from the parent directories and push them down. */
1186 current
= stk
? stk
->baselen
: -1;
1187 strbuf_setlen(&dir
->basebuf
, current
< 0 ? 0 : current
);
1189 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1193 while (current
< baselen
) {
1195 struct oid_stat oid_stat
;
1197 stk
= xcalloc(1, sizeof(*stk
));
1202 cp
= strchr(base
+ current
+ 1, '/');
1204 die("oops in prep_exclude");
1207 lookup_untracked(dir
->untracked
, untracked
,
1209 cp
- base
- current
);
1211 stk
->prev
= dir
->exclude_stack
;
1212 stk
->baselen
= cp
- base
;
1213 stk
->exclude_ix
= group
->nr
;
1214 stk
->ucd
= untracked
;
1215 pl
= add_pattern_list(dir
, EXC_DIRS
, NULL
);
1216 strbuf_add(&dir
->basebuf
, base
+ current
, stk
->baselen
- current
);
1217 assert(stk
->baselen
== dir
->basebuf
.len
);
1219 /* Abort if the directory is excluded */
1222 dir
->basebuf
.buf
[stk
->baselen
- 1] = 0;
1223 dir
->pattern
= last_matching_pattern_from_lists(dir
,
1225 dir
->basebuf
.buf
, stk
->baselen
- 1,
1226 dir
->basebuf
.buf
+ current
, &dt
);
1227 dir
->basebuf
.buf
[stk
->baselen
- 1] = '/';
1229 dir
->pattern
->flags
& PATTERN_FLAG_NEGATIVE
)
1230 dir
->pattern
= NULL
;
1232 dir
->exclude_stack
= stk
;
1237 /* Try to read per-directory file */
1238 oidclr(&oid_stat
.oid
);
1240 if (dir
->exclude_per_dir
&&
1242 * If we know that no files have been added in
1243 * this directory (i.e. valid_cached_dir() has
1244 * been executed and set untracked->valid) ..
1246 (!untracked
|| !untracked
->valid
||
1248 * .. and .gitignore does not exist before
1249 * (i.e. null exclude_oid). Then we can skip
1250 * loading .gitignore, which would result in
1253 !is_null_oid(&untracked
->exclude_oid
))) {
1255 * dir->basebuf gets reused by the traversal, but we
1256 * need fname to remain unchanged to ensure the src
1257 * member of each struct path_pattern correctly
1258 * back-references its source file. Other invocations
1259 * of add_pattern_list provide stable strings, so we
1260 * strbuf_detach() and free() here in the caller.
1262 struct strbuf sb
= STRBUF_INIT
;
1263 strbuf_addbuf(&sb
, &dir
->basebuf
);
1264 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1265 pl
->src
= strbuf_detach(&sb
, NULL
);
1266 add_patterns(pl
->src
, pl
->src
, stk
->baselen
, pl
, istate
,
1267 untracked
? &oid_stat
: NULL
);
1270 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1271 * will first be called in valid_cached_dir() then maybe many
1272 * times more in last_matching_pattern(). When the cache is
1273 * used, last_matching_pattern() will not be called and
1274 * reading .gitignore content will be a waste.
1276 * So when it's called by valid_cached_dir() and we can get
1277 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1278 * modified on work tree), we could delay reading the
1279 * .gitignore content until we absolutely need it in
1280 * last_matching_pattern(). Be careful about ignore rule
1281 * order, though, if you do that.
1284 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1285 invalidate_gitignore(dir
->untracked
, untracked
);
1286 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1288 dir
->exclude_stack
= stk
;
1289 current
= stk
->baselen
;
1291 strbuf_setlen(&dir
->basebuf
, baselen
);
1295 * Loads the exclude lists for the directory containing pathname, then
1296 * scans all exclude lists to determine whether pathname is excluded.
1297 * Returns the exclude_list element which matched, or NULL for
1300 struct path_pattern
*last_matching_pattern(struct dir_struct
*dir
,
1301 struct index_state
*istate
,
1302 const char *pathname
,
1305 int pathlen
= strlen(pathname
);
1306 const char *basename
= strrchr(pathname
, '/');
1307 basename
= (basename
) ? basename
+1 : pathname
;
1309 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1312 return dir
->pattern
;
1314 return last_matching_pattern_from_lists(dir
, istate
, pathname
, pathlen
,
1319 * Loads the exclude lists for the directory containing pathname, then
1320 * scans all exclude lists to determine whether pathname is excluded.
1321 * Returns 1 if true, otherwise 0.
1323 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1324 const char *pathname
, int *dtype_p
)
1326 struct path_pattern
*pattern
=
1327 last_matching_pattern(dir
, istate
, pathname
, dtype_p
);
1329 return pattern
->flags
& PATTERN_FLAG_NEGATIVE
? 0 : 1;
1333 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1335 struct dir_entry
*ent
;
1337 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1342 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1343 struct index_state
*istate
,
1344 const char *pathname
, int len
)
1346 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1349 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->alloc
);
1350 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1353 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1354 struct index_state
*istate
,
1355 const char *pathname
, int len
)
1357 if (!index_name_is_other(istate
, pathname
, len
))
1360 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->ignored_alloc
);
1361 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1365 index_nonexistent
= 0,
1371 * Do not use the alphabetically sorted index to look up
1372 * the directory name; instead, use the case insensitive
1375 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1376 const char *dirname
, int len
)
1378 struct cache_entry
*ce
;
1380 if (index_dir_exists(istate
, dirname
, len
))
1381 return index_directory
;
1383 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1384 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1385 return index_gitdir
;
1387 return index_nonexistent
;
1391 * The index sorts alphabetically by entry name, which
1392 * means that a gitlink sorts as '\0' at the end, while
1393 * a directory (which is defined not as an entry, but as
1394 * the files it contains) will sort with the '/' at the
1397 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1398 const char *dirname
, int len
)
1403 return directory_exists_in_index_icase(istate
, dirname
, len
);
1405 pos
= index_name_pos(istate
, dirname
, len
);
1408 while (pos
< istate
->cache_nr
) {
1409 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1410 unsigned char endchar
;
1412 if (strncmp(ce
->name
, dirname
, len
))
1414 endchar
= ce
->name
[len
];
1418 return index_directory
;
1419 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1420 return index_gitdir
;
1422 return index_nonexistent
;
1426 * When we find a directory when traversing the filesystem, we
1427 * have three distinct cases:
1430 * - see it as a directory
1433 * and which one we choose depends on a combination of existing
1434 * git index contents and the flags passed into the directory
1435 * traversal routine.
1437 * Case 1: If we *already* have entries in the index under that
1438 * directory name, we always recurse into the directory to see
1441 * Case 2: If we *already* have that directory name as a gitlink,
1442 * we always continue to see it as a gitlink, regardless of whether
1443 * there is an actual git directory there or not (it might not
1444 * be checked out as a subproject!)
1446 * Case 3: if we didn't have it in the index previously, we
1447 * have a few sub-cases:
1449 * (a) if "show_other_directories" is true, we show it as
1450 * just a directory, unless "hide_empty_directories" is
1451 * also true, in which case we need to check if it contains any
1452 * untracked and / or ignored files.
1453 * (b) if it looks like a git directory, and we don't have
1454 * 'no_gitlinks' set we treat it as a gitlink, and show it
1456 * (c) otherwise, we recurse into it.
1458 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1459 struct index_state
*istate
,
1460 struct untracked_cache_dir
*untracked
,
1461 const char *dirname
, int len
, int baselen
, int exclude
,
1462 const struct pathspec
*pathspec
)
1464 /* The "len-1" is to strip the final '/' */
1465 switch (directory_exists_in_index(istate
, dirname
, len
-1)) {
1466 case index_directory
:
1467 return path_recurse
;
1472 case index_nonexistent
:
1473 if (dir
->flags
& DIR_SKIP_NESTED_GIT
) {
1475 struct strbuf sb
= STRBUF_INIT
;
1476 strbuf_addstr(&sb
, dirname
);
1477 nested_repo
= is_nonbare_repository_dir(&sb
);
1478 strbuf_release(&sb
);
1483 if (dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)
1486 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1487 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1490 * This is an excluded directory and we are
1491 * showing ignored paths that match an exclude
1492 * pattern. (e.g. show directory as ignored
1493 * only if it matches an exclude pattern).
1494 * This path will either be 'path_excluded`
1495 * (if we are showing empty directories or if
1496 * the directory is not empty), or will be
1497 * 'path_none' (empty directory, and we are
1498 * not showing empty directories).
1500 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1501 return path_excluded
;
1503 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1504 untracked
, 1, 1, pathspec
) == path_excluded
)
1505 return path_excluded
;
1509 if (!(dir
->flags
& DIR_NO_GITLINKS
)) {
1510 struct strbuf sb
= STRBUF_INIT
;
1511 strbuf_addstr(&sb
, dirname
);
1512 if (is_nonbare_repository_dir(&sb
))
1513 return exclude
? path_excluded
: path_untracked
;
1514 strbuf_release(&sb
);
1516 return path_recurse
;
1519 /* This is the "show_other_directories" case */
1521 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1522 return exclude
? path_excluded
: path_untracked
;
1524 untracked
= lookup_untracked(dir
->untracked
, untracked
,
1525 dirname
+ baselen
, len
- baselen
);
1528 * If this is an excluded directory, then we only need to check if
1529 * the directory contains any files.
1531 return read_directory_recursive(dir
, istate
, dirname
, len
,
1532 untracked
, 1, exclude
, pathspec
);
1536 * This is an inexact early pruning of any recursive directory
1537 * reading - if the path cannot possibly be in the pathspec,
1538 * return true, and we'll skip it early.
1540 static int simplify_away(const char *path
, int pathlen
,
1541 const struct pathspec
*pathspec
)
1545 if (!pathspec
|| !pathspec
->nr
)
1548 GUARD_PATHSPEC(pathspec
,
1557 for (i
= 0; i
< pathspec
->nr
; i
++) {
1558 const struct pathspec_item
*item
= &pathspec
->items
[i
];
1559 int len
= item
->nowildcard_len
;
1563 if (!ps_strncmp(item
, item
->match
, path
, len
))
1571 * This function tells us whether an excluded path matches a
1572 * list of "interesting" pathspecs. That is, whether a path matched
1573 * by any of the pathspecs could possibly be ignored by excluding
1574 * the specified path. This can happen if:
1576 * 1. the path is mentioned explicitly in the pathspec
1578 * 2. the path is a directory prefix of some element in the
1581 static int exclude_matches_pathspec(const char *path
, int pathlen
,
1582 const struct pathspec
*pathspec
)
1586 if (!pathspec
|| !pathspec
->nr
)
1589 GUARD_PATHSPEC(pathspec
,
1597 for (i
= 0; i
< pathspec
->nr
; i
++) {
1598 const struct pathspec_item
*item
= &pathspec
->items
[i
];
1599 int len
= item
->nowildcard_len
;
1601 if (len
== pathlen
&&
1602 !ps_strncmp(item
, item
->match
, path
, pathlen
))
1604 if (len
> pathlen
&&
1605 item
->match
[pathlen
] == '/' &&
1606 !ps_strncmp(item
, item
->match
, path
, pathlen
))
1612 static int get_index_dtype(struct index_state
*istate
,
1613 const char *path
, int len
)
1616 const struct cache_entry
*ce
;
1618 ce
= index_file_exists(istate
, path
, len
, 0);
1620 if (!ce_uptodate(ce
))
1622 if (S_ISGITLINK(ce
->ce_mode
))
1625 * Nobody actually cares about the
1626 * difference between DT_LNK and DT_REG
1631 /* Try to look it up as a directory */
1632 pos
= index_name_pos(istate
, path
, len
);
1636 while (pos
< istate
->cache_nr
) {
1637 ce
= istate
->cache
[pos
++];
1638 if (strncmp(ce
->name
, path
, len
))
1640 if (ce
->name
[len
] > '/')
1642 if (ce
->name
[len
] < '/')
1644 if (!ce_uptodate(ce
))
1645 break; /* continue? */
1651 static int get_dtype(struct dirent
*de
, struct index_state
*istate
,
1652 const char *path
, int len
)
1654 int dtype
= de
? DTYPE(de
) : DT_UNKNOWN
;
1657 if (dtype
!= DT_UNKNOWN
)
1659 dtype
= get_index_dtype(istate
, path
, len
);
1660 if (dtype
!= DT_UNKNOWN
)
1662 if (lstat(path
, &st
))
1664 if (S_ISREG(st
.st_mode
))
1666 if (S_ISDIR(st
.st_mode
))
1668 if (S_ISLNK(st
.st_mode
))
1673 static enum path_treatment
treat_one_path(struct dir_struct
*dir
,
1674 struct untracked_cache_dir
*untracked
,
1675 struct index_state
*istate
,
1676 struct strbuf
*path
,
1678 const struct pathspec
*pathspec
,
1679 int dtype
, struct dirent
*de
)
1682 int has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
, ignore_case
);
1683 enum path_treatment path_treatment
;
1685 if (dtype
== DT_UNKNOWN
)
1686 dtype
= get_dtype(de
, istate
, path
->buf
, path
->len
);
1688 /* Always exclude indexed files */
1689 if (dtype
!= DT_DIR
&& has_path_in_index
)
1693 * When we are looking at a directory P in the working tree,
1694 * there are three cases:
1696 * (1) P exists in the index. Everything inside the directory P in
1697 * the working tree needs to go when P is checked out from the
1700 * (2) P does not exist in the index, but there is P/Q in the index.
1701 * We know P will stay a directory when we check out the contents
1702 * of the index, but we do not know yet if there is a directory
1703 * P/Q in the working tree to be killed, so we need to recurse.
1705 * (3) P does not exist in the index, and there is no P/Q in the index
1706 * to require P to be a directory, either. Only in this case, we
1707 * know that everything inside P will not be killed without
1710 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
1711 (dtype
== DT_DIR
) &&
1712 !has_path_in_index
&&
1713 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
1716 exclude
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
1719 * Excluded? If we don't explicitly want to show
1720 * ignored files, ignore it
1722 if (exclude
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
1723 return path_excluded
;
1729 strbuf_addch(path
, '/');
1730 path_treatment
= treat_directory(dir
, istate
, untracked
,
1731 path
->buf
, path
->len
,
1732 baselen
, exclude
, pathspec
);
1734 * If 1) we only want to return directories that
1735 * match an exclude pattern and 2) this directory does
1736 * not match an exclude pattern but all of its
1737 * contents are excluded, then indicate that we should
1738 * recurse into this directory (instead of marking the
1739 * directory itself as an ignored path).
1742 path_treatment
== path_excluded
&&
1743 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1744 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
1745 return path_recurse
;
1746 return path_treatment
;
1749 return exclude
? path_excluded
: path_untracked
;
1753 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
1754 struct untracked_cache_dir
*untracked
,
1755 struct cached_dir
*cdir
,
1756 struct index_state
*istate
,
1757 struct strbuf
*path
,
1759 const struct pathspec
*pathspec
)
1761 strbuf_setlen(path
, baselen
);
1763 strbuf_addstr(path
, cdir
->file
);
1764 return path_untracked
;
1766 strbuf_addstr(path
, cdir
->ucd
->name
);
1767 /* treat_one_path() does this before it calls treat_directory() */
1768 strbuf_complete(path
, '/');
1769 if (cdir
->ucd
->check_only
)
1771 * check_only is set as a result of treat_directory() getting
1772 * to its bottom. Verify again the same set of directories
1773 * with check_only set.
1775 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
1776 cdir
->ucd
, 1, 0, pathspec
);
1778 * We get path_recurse in the first run when
1779 * directory_exists_in_index() returns index_nonexistent. We
1780 * are sure that new changes in the index does not impact the
1781 * outcome. Return now.
1783 return path_recurse
;
1786 static enum path_treatment
treat_path(struct dir_struct
*dir
,
1787 struct untracked_cache_dir
*untracked
,
1788 struct cached_dir
*cdir
,
1789 struct index_state
*istate
,
1790 struct strbuf
*path
,
1792 const struct pathspec
*pathspec
)
1795 struct dirent
*de
= cdir
->de
;
1798 return treat_path_fast(dir
, untracked
, cdir
, istate
, path
,
1800 if (is_dot_or_dotdot(de
->d_name
) || !fspathcmp(de
->d_name
, ".git"))
1802 strbuf_setlen(path
, baselen
);
1803 strbuf_addstr(path
, de
->d_name
);
1804 if (simplify_away(path
->buf
, path
->len
, pathspec
))
1808 return treat_one_path(dir
, untracked
, istate
, path
, baselen
, pathspec
, dtype
, de
);
1811 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
1815 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
1816 dir
->untracked_alloc
);
1817 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
1820 static int valid_cached_dir(struct dir_struct
*dir
,
1821 struct untracked_cache_dir
*untracked
,
1822 struct index_state
*istate
,
1823 struct strbuf
*path
,
1832 * With fsmonitor, we can trust the untracked cache's valid field.
1834 refresh_fsmonitor(istate
);
1835 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
1836 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
1837 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
1840 if (!untracked
->valid
||
1841 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
1842 fill_stat_data(&untracked
->stat_data
, &st
);
1847 if (untracked
->check_only
!= !!check_only
)
1851 * prep_exclude will be called eventually on this directory,
1852 * but it's called much later in last_matching_pattern(). We
1853 * need it now to determine the validity of the cache for this
1854 * path. The next calls will be nearly no-op, the way
1855 * prep_exclude() is designed.
1857 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
1858 strbuf_addch(path
, '/');
1859 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
1860 strbuf_setlen(path
, path
->len
- 1);
1862 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
1864 /* hopefully prep_exclude() haven't invalidated this entry... */
1865 return untracked
->valid
;
1868 static int open_cached_dir(struct cached_dir
*cdir
,
1869 struct dir_struct
*dir
,
1870 struct untracked_cache_dir
*untracked
,
1871 struct index_state
*istate
,
1872 struct strbuf
*path
,
1877 memset(cdir
, 0, sizeof(*cdir
));
1878 cdir
->untracked
= untracked
;
1879 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
1881 c_path
= path
->len
? path
->buf
: ".";
1882 cdir
->fdir
= opendir(c_path
);
1884 warning_errno(_("could not open directory '%s'"), c_path
);
1885 if (dir
->untracked
) {
1886 invalidate_directory(dir
->untracked
, untracked
);
1887 dir
->untracked
->dir_opened
++;
1894 static int read_cached_dir(struct cached_dir
*cdir
)
1897 cdir
->de
= readdir(cdir
->fdir
);
1902 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
1903 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
1913 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
1914 struct untracked_cache_dir
*d
= cdir
->untracked
;
1915 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
1921 static void close_cached_dir(struct cached_dir
*cdir
)
1924 closedir(cdir
->fdir
);
1926 * We have gone through this directory and found no untracked
1927 * entries. Mark it valid.
1929 if (cdir
->untracked
) {
1930 cdir
->untracked
->valid
= 1;
1931 cdir
->untracked
->recurse
= 1;
1935 static void add_path_to_appropriate_result_list(struct dir_struct
*dir
,
1936 struct untracked_cache_dir
*untracked
,
1937 struct cached_dir
*cdir
,
1938 struct index_state
*istate
,
1939 struct strbuf
*path
,
1941 const struct pathspec
*pathspec
,
1942 enum path_treatment state
)
1944 /* add the path to the appropriate result list */
1947 if (dir
->flags
& DIR_SHOW_IGNORED
)
1948 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
1949 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
1950 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
1951 exclude_matches_pathspec(path
->buf
, path
->len
,
1953 dir_add_ignored(dir
, istate
, path
->buf
, path
->len
);
1956 case path_untracked
:
1957 if (dir
->flags
& DIR_SHOW_IGNORED
)
1959 dir_add_name(dir
, istate
, path
->buf
, path
->len
);
1961 add_untracked(untracked
, path
->buf
+ baselen
);
1970 * Read a directory tree. We currently ignore anything but
1971 * directories, regular files and symlinks. That's because git
1972 * doesn't handle them at all yet. Maybe that will change some
1975 * Also, we ignore the name ".git" (even if it is not a directory).
1976 * That likely will not change.
1978 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1979 * to signal that a file was found. This is the least significant value that
1980 * indicates that a file was encountered that does not depend on the order of
1981 * whether an untracked or exluded path was encountered first.
1983 * Returns the most significant path_treatment value encountered in the scan.
1984 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1985 * significant path_treatment value that will be returned.
1988 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
1989 struct index_state
*istate
, const char *base
, int baselen
,
1990 struct untracked_cache_dir
*untracked
, int check_only
,
1991 int stop_at_first_file
, const struct pathspec
*pathspec
)
1994 * WARNING WARNING WARNING:
1996 * Any updates to the traversal logic here may need corresponding
1997 * updates in treat_leading_path(). See the commit message for the
1998 * commit adding this warning as well as the commit preceding it
2002 struct cached_dir cdir
;
2003 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
2004 struct strbuf path
= STRBUF_INIT
;
2006 strbuf_add(&path
, base
, baselen
);
2008 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
2012 untracked
->check_only
= !!check_only
;
2014 while (!read_cached_dir(&cdir
)) {
2015 /* check how the file or directory should be treated */
2016 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
2019 if (state
> dir_state
)
2022 /* recurse into subdir if instructed by treat_path */
2023 if ((state
== path_recurse
) ||
2024 ((state
== path_untracked
) &&
2025 (get_dtype(cdir
.de
, istate
, path
.buf
, path
.len
) == DT_DIR
) &&
2026 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2028 do_match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2029 baselen
, NULL
, DO_MATCH_LEADING_PATHSPEC
) == MATCHED_RECURSIVELY_LEADING_PATHSPEC
)))) {
2030 struct untracked_cache_dir
*ud
;
2031 ud
= lookup_untracked(dir
->untracked
, untracked
,
2033 path
.len
- baselen
);
2035 read_directory_recursive(dir
, istate
, path
.buf
,
2037 check_only
, stop_at_first_file
, pathspec
);
2038 if (subdir_state
> dir_state
)
2039 dir_state
= subdir_state
;
2042 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
2043 0 /* prefix */, NULL
,
2044 0 /* do NOT special case dirs */))
2049 if (stop_at_first_file
) {
2051 * If stopping at first file, then
2052 * signal that a file was found by
2053 * returning `path_excluded`. This is
2054 * to return a consistent value
2055 * regardless of whether an ignored or
2056 * excluded file happened to be
2059 * In current usage, the
2060 * `stop_at_first_file` is passed when
2061 * an ancestor directory has matched
2062 * an exclude pattern, so any found
2063 * files will be excluded.
2065 if (dir_state
>= path_excluded
) {
2066 dir_state
= path_excluded
;
2071 /* abort early if maximum state has been reached */
2072 if (dir_state
== path_untracked
) {
2074 add_untracked(untracked
, path
.buf
+ baselen
);
2077 /* skip the dir_add_* part */
2081 add_path_to_appropriate_result_list(dir
, untracked
, &cdir
,
2082 istate
, &path
, baselen
,
2085 close_cached_dir(&cdir
);
2087 strbuf_release(&path
);
2092 int cmp_dir_entry(const void *p1
, const void *p2
)
2094 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2095 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2097 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2100 /* check if *out lexically strictly contains *in */
2101 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2103 return (out
->len
< in
->len
) &&
2104 (out
->name
[out
->len
- 1] == '/') &&
2105 !memcmp(out
->name
, in
->name
, out
->len
);
2108 static int treat_leading_path(struct dir_struct
*dir
,
2109 struct index_state
*istate
,
2110 const char *path
, int len
,
2111 const struct pathspec
*pathspec
)
2114 * WARNING WARNING WARNING:
2116 * Any updates to the traversal logic here may need corresponding
2117 * updates in treat_leading_path(). See the commit message for the
2118 * commit adding this warning as well as the commit preceding it
2122 struct strbuf sb
= STRBUF_INIT
;
2123 int prevlen
, baselen
;
2125 struct cached_dir cdir
;
2127 enum path_treatment state
= path_none
;
2130 * For each directory component of path, we are going to check whether
2131 * that path is relevant given the pathspec. For example, if path is
2133 * then we will ask treat_path() whether we should go into foo, then
2134 * whether we should go into bar, then whether baz is relevant.
2135 * Checking each is important because e.g. if path is
2137 * then we need to check .git to know we shouldn't traverse it.
2138 * If the return from treat_path() is:
2139 * * path_none, for any path, we return false.
2140 * * path_recurse, for all path components, we return true
2141 * * <anything else> for some intermediate component, we make sure
2142 * to add that path to the relevant list but return false
2143 * signifying that we shouldn't recurse into it.
2146 while (len
&& path
[len
- 1] == '/')
2152 * We need a manufactured dirent with sufficient space to store a
2153 * leading directory component of path in its d_name. Here, we
2154 * assume that the dirent's d_name is either declared as
2155 * char d_name[BIG_ENOUGH]
2156 * or that it is declared at the end of the struct as
2158 * For either case, padding with len+1 bytes at the end will ensure
2159 * sufficient storage space.
2161 de
= xcalloc(1, sizeof(struct dirent
)+len
+1);
2162 memset(&cdir
, 0, sizeof(cdir
));
2164 #if defined(DT_UNKNOWN) && !defined(NO_D_TYPE_IN_DIRENT)
2165 de
->d_type
= DT_DIR
;
2170 prevlen
= baselen
+ !!baselen
;
2171 cp
= path
+ prevlen
;
2172 cp
= memchr(cp
, '/', path
+ len
- cp
);
2176 baselen
= cp
- path
;
2178 strbuf_add(&sb
, path
, baselen
);
2179 if (!is_directory(sb
.buf
))
2182 strbuf_add(&sb
, path
, prevlen
);
2183 memcpy(de
->d_name
, path
+prevlen
, baselen
-prevlen
);
2184 de
->d_name
[baselen
-prevlen
] = '\0';
2185 state
= treat_path(dir
, NULL
, &cdir
, istate
, &sb
, prevlen
,
2187 if (state
== path_untracked
&&
2188 get_dtype(cdir
.de
, istate
, sb
.buf
, sb
.len
) == DT_DIR
&&
2189 (dir
->flags
& DIR_SHOW_IGNORED_TOO
||
2190 do_match_pathspec(istate
, pathspec
, sb
.buf
, sb
.len
,
2191 baselen
, NULL
, DO_MATCH_LEADING_PATHSPEC
) == MATCHED_RECURSIVELY_LEADING_PATHSPEC
)) {
2192 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
,
2196 state
= path_recurse
;
2199 if (state
!= path_recurse
)
2200 break; /* do not recurse into it */
2202 break; /* finished checking */
2204 add_path_to_appropriate_result_list(dir
, NULL
, &cdir
, istate
,
2205 &sb
, baselen
, pathspec
,
2209 strbuf_release(&sb
);
2210 return state
== path_recurse
;
2213 static const char *get_ident_string(void)
2215 static struct strbuf sb
= STRBUF_INIT
;
2220 if (uname(&uts
) < 0)
2221 die_errno(_("failed to get kernel name and information"));
2222 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2227 static int ident_in_untracked(const struct untracked_cache
*uc
)
2230 * Previous git versions may have saved many NUL separated
2231 * strings in the "ident" field, but it is insane to manage
2232 * many locations, so just take care of the first one.
2235 return !strcmp(uc
->ident
.buf
, get_ident_string());
2238 static void set_untracked_ident(struct untracked_cache
*uc
)
2240 strbuf_reset(&uc
->ident
);
2241 strbuf_addstr(&uc
->ident
, get_ident_string());
2244 * This strbuf used to contain a list of NUL separated
2245 * strings, so save NUL too for backward compatibility.
2247 strbuf_addch(&uc
->ident
, 0);
2250 static void new_untracked_cache(struct index_state
*istate
)
2252 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2253 strbuf_init(&uc
->ident
, 100);
2254 uc
->exclude_per_dir
= ".gitignore";
2255 /* should be the same flags used by git-status */
2256 uc
->dir_flags
= DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2257 set_untracked_ident(uc
);
2258 istate
->untracked
= uc
;
2259 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2262 void add_untracked_cache(struct index_state
*istate
)
2264 if (!istate
->untracked
) {
2265 new_untracked_cache(istate
);
2267 if (!ident_in_untracked(istate
->untracked
)) {
2268 free_untracked_cache(istate
->untracked
);
2269 new_untracked_cache(istate
);
2274 void remove_untracked_cache(struct index_state
*istate
)
2276 if (istate
->untracked
) {
2277 free_untracked_cache(istate
->untracked
);
2278 istate
->untracked
= NULL
;
2279 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2283 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2285 const struct pathspec
*pathspec
)
2287 struct untracked_cache_dir
*root
;
2288 static int untracked_cache_disabled
= -1;
2290 if (!dir
->untracked
)
2292 if (untracked_cache_disabled
< 0)
2293 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2294 if (untracked_cache_disabled
)
2298 * We only support $GIT_DIR/info/exclude and core.excludesfile
2299 * as the global ignore rule files. Any other additions
2300 * (e.g. from command line) invalidate the cache. This
2301 * condition also catches running setup_standard_excludes()
2302 * before setting dir->untracked!
2304 if (dir
->unmanaged_exclude_files
)
2308 * Optimize for the main use case only: whole-tree git
2309 * status. More work involved in treat_leading_path() if we
2310 * use cache on just a subset of the worktree. pathspec
2311 * support could make the matter even worse.
2313 if (base_len
|| (pathspec
&& pathspec
->nr
))
2316 /* Different set of flags may produce different results */
2317 if (dir
->flags
!= dir
->untracked
->dir_flags
||
2319 * See treat_directory(), case index_nonexistent. Without
2320 * this flag, we may need to also cache .git file content
2321 * for the resolve_gitlink_ref() call, which we don't.
2323 !(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
) ||
2324 /* We don't support collecting ignore files */
2325 (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2326 DIR_COLLECT_IGNORED
)))
2330 * If we use .gitignore in the cache and now you change it to
2331 * .gitexclude, everything will go wrong.
2333 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2334 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2338 * EXC_CMDL is not considered in the cache. If people set it,
2341 if (dir
->exclude_list_group
[EXC_CMDL
].nr
)
2344 if (!ident_in_untracked(dir
->untracked
)) {
2345 warning(_("untracked cache is disabled on this system or location"));
2349 if (!dir
->untracked
->root
) {
2350 const int len
= sizeof(*dir
->untracked
->root
);
2351 dir
->untracked
->root
= xmalloc(len
);
2352 memset(dir
->untracked
->root
, 0, len
);
2355 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2356 root
= dir
->untracked
->root
;
2357 if (!oideq(&dir
->ss_info_exclude
.oid
,
2358 &dir
->untracked
->ss_info_exclude
.oid
)) {
2359 invalidate_gitignore(dir
->untracked
, root
);
2360 dir
->untracked
->ss_info_exclude
= dir
->ss_info_exclude
;
2362 if (!oideq(&dir
->ss_excludes_file
.oid
,
2363 &dir
->untracked
->ss_excludes_file
.oid
)) {
2364 invalidate_gitignore(dir
->untracked
, root
);
2365 dir
->untracked
->ss_excludes_file
= dir
->ss_excludes_file
;
2368 /* Make sure this directory is not dropped out at saving phase */
2373 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
2374 const char *path
, int len
, const struct pathspec
*pathspec
)
2376 struct untracked_cache_dir
*untracked
;
2378 trace_performance_enter();
2380 if (has_symlink_leading_path(path
, len
)) {
2381 trace_performance_leave("read directory %.*s", len
, path
);
2385 untracked
= validate_untracked_cache(dir
, len
, pathspec
);
2388 * make sure untracked cache code path is disabled,
2389 * e.g. prep_exclude()
2391 dir
->untracked
= NULL
;
2392 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
2393 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
2394 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
2395 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
2398 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2399 * also pick up untracked contents of untracked dirs; by default
2400 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2402 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2403 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
2406 /* remove from dir->entries untracked contents of untracked dirs */
2407 for (i
= j
= 0; j
< dir
->nr
; j
++) {
2409 check_dir_entry_contains(dir
->entries
[i
- 1], dir
->entries
[j
])) {
2410 FREE_AND_NULL(dir
->entries
[j
]);
2412 dir
->entries
[i
++] = dir
->entries
[j
];
2419 trace_performance_leave("read directory %.*s", len
, path
);
2420 if (dir
->untracked
) {
2421 static int force_untracked_cache
= -1;
2422 static struct trace_key trace_untracked_stats
= TRACE_KEY_INIT(UNTRACKED_STATS
);
2424 if (force_untracked_cache
< 0)
2425 force_untracked_cache
=
2426 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2427 trace_printf_key(&trace_untracked_stats
,
2428 "node creation: %u\n"
2429 "gitignore invalidation: %u\n"
2430 "directory invalidation: %u\n"
2432 dir
->untracked
->dir_created
,
2433 dir
->untracked
->gitignore_invalidated
,
2434 dir
->untracked
->dir_invalidated
,
2435 dir
->untracked
->dir_opened
);
2436 if (force_untracked_cache
&&
2437 dir
->untracked
== istate
->untracked
&&
2438 (dir
->untracked
->dir_opened
||
2439 dir
->untracked
->gitignore_invalidated
||
2440 dir
->untracked
->dir_invalidated
))
2441 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2442 if (dir
->untracked
!= istate
->untracked
) {
2443 FREE_AND_NULL(dir
->untracked
);
2449 int file_exists(const char *f
)
2452 return lstat(f
, &sb
) == 0;
2455 int repo_file_exists(struct repository
*repo
, const char *path
)
2457 if (repo
!= the_repository
)
2458 BUG("do not know how to check file existence in arbitrary repo");
2460 return file_exists(path
);
2463 static int cmp_icase(char a
, char b
)
2468 return toupper(a
) - toupper(b
);
2473 * Given two normalized paths (a trailing slash is ok), if subdir is
2474 * outside dir, return -1. Otherwise return the offset in subdir that
2475 * can be used as relative path to dir.
2477 int dir_inside_of(const char *subdir
, const char *dir
)
2481 assert(dir
&& subdir
&& *dir
&& *subdir
);
2483 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
2489 /* hel[p]/me vs hel[l]/yeah */
2490 if (*dir
&& *subdir
)
2494 return !*dir
? offset
: -1; /* same dir */
2496 /* foo/[b]ar vs foo/[] */
2497 if (is_dir_sep(dir
[-1]))
2498 return is_dir_sep(subdir
[-1]) ? offset
: -1;
2500 /* foo[/]bar vs foo[] */
2501 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
2504 int is_inside_dir(const char *dir
)
2513 rc
= (dir_inside_of(cwd
, dir
) >= 0);
2518 int is_empty_dir(const char *path
)
2520 DIR *dir
= opendir(path
);
2527 while ((e
= readdir(dir
)) != NULL
)
2528 if (!is_dot_or_dotdot(e
->d_name
)) {
2537 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
2541 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
2542 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
2543 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
2544 struct object_id submodule_head
;
2546 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
2547 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
2548 /* Do not descend and nuke a nested git work tree. */
2554 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
2555 dir
= opendir(path
->buf
);
2557 if (errno
== ENOENT
)
2558 return keep_toplevel
? -1 : 0;
2559 else if (errno
== EACCES
&& !keep_toplevel
)
2561 * An empty dir could be removable even if it
2564 return rmdir(path
->buf
);
2568 strbuf_complete(path
, '/');
2571 while ((e
= readdir(dir
)) != NULL
) {
2573 if (is_dot_or_dotdot(e
->d_name
))
2576 strbuf_setlen(path
, len
);
2577 strbuf_addstr(path
, e
->d_name
);
2578 if (lstat(path
->buf
, &st
)) {
2579 if (errno
== ENOENT
)
2581 * file disappeared, which is what we
2586 } else if (S_ISDIR(st
.st_mode
)) {
2587 if (!remove_dir_recurse(path
, flag
, &kept_down
))
2588 continue; /* happy */
2589 } else if (!only_empty
&&
2590 (!unlink(path
->buf
) || errno
== ENOENT
)) {
2591 continue; /* happy, too */
2594 /* path too long, stat fails, or non-directory still exists */
2600 strbuf_setlen(path
, original_len
);
2601 if (!ret
&& !keep_toplevel
&& !kept_down
)
2602 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
2605 * report the uplevel that it is not an error that we
2606 * did not rmdir() our directory.
2612 int remove_dir_recursively(struct strbuf
*path
, int flag
)
2614 return remove_dir_recurse(path
, flag
, NULL
);
2617 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
2619 void setup_standard_excludes(struct dir_struct
*dir
)
2621 dir
->exclude_per_dir
= ".gitignore";
2623 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2625 excludes_file
= xdg_config_home("ignore");
2626 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
2627 add_patterns_from_file_1(dir
, excludes_file
,
2628 dir
->untracked
? &dir
->ss_excludes_file
: NULL
);
2630 /* per repository user preference */
2631 if (startup_info
->have_repository
) {
2632 const char *path
= git_path_info_exclude();
2633 if (!access_or_warn(path
, R_OK
, 0))
2634 add_patterns_from_file_1(dir
, path
,
2635 dir
->untracked
? &dir
->ss_info_exclude
: NULL
);
2639 int remove_path(const char *name
)
2643 if (unlink(name
) && !is_missing_file_error(errno
))
2646 slash
= strrchr(name
, '/');
2648 char *dirs
= xstrdup(name
);
2649 slash
= dirs
+ (slash
- name
);
2652 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
2659 * Frees memory within dir which was allocated for exclude lists and
2660 * the exclude_stack. Does not free dir itself.
2662 void clear_directory(struct dir_struct
*dir
)
2665 struct exclude_list_group
*group
;
2666 struct pattern_list
*pl
;
2667 struct exclude_stack
*stk
;
2669 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
2670 group
= &dir
->exclude_list_group
[i
];
2671 for (j
= 0; j
< group
->nr
; j
++) {
2674 free((char *)pl
->src
);
2675 clear_pattern_list(pl
);
2680 stk
= dir
->exclude_stack
;
2682 struct exclude_stack
*prev
= stk
->prev
;
2686 strbuf_release(&dir
->basebuf
);
2689 struct ondisk_untracked_cache
{
2690 struct stat_data info_exclude_stat
;
2691 struct stat_data excludes_file_stat
;
2695 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2698 int index
; /* number of written untracked_cache_dir */
2699 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
2700 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
2701 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
2703 struct strbuf sb_stat
;
2704 struct strbuf sb_sha1
;
2707 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
2709 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
2710 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
2711 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
2712 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
2713 to
->sd_dev
= htonl(from
->sd_dev
);
2714 to
->sd_ino
= htonl(from
->sd_ino
);
2715 to
->sd_uid
= htonl(from
->sd_uid
);
2716 to
->sd_gid
= htonl(from
->sd_gid
);
2717 to
->sd_size
= htonl(from
->sd_size
);
2720 static void write_one_dir(struct untracked_cache_dir
*untracked
,
2721 struct write_data
*wd
)
2723 struct stat_data stat_data
;
2724 struct strbuf
*out
= &wd
->out
;
2725 unsigned char intbuf
[16];
2726 unsigned int intlen
, value
;
2727 int i
= wd
->index
++;
2730 * untracked_nr should be reset whenever valid is clear, but
2733 if (!untracked
->valid
) {
2734 untracked
->untracked_nr
= 0;
2735 untracked
->check_only
= 0;
2738 if (untracked
->check_only
)
2739 ewah_set(wd
->check_only
, i
);
2740 if (untracked
->valid
) {
2741 ewah_set(wd
->valid
, i
);
2742 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
2743 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
2745 if (!is_null_oid(&untracked
->exclude_oid
)) {
2746 ewah_set(wd
->sha1_valid
, i
);
2747 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
2748 the_hash_algo
->rawsz
);
2751 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
2752 strbuf_add(out
, intbuf
, intlen
);
2754 /* skip non-recurse directories */
2755 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
2756 if (untracked
->dirs
[i
]->recurse
)
2758 intlen
= encode_varint(value
, intbuf
);
2759 strbuf_add(out
, intbuf
, intlen
);
2761 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
2763 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
2764 strbuf_add(out
, untracked
->untracked
[i
],
2765 strlen(untracked
->untracked
[i
]) + 1);
2767 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
2768 if (untracked
->dirs
[i
]->recurse
)
2769 write_one_dir(untracked
->dirs
[i
], wd
);
2772 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
2774 struct ondisk_untracked_cache
*ouc
;
2775 struct write_data wd
;
2776 unsigned char varbuf
[16];
2778 const unsigned hashsz
= the_hash_algo
->rawsz
;
2780 ouc
= xcalloc(1, sizeof(*ouc
));
2781 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
2782 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
2783 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
2785 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
2786 strbuf_add(out
, varbuf
, varint_len
);
2787 strbuf_addbuf(out
, &untracked
->ident
);
2789 strbuf_add(out
, ouc
, sizeof(*ouc
));
2790 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
2791 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
2792 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
2795 if (!untracked
->root
) {
2796 varint_len
= encode_varint(0, varbuf
);
2797 strbuf_add(out
, varbuf
, varint_len
);
2802 wd
.check_only
= ewah_new();
2803 wd
.valid
= ewah_new();
2804 wd
.sha1_valid
= ewah_new();
2805 strbuf_init(&wd
.out
, 1024);
2806 strbuf_init(&wd
.sb_stat
, 1024);
2807 strbuf_init(&wd
.sb_sha1
, 1024);
2808 write_one_dir(untracked
->root
, &wd
);
2810 varint_len
= encode_varint(wd
.index
, varbuf
);
2811 strbuf_add(out
, varbuf
, varint_len
);
2812 strbuf_addbuf(out
, &wd
.out
);
2813 ewah_serialize_strbuf(wd
.valid
, out
);
2814 ewah_serialize_strbuf(wd
.check_only
, out
);
2815 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
2816 strbuf_addbuf(out
, &wd
.sb_stat
);
2817 strbuf_addbuf(out
, &wd
.sb_sha1
);
2818 strbuf_addch(out
, '\0'); /* safe guard for string lists */
2820 ewah_free(wd
.valid
);
2821 ewah_free(wd
.check_only
);
2822 ewah_free(wd
.sha1_valid
);
2823 strbuf_release(&wd
.out
);
2824 strbuf_release(&wd
.sb_stat
);
2825 strbuf_release(&wd
.sb_sha1
);
2828 static void free_untracked(struct untracked_cache_dir
*ucd
)
2833 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
2834 free_untracked(ucd
->dirs
[i
]);
2835 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
2836 free(ucd
->untracked
[i
]);
2837 free(ucd
->untracked
);
2842 void free_untracked_cache(struct untracked_cache
*uc
)
2845 free_untracked(uc
->root
);
2851 struct untracked_cache_dir
**ucd
;
2852 struct ewah_bitmap
*check_only
;
2853 struct ewah_bitmap
*valid
;
2854 struct ewah_bitmap
*sha1_valid
;
2855 const unsigned char *data
;
2856 const unsigned char *end
;
2859 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
2861 memcpy(to
, data
, sizeof(*to
));
2862 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
2863 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
2864 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
2865 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
2866 to
->sd_dev
= ntohl(to
->sd_dev
);
2867 to
->sd_ino
= ntohl(to
->sd_ino
);
2868 to
->sd_uid
= ntohl(to
->sd_uid
);
2869 to
->sd_gid
= ntohl(to
->sd_gid
);
2870 to
->sd_size
= ntohl(to
->sd_size
);
2873 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
2874 struct read_data
*rd
)
2876 struct untracked_cache_dir ud
, *untracked
;
2877 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
2878 const unsigned char *eos
;
2882 memset(&ud
, 0, sizeof(ud
));
2884 value
= decode_varint(&data
);
2888 ud
.untracked_alloc
= value
;
2889 ud
.untracked_nr
= value
;
2890 if (ud
.untracked_nr
)
2891 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
2893 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
2896 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
2898 eos
= memchr(data
, '\0', end
- data
);
2899 if (!eos
|| eos
== end
)
2902 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
2903 memcpy(untracked
, &ud
, sizeof(ud
));
2904 memcpy(untracked
->name
, data
, eos
- data
+ 1);
2907 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
2908 eos
= memchr(data
, '\0', end
- data
);
2909 if (!eos
|| eos
== end
)
2911 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
2915 rd
->ucd
[rd
->index
++] = untracked
;
2918 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
2919 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
2925 static void set_check_only(size_t pos
, void *cb
)
2927 struct read_data
*rd
= cb
;
2928 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
2932 static void read_stat(size_t pos
, void *cb
)
2934 struct read_data
*rd
= cb
;
2935 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
2936 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
2937 rd
->data
= rd
->end
+ 1;
2940 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
2941 rd
->data
+= sizeof(struct stat_data
);
2945 static void read_oid(size_t pos
, void *cb
)
2947 struct read_data
*rd
= cb
;
2948 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
2949 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
2950 rd
->data
= rd
->end
+ 1;
2953 hashcpy(ud
->exclude_oid
.hash
, rd
->data
);
2954 rd
->data
+= the_hash_algo
->rawsz
;
2957 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
2958 const unsigned char *sha1
)
2960 stat_data_from_disk(&oid_stat
->stat
, data
);
2961 hashcpy(oid_stat
->oid
.hash
, sha1
);
2962 oid_stat
->valid
= 1;
2965 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
2967 struct untracked_cache
*uc
;
2968 struct read_data rd
;
2969 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
2973 const char *exclude_per_dir
;
2974 const unsigned hashsz
= the_hash_algo
->rawsz
;
2975 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
2976 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
2978 if (sz
<= 1 || end
[-1] != '\0')
2982 ident_len
= decode_varint(&next
);
2983 if (next
+ ident_len
> end
)
2985 ident
= (const char *)next
;
2988 if (next
+ exclude_per_dir_offset
+ 1 > end
)
2991 uc
= xcalloc(1, sizeof(*uc
));
2992 strbuf_init(&uc
->ident
, ident_len
);
2993 strbuf_add(&uc
->ident
, ident
, ident_len
);
2994 load_oid_stat(&uc
->ss_info_exclude
,
2995 next
+ ouc_offset(info_exclude_stat
),
2997 load_oid_stat(&uc
->ss_excludes_file
,
2998 next
+ ouc_offset(excludes_file_stat
),
2999 next
+ offset
+ hashsz
);
3000 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
3001 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
3002 uc
->exclude_per_dir
= xstrdup(exclude_per_dir
);
3003 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
3004 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
3008 len
= decode_varint(&next
);
3009 if (next
> end
|| len
== 0)
3012 rd
.valid
= ewah_new();
3013 rd
.check_only
= ewah_new();
3014 rd
.sha1_valid
= ewah_new();
3018 ALLOC_ARRAY(rd
.ucd
, len
);
3020 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
3024 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
3029 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
3034 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
3038 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
3039 rd
.data
= next
+ len
;
3040 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
3041 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
3046 ewah_free(rd
.valid
);
3047 ewah_free(rd
.check_only
);
3048 ewah_free(rd
.sha1_valid
);
3051 free_untracked_cache(uc
);
3057 static void invalidate_one_directory(struct untracked_cache
*uc
,
3058 struct untracked_cache_dir
*ucd
)
3060 uc
->dir_invalidated
++;
3062 ucd
->untracked_nr
= 0;
3066 * Normally when an entry is added or removed from a directory,
3067 * invalidating that directory is enough. No need to touch its
3068 * ancestors. When a directory is shown as "foo/bar/" in git-status
3069 * however, deleting or adding an entry may have cascading effect.
3071 * Say the "foo/bar/file" has become untracked, we need to tell the
3072 * untracked_cache_dir of "foo" that "bar/" is not an untracked
3073 * directory any more (because "bar" is managed by foo as an untracked
3076 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
3077 * was the last untracked entry in the entire "foo", we should show
3078 * "foo/" instead. Which means we have to invalidate past "bar" up to
3081 * This function traverses all directories from root to leaf. If there
3082 * is a chance of one of the above cases happening, we invalidate back
3083 * to root. Otherwise we just invalidate the leaf. There may be a more
3084 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
3085 * detect these cases and avoid unnecessary invalidation, for example,
3086 * checking for the untracked entry named "bar/" in "foo", but for now
3087 * stick to something safe and simple.
3089 static int invalidate_one_component(struct untracked_cache
*uc
,
3090 struct untracked_cache_dir
*dir
,
3091 const char *path
, int len
)
3093 const char *rest
= strchr(path
, '/');
3096 int component_len
= rest
- path
;
3097 struct untracked_cache_dir
*d
=
3098 lookup_untracked(uc
, dir
, path
, component_len
);
3100 invalidate_one_component(uc
, d
, rest
+ 1,
3101 len
- (component_len
+ 1));
3103 invalidate_one_directory(uc
, dir
);
3107 invalidate_one_directory(uc
, dir
);
3108 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3111 void untracked_cache_invalidate_path(struct index_state
*istate
,
3112 const char *path
, int safe_path
)
3114 if (!istate
->untracked
|| !istate
->untracked
->root
)
3116 if (!safe_path
&& !verify_path(path
, 0))
3118 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3119 path
, strlen(path
));
3122 void untracked_cache_remove_from_index(struct index_state
*istate
,
3125 untracked_cache_invalidate_path(istate
, path
, 1);
3128 void untracked_cache_add_to_index(struct index_state
*istate
,
3131 untracked_cache_invalidate_path(istate
, path
, 1);
3134 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3135 const char *sub_gitdir
)
3138 struct repository subrepo
;
3139 struct strbuf sub_wt
= STRBUF_INIT
;
3140 struct strbuf sub_gd
= STRBUF_INIT
;
3142 const struct submodule
*sub
;
3144 /* If the submodule has no working tree, we can ignore it. */
3145 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3148 if (repo_read_index(&subrepo
) < 0)
3149 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3151 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3152 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3154 if (!S_ISGITLINK(ce
->ce_mode
))
3157 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3158 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3160 * Skip entries with the same name in different stages
3161 * to make sure an entry is returned only once.
3165 sub
= submodule_from_path(&subrepo
, &null_oid
, ce
->name
);
3166 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3167 /* .gitmodules broken or inactive sub */
3170 strbuf_reset(&sub_wt
);
3171 strbuf_reset(&sub_gd
);
3172 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3173 strbuf_addf(&sub_gd
, "%s/modules/%s", sub_gitdir
, sub
->name
);
3175 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3177 strbuf_release(&sub_wt
);
3178 strbuf_release(&sub_gd
);
3179 repo_clear(&subrepo
);
3182 void connect_work_tree_and_git_dir(const char *work_tree_
,
3183 const char *git_dir_
,
3184 int recurse_into_nested
)
3186 struct strbuf gitfile_sb
= STRBUF_INIT
;
3187 struct strbuf cfg_sb
= STRBUF_INIT
;
3188 struct strbuf rel_path
= STRBUF_INIT
;
3189 char *git_dir
, *work_tree
;
3191 /* Prepare .git file */
3192 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3193 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3194 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3196 /* Prepare config file */
3197 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3198 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3199 die(_("could not create directories for %s"), cfg_sb
.buf
);
3201 git_dir
= real_pathdup(git_dir_
, 1);
3202 work_tree
= real_pathdup(work_tree_
, 1);
3204 /* Write .git file */
3205 write_file(gitfile_sb
.buf
, "gitdir: %s",
3206 relative_path(git_dir
, work_tree
, &rel_path
));
3207 /* Update core.worktree setting */
3208 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3209 relative_path(work_tree
, git_dir
, &rel_path
));
3211 strbuf_release(&gitfile_sb
);
3212 strbuf_release(&cfg_sb
);
3213 strbuf_release(&rel_path
);
3215 if (recurse_into_nested
)
3216 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
3223 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3225 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
3227 if (rename(old_git_dir
, new_git_dir
) < 0)
3228 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3229 old_git_dir
, new_git_dir
);
3231 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);