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 * Here is where we would perform a wildmatch to check if
384 * "name" can be matched as a directory (or a prefix) against
385 * the pathspec. Since wildmatch doesn't have this capability
386 * at the present we have to punt and say that it is a match,
387 * potentially returning a false positive
388 * The submodules themselves will be able to perform more
389 * accurate matching to determine if the pathspec matches.
391 return MATCHED_RECURSIVELY_LEADING_PATHSPEC
;
398 * Given a name and a list of pathspecs, returns the nature of the
399 * closest (i.e. most specific) match of the name to any of the
402 * The caller typically calls this multiple times with the same
403 * pathspec and seen[] array but with different name/namelen
404 * (e.g. entries from the index) and is interested in seeing if and
405 * how each pathspec matches all the names it calls this function
406 * with. A mark is left in the seen[] array for each pathspec element
407 * indicating the closest type of match that element achieved, so if
408 * seen[n] remains zero after multiple invocations, that means the nth
409 * pathspec did not match any names, which could indicate that the
410 * user mistyped the nth pathspec.
412 static int do_match_pathspec(const struct index_state
*istate
,
413 const struct pathspec
*ps
,
414 const char *name
, int namelen
,
415 int prefix
, char *seen
,
418 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
430 if (!ps
->recursive
||
431 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
433 return MATCHED_RECURSIVELY
;
435 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
436 return MATCHED_EXACTLY
;
444 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
447 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
448 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
451 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
454 * Make exclude patterns optional and never report
455 * "pathspec ':(exclude)foo' matches no files"
457 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
458 seen
[i
] = MATCHED_FNMATCH
;
459 how
= match_pathspec_item(istate
, ps
->items
+i
, prefix
, name
,
462 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
463 ps
->max_depth
!= -1 &&
464 how
&& how
!= MATCHED_FNMATCH
) {
465 int len
= ps
->items
[i
].len
;
466 if (name
[len
] == '/')
468 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
469 how
= MATCHED_EXACTLY
;
476 if (seen
&& seen
[i
] < how
)
483 int match_pathspec(const struct index_state
*istate
,
484 const struct pathspec
*ps
,
485 const char *name
, int namelen
,
486 int prefix
, char *seen
, int is_dir
)
488 int positive
, negative
;
489 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
490 positive
= do_match_pathspec(istate
, ps
, name
, namelen
,
491 prefix
, seen
, flags
);
492 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
494 negative
= do_match_pathspec(istate
, ps
, name
, namelen
,
496 flags
| DO_MATCH_EXCLUDE
);
497 return negative
? 0 : positive
;
501 * Check if a submodule is a superset of the pathspec
503 int submodule_path_match(const struct index_state
*istate
,
504 const struct pathspec
*ps
,
505 const char *submodule_name
,
508 int matched
= do_match_pathspec(istate
, ps
, submodule_name
,
509 strlen(submodule_name
),
512 DO_MATCH_LEADING_PATHSPEC
);
516 int report_path_error(const char *ps_matched
,
517 const struct pathspec
*pathspec
)
520 * Make sure all pathspec matched; otherwise it is an error.
523 for (num
= 0; num
< pathspec
->nr
; num
++) {
524 int other
, found_dup
;
529 * The caller might have fed identical pathspec
530 * twice. Do not barf on such a mistake.
531 * FIXME: parse_pathspec should have eliminated
532 * duplicate pathspec.
534 for (found_dup
= other
= 0;
535 !found_dup
&& other
< pathspec
->nr
;
537 if (other
== num
|| !ps_matched
[other
])
539 if (!strcmp(pathspec
->items
[other
].original
,
540 pathspec
->items
[num
].original
))
542 * Ok, we have a match already.
549 error(_("pathspec '%s' did not match any file(s) known to git"),
550 pathspec
->items
[num
].original
);
557 * Return the length of the "simple" part of a path match limiter.
559 int simple_length(const char *match
)
564 unsigned char c
= *match
++;
566 if (c
== '\0' || is_glob_special(c
))
571 int no_wildcard(const char *string
)
573 return string
[simple_length(string
)] == '\0';
576 void parse_exclude_pattern(const char **pattern
,
581 const char *p
= *pattern
;
586 *flags
|= EXC_FLAG_NEGATIVE
;
590 if (len
&& p
[len
- 1] == '/') {
592 *flags
|= EXC_FLAG_MUSTBEDIR
;
594 for (i
= 0; i
< len
; i
++) {
599 *flags
|= EXC_FLAG_NODIR
;
600 *nowildcardlen
= simple_length(p
);
602 * we should have excluded the trailing slash from 'p' too,
603 * but that's one more allocation. Instead just make sure
604 * nowildcardlen does not exceed real patternlen
606 if (*nowildcardlen
> len
)
607 *nowildcardlen
= len
;
608 if (*p
== '*' && no_wildcard(p
+ 1))
609 *flags
|= EXC_FLAG_ENDSWITH
;
614 void add_exclude(const char *string
, const char *base
,
615 int baselen
, struct exclude_list
*el
, int srcpos
)
622 parse_exclude_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
623 if (flags
& EXC_FLAG_MUSTBEDIR
) {
624 FLEXPTR_ALLOC_MEM(x
, pattern
, string
, patternlen
);
626 x
= xmalloc(sizeof(*x
));
629 x
->patternlen
= patternlen
;
630 x
->nowildcardlen
= nowildcardlen
;
632 x
->baselen
= baselen
;
635 ALLOC_GROW(el
->excludes
, el
->nr
+ 1, el
->alloc
);
636 el
->excludes
[el
->nr
++] = x
;
640 static int read_skip_worktree_file_from_index(const struct index_state
*istate
,
642 size_t *size_out
, char **data_out
,
643 struct oid_stat
*oid_stat
)
648 pos
= index_name_pos(istate
, path
, len
);
651 if (!ce_skip_worktree(istate
->cache
[pos
]))
654 return do_read_blob(&istate
->cache
[pos
]->oid
, oid_stat
, size_out
, data_out
);
658 * Frees memory within el which was allocated for exclude patterns and
659 * the file buffer. Does not free el itself.
661 void clear_exclude_list(struct exclude_list
*el
)
665 for (i
= 0; i
< el
->nr
; i
++)
666 free(el
->excludes
[i
]);
670 memset(el
, 0, sizeof(*el
));
673 static void trim_trailing_spaces(char *buf
)
675 char *p
, *last_space
= NULL
;
677 for (p
= buf
; *p
; p
++)
697 * Given a subdirectory name and "dir" of the current directory,
698 * search the subdir in "dir" and return it, or create a new one if it
699 * does not exist in "dir".
701 * If "name" has the trailing slash, it'll be excluded in the search.
703 static struct untracked_cache_dir
*lookup_untracked(struct untracked_cache
*uc
,
704 struct untracked_cache_dir
*dir
,
705 const char *name
, int len
)
708 struct untracked_cache_dir
*d
;
711 if (len
&& name
[len
- 1] == '/')
715 while (last
> first
) {
716 int cmp
, next
= first
+ ((last
- first
) >> 1);
718 cmp
= strncmp(name
, d
->name
, len
);
719 if (!cmp
&& strlen(d
->name
) > len
)
731 FLEX_ALLOC_MEM(d
, name
, name
, len
);
733 ALLOC_GROW(dir
->dirs
, dir
->dirs_nr
+ 1, dir
->dirs_alloc
);
734 MOVE_ARRAY(dir
->dirs
+ first
+ 1, dir
->dirs
+ first
,
735 dir
->dirs_nr
- first
);
737 dir
->dirs
[first
] = d
;
741 static void do_invalidate_gitignore(struct untracked_cache_dir
*dir
)
745 dir
->untracked_nr
= 0;
746 for (i
= 0; i
< dir
->dirs_nr
; i
++)
747 do_invalidate_gitignore(dir
->dirs
[i
]);
750 static void invalidate_gitignore(struct untracked_cache
*uc
,
751 struct untracked_cache_dir
*dir
)
753 uc
->gitignore_invalidated
++;
754 do_invalidate_gitignore(dir
);
757 static void invalidate_directory(struct untracked_cache
*uc
,
758 struct untracked_cache_dir
*dir
)
763 * Invalidation increment here is just roughly correct. If
764 * untracked_nr or any of dirs[].recurse is non-zero, we
765 * should increment dir_invalidated too. But that's more
769 uc
->dir_invalidated
++;
772 dir
->untracked_nr
= 0;
773 for (i
= 0; i
< dir
->dirs_nr
; i
++)
774 dir
->dirs
[i
]->recurse
= 0;
777 static int add_excludes_from_buffer(char *buf
, size_t size
,
778 const char *base
, int baselen
,
779 struct exclude_list
*el
);
782 * Given a file with name "fname", read it (either from disk, or from
783 * an index if 'istate' is non-null), parse it and store the
784 * exclude rules in "el".
786 * If "ss" is not NULL, compute SHA-1 of the exclude file and fill
787 * stat data from disk (only valid if add_excludes returns zero). If
788 * ss_valid is non-zero, "ss" must contain good value as input.
790 static int add_excludes(const char *fname
, const char *base
, int baselen
,
791 struct exclude_list
*el
, struct index_state
*istate
,
792 struct oid_stat
*oid_stat
)
800 fd
= open(fname
, O_RDONLY
);
801 if (fd
< 0 || fstat(fd
, &st
) < 0) {
803 warn_on_fopen_errors(fname
);
808 r
= read_skip_worktree_file_from_index(istate
, fname
,
814 size
= xsize_t(st
.st_size
);
817 fill_stat_data(&oid_stat
->stat
, &st
);
818 oidcpy(&oid_stat
->oid
, the_hash_algo
->empty_blob
);
824 buf
= xmallocz(size
);
825 if (read_in_full(fd
, buf
, size
) != size
) {
834 if (oid_stat
->valid
&&
835 !match_stat_data_racy(istate
, &oid_stat
->stat
, &st
))
836 ; /* no content change, ss->sha1 still good */
838 (pos
= index_name_pos(istate
, fname
, strlen(fname
))) >= 0 &&
839 !ce_stage(istate
->cache
[pos
]) &&
840 ce_uptodate(istate
->cache
[pos
]) &&
841 !would_convert_to_git(istate
, fname
))
842 oidcpy(&oid_stat
->oid
,
843 &istate
->cache
[pos
]->oid
);
845 hash_object_file(buf
, size
, "blob",
847 fill_stat_data(&oid_stat
->stat
, &st
);
852 add_excludes_from_buffer(buf
, size
, base
, baselen
, el
);
856 static int add_excludes_from_buffer(char *buf
, size_t size
,
857 const char *base
, int baselen
,
858 struct exclude_list
*el
)
865 if (skip_utf8_bom(&buf
, size
))
866 size
-= buf
- el
->filebuf
;
870 for (i
= 0; i
< size
; i
++) {
871 if (buf
[i
] == '\n') {
872 if (entry
!= buf
+ i
&& entry
[0] != '#') {
873 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
874 trim_trailing_spaces(entry
);
875 add_exclude(entry
, base
, baselen
, el
, lineno
);
884 int add_excludes_from_file_to_list(const char *fname
, const char *base
,
885 int baselen
, struct exclude_list
*el
,
886 struct index_state
*istate
)
888 return add_excludes(fname
, base
, baselen
, el
, istate
, NULL
);
891 int add_excludes_from_blob_to_list(
892 struct object_id
*oid
,
893 const char *base
, int baselen
,
894 struct exclude_list
*el
)
900 r
= do_read_blob(oid
, NULL
, &size
, &buf
);
904 add_excludes_from_buffer(buf
, size
, base
, baselen
, el
);
908 struct exclude_list
*add_exclude_list(struct dir_struct
*dir
,
909 int group_type
, const char *src
)
911 struct exclude_list
*el
;
912 struct exclude_list_group
*group
;
914 group
= &dir
->exclude_list_group
[group_type
];
915 ALLOC_GROW(group
->el
, group
->nr
+ 1, group
->alloc
);
916 el
= &group
->el
[group
->nr
++];
917 memset(el
, 0, sizeof(*el
));
923 * Used to set up core.excludesfile and .git/info/exclude lists.
925 static void add_excludes_from_file_1(struct dir_struct
*dir
, const char *fname
,
926 struct oid_stat
*oid_stat
)
928 struct exclude_list
*el
;
930 * catch setup_standard_excludes() that's called before
931 * dir->untracked is assigned. That function behaves
932 * differently when dir->untracked is non-NULL.
935 dir
->unmanaged_exclude_files
++;
936 el
= add_exclude_list(dir
, EXC_FILE
, fname
);
937 if (add_excludes(fname
, "", 0, el
, NULL
, oid_stat
) < 0)
938 die(_("cannot use %s as an exclude file"), fname
);
941 void add_excludes_from_file(struct dir_struct
*dir
, const char *fname
)
943 dir
->unmanaged_exclude_files
++; /* see validate_untracked_cache() */
944 add_excludes_from_file_1(dir
, fname
, NULL
);
947 int match_basename(const char *basename
, int basenamelen
,
948 const char *pattern
, int prefix
, int patternlen
,
951 if (prefix
== patternlen
) {
952 if (patternlen
== basenamelen
&&
953 !fspathncmp(pattern
, basename
, basenamelen
))
955 } else if (flags
& EXC_FLAG_ENDSWITH
) {
956 /* "*literal" matching against "fooliteral" */
957 if (patternlen
- 1 <= basenamelen
&&
958 !fspathncmp(pattern
+ 1,
959 basename
+ basenamelen
- (patternlen
- 1),
963 if (fnmatch_icase_mem(pattern
, patternlen
,
964 basename
, basenamelen
,
971 int match_pathname(const char *pathname
, int pathlen
,
972 const char *base
, int baselen
,
973 const char *pattern
, int prefix
, int patternlen
,
980 * match with FNM_PATHNAME; the pattern has base implicitly
983 if (*pattern
== '/') {
990 * baselen does not count the trailing slash. base[] may or
991 * may not end with a trailing slash though.
993 if (pathlen
< baselen
+ 1 ||
994 (baselen
&& pathname
[baselen
] != '/') ||
995 fspathncmp(pathname
, base
, baselen
))
998 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
999 name
= pathname
+ pathlen
- namelen
;
1003 * if the non-wildcard part is longer than the
1004 * remaining pathname, surely it cannot match.
1006 if (prefix
> namelen
)
1009 if (fspathncmp(pattern
, name
, prefix
))
1012 patternlen
-= prefix
;
1017 * If the whole pattern did not have a wildcard,
1018 * then our prefix match is all we need; we
1019 * do not need to call fnmatch at all.
1021 if (!patternlen
&& !namelen
)
1025 return fnmatch_icase_mem(pattern
, patternlen
,
1031 * Scan the given exclude list in reverse to see whether pathname
1032 * should be ignored. The first match (i.e. the last on the list), if
1033 * any, determines the fate. Returns the exclude_list element which
1034 * matched, or NULL for undecided.
1036 static struct exclude
*last_exclude_matching_from_list(const char *pathname
,
1038 const char *basename
,
1040 struct exclude_list
*el
,
1041 struct index_state
*istate
)
1043 struct exclude
*exc
= NULL
; /* undecided */
1047 return NULL
; /* undefined */
1049 for (i
= el
->nr
- 1; 0 <= i
; i
--) {
1050 struct exclude
*x
= el
->excludes
[i
];
1051 const char *exclude
= x
->pattern
;
1052 int prefix
= x
->nowildcardlen
;
1054 if (x
->flags
& EXC_FLAG_MUSTBEDIR
) {
1055 if (*dtype
== DT_UNKNOWN
)
1056 *dtype
= get_dtype(NULL
, istate
, pathname
, pathlen
);
1057 if (*dtype
!= DT_DIR
)
1061 if (x
->flags
& EXC_FLAG_NODIR
) {
1062 if (match_basename(basename
,
1063 pathlen
- (basename
- pathname
),
1064 exclude
, prefix
, x
->patternlen
,
1072 assert(x
->baselen
== 0 || x
->base
[x
->baselen
- 1] == '/');
1073 if (match_pathname(pathname
, pathlen
,
1074 x
->base
, x
->baselen
? x
->baselen
- 1 : 0,
1075 exclude
, prefix
, x
->patternlen
, x
->flags
)) {
1084 * Scan the list and let the last match determine the fate.
1085 * Return 1 for exclude, 0 for include and -1 for undecided.
1087 int is_excluded_from_list(const char *pathname
,
1088 int pathlen
, const char *basename
, int *dtype
,
1089 struct exclude_list
*el
, struct index_state
*istate
)
1091 struct exclude
*exclude
;
1092 exclude
= last_exclude_matching_from_list(pathname
, pathlen
, basename
,
1095 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
1096 return -1; /* undecided */
1099 static struct exclude
*last_exclude_matching_from_lists(struct dir_struct
*dir
,
1100 struct index_state
*istate
,
1101 const char *pathname
, int pathlen
, const char *basename
,
1105 struct exclude_list_group
*group
;
1106 struct exclude
*exclude
;
1107 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1108 group
= &dir
->exclude_list_group
[i
];
1109 for (j
= group
->nr
- 1; j
>= 0; j
--) {
1110 exclude
= last_exclude_matching_from_list(
1111 pathname
, pathlen
, basename
, dtype_p
,
1112 &group
->el
[j
], istate
);
1121 * Loads the per-directory exclude list for the substring of base
1122 * which has a char length of baselen.
1124 static void prep_exclude(struct dir_struct
*dir
,
1125 struct index_state
*istate
,
1126 const char *base
, int baselen
)
1128 struct exclude_list_group
*group
;
1129 struct exclude_list
*el
;
1130 struct exclude_stack
*stk
= NULL
;
1131 struct untracked_cache_dir
*untracked
;
1134 group
= &dir
->exclude_list_group
[EXC_DIRS
];
1137 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
1138 * which originate from directories not in the prefix of the
1139 * path being checked.
1141 while ((stk
= dir
->exclude_stack
) != NULL
) {
1142 if (stk
->baselen
<= baselen
&&
1143 !strncmp(dir
->basebuf
.buf
, base
, stk
->baselen
))
1145 el
= &group
->el
[dir
->exclude_stack
->exclude_ix
];
1146 dir
->exclude_stack
= stk
->prev
;
1147 dir
->exclude
= NULL
;
1148 free((char *)el
->src
); /* see strbuf_detach() below */
1149 clear_exclude_list(el
);
1154 /* Skip traversing into sub directories if the parent is excluded */
1159 * Lazy initialization. All call sites currently just
1160 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
1161 * them seems lots of work for little benefit.
1163 if (!dir
->basebuf
.buf
)
1164 strbuf_init(&dir
->basebuf
, PATH_MAX
);
1166 /* Read from the parent directories and push them down. */
1167 current
= stk
? stk
->baselen
: -1;
1168 strbuf_setlen(&dir
->basebuf
, current
< 0 ? 0 : current
);
1170 untracked
= stk
? stk
->ucd
: dir
->untracked
->root
;
1174 while (current
< baselen
) {
1176 struct oid_stat oid_stat
;
1178 stk
= xcalloc(1, sizeof(*stk
));
1183 cp
= strchr(base
+ current
+ 1, '/');
1185 die("oops in prep_exclude");
1188 lookup_untracked(dir
->untracked
, untracked
,
1190 cp
- base
- current
);
1192 stk
->prev
= dir
->exclude_stack
;
1193 stk
->baselen
= cp
- base
;
1194 stk
->exclude_ix
= group
->nr
;
1195 stk
->ucd
= untracked
;
1196 el
= add_exclude_list(dir
, EXC_DIRS
, NULL
);
1197 strbuf_add(&dir
->basebuf
, base
+ current
, stk
->baselen
- current
);
1198 assert(stk
->baselen
== dir
->basebuf
.len
);
1200 /* Abort if the directory is excluded */
1203 dir
->basebuf
.buf
[stk
->baselen
- 1] = 0;
1204 dir
->exclude
= last_exclude_matching_from_lists(dir
,
1206 dir
->basebuf
.buf
, stk
->baselen
- 1,
1207 dir
->basebuf
.buf
+ current
, &dt
);
1208 dir
->basebuf
.buf
[stk
->baselen
- 1] = '/';
1210 dir
->exclude
->flags
& EXC_FLAG_NEGATIVE
)
1211 dir
->exclude
= NULL
;
1213 dir
->exclude_stack
= stk
;
1218 /* Try to read per-directory file */
1219 oidclr(&oid_stat
.oid
);
1221 if (dir
->exclude_per_dir
&&
1223 * If we know that no files have been added in
1224 * this directory (i.e. valid_cached_dir() has
1225 * been executed and set untracked->valid) ..
1227 (!untracked
|| !untracked
->valid
||
1229 * .. and .gitignore does not exist before
1230 * (i.e. null exclude_oid). Then we can skip
1231 * loading .gitignore, which would result in
1234 !is_null_oid(&untracked
->exclude_oid
))) {
1236 * dir->basebuf gets reused by the traversal, but we
1237 * need fname to remain unchanged to ensure the src
1238 * member of each struct exclude correctly
1239 * back-references its source file. Other invocations
1240 * of add_exclude_list provide stable strings, so we
1241 * strbuf_detach() and free() here in the caller.
1243 struct strbuf sb
= STRBUF_INIT
;
1244 strbuf_addbuf(&sb
, &dir
->basebuf
);
1245 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
1246 el
->src
= strbuf_detach(&sb
, NULL
);
1247 add_excludes(el
->src
, el
->src
, stk
->baselen
, el
, istate
,
1248 untracked
? &oid_stat
: NULL
);
1251 * NEEDSWORK: when untracked cache is enabled, prep_exclude()
1252 * will first be called in valid_cached_dir() then maybe many
1253 * times more in last_exclude_matching(). When the cache is
1254 * used, last_exclude_matching() will not be called and
1255 * reading .gitignore content will be a waste.
1257 * So when it's called by valid_cached_dir() and we can get
1258 * .gitignore SHA-1 from the index (i.e. .gitignore is not
1259 * modified on work tree), we could delay reading the
1260 * .gitignore content until we absolutely need it in
1261 * last_exclude_matching(). Be careful about ignore rule
1262 * order, though, if you do that.
1265 !oideq(&oid_stat
.oid
, &untracked
->exclude_oid
)) {
1266 invalidate_gitignore(dir
->untracked
, untracked
);
1267 oidcpy(&untracked
->exclude_oid
, &oid_stat
.oid
);
1269 dir
->exclude_stack
= stk
;
1270 current
= stk
->baselen
;
1272 strbuf_setlen(&dir
->basebuf
, baselen
);
1276 * Loads the exclude lists for the directory containing pathname, then
1277 * scans all exclude lists to determine whether pathname is excluded.
1278 * Returns the exclude_list element which matched, or NULL for
1281 struct exclude
*last_exclude_matching(struct dir_struct
*dir
,
1282 struct index_state
*istate
,
1283 const char *pathname
,
1286 int pathlen
= strlen(pathname
);
1287 const char *basename
= strrchr(pathname
, '/');
1288 basename
= (basename
) ? basename
+1 : pathname
;
1290 prep_exclude(dir
, istate
, pathname
, basename
-pathname
);
1293 return dir
->exclude
;
1295 return last_exclude_matching_from_lists(dir
, istate
, pathname
, pathlen
,
1300 * Loads the exclude lists for the directory containing pathname, then
1301 * scans all exclude lists to determine whether pathname is excluded.
1302 * Returns 1 if true, otherwise 0.
1304 int is_excluded(struct dir_struct
*dir
, struct index_state
*istate
,
1305 const char *pathname
, int *dtype_p
)
1307 struct exclude
*exclude
=
1308 last_exclude_matching(dir
, istate
, pathname
, dtype_p
);
1310 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
1314 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
1316 struct dir_entry
*ent
;
1318 FLEX_ALLOC_MEM(ent
, name
, pathname
, len
);
1323 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
,
1324 struct index_state
*istate
,
1325 const char *pathname
, int len
)
1327 if (index_file_exists(istate
, pathname
, len
, ignore_case
))
1330 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->alloc
);
1331 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
1334 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
,
1335 struct index_state
*istate
,
1336 const char *pathname
, int len
)
1338 if (!index_name_is_other(istate
, pathname
, len
))
1341 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->ignored_alloc
);
1342 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1346 index_nonexistent
= 0,
1352 * Do not use the alphabetically sorted index to look up
1353 * the directory name; instead, use the case insensitive
1356 static enum exist_status
directory_exists_in_index_icase(struct index_state
*istate
,
1357 const char *dirname
, int len
)
1359 struct cache_entry
*ce
;
1361 if (index_dir_exists(istate
, dirname
, len
))
1362 return index_directory
;
1364 ce
= index_file_exists(istate
, dirname
, len
, ignore_case
);
1365 if (ce
&& S_ISGITLINK(ce
->ce_mode
))
1366 return index_gitdir
;
1368 return index_nonexistent
;
1372 * The index sorts alphabetically by entry name, which
1373 * means that a gitlink sorts as '\0' at the end, while
1374 * a directory (which is defined not as an entry, but as
1375 * the files it contains) will sort with the '/' at the
1378 static enum exist_status
directory_exists_in_index(struct index_state
*istate
,
1379 const char *dirname
, int len
)
1384 return directory_exists_in_index_icase(istate
, dirname
, len
);
1386 pos
= index_name_pos(istate
, dirname
, len
);
1389 while (pos
< istate
->cache_nr
) {
1390 const struct cache_entry
*ce
= istate
->cache
[pos
++];
1391 unsigned char endchar
;
1393 if (strncmp(ce
->name
, dirname
, len
))
1395 endchar
= ce
->name
[len
];
1399 return index_directory
;
1400 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1401 return index_gitdir
;
1403 return index_nonexistent
;
1407 * When we find a directory when traversing the filesystem, we
1408 * have three distinct cases:
1411 * - see it as a directory
1414 * and which one we choose depends on a combination of existing
1415 * git index contents and the flags passed into the directory
1416 * traversal routine.
1418 * Case 1: If we *already* have entries in the index under that
1419 * directory name, we always recurse into the directory to see
1422 * Case 2: If we *already* have that directory name as a gitlink,
1423 * we always continue to see it as a gitlink, regardless of whether
1424 * there is an actual git directory there or not (it might not
1425 * be checked out as a subproject!)
1427 * Case 3: if we didn't have it in the index previously, we
1428 * have a few sub-cases:
1430 * (a) if "show_other_directories" is true, we show it as
1431 * just a directory, unless "hide_empty_directories" is
1432 * also true, in which case we need to check if it contains any
1433 * untracked and / or ignored files.
1434 * (b) if it looks like a git directory, and we don't have
1435 * 'no_gitlinks' set we treat it as a gitlink, and show it
1437 * (c) otherwise, we recurse into it.
1439 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1440 struct index_state
*istate
,
1441 struct untracked_cache_dir
*untracked
,
1442 const char *dirname
, int len
, int baselen
, int exclude
,
1443 const struct pathspec
*pathspec
)
1445 /* The "len-1" is to strip the final '/' */
1446 switch (directory_exists_in_index(istate
, dirname
, len
-1)) {
1447 case index_directory
:
1448 return path_recurse
;
1453 case index_nonexistent
:
1454 if (dir
->flags
& DIR_SKIP_NESTED_GIT
) {
1456 struct strbuf sb
= STRBUF_INIT
;
1457 strbuf_addstr(&sb
, dirname
);
1458 nested_repo
= is_nonbare_repository_dir(&sb
);
1459 strbuf_release(&sb
);
1464 if (dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)
1467 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1468 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
)) {
1471 * This is an excluded directory and we are
1472 * showing ignored paths that match an exclude
1473 * pattern. (e.g. show directory as ignored
1474 * only if it matches an exclude pattern).
1475 * This path will either be 'path_excluded`
1476 * (if we are showing empty directories or if
1477 * the directory is not empty), or will be
1478 * 'path_none' (empty directory, and we are
1479 * not showing empty directories).
1481 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1482 return path_excluded
;
1484 if (read_directory_recursive(dir
, istate
, dirname
, len
,
1485 untracked
, 1, 1, pathspec
) == path_excluded
)
1486 return path_excluded
;
1490 if (!(dir
->flags
& DIR_NO_GITLINKS
)) {
1491 struct strbuf sb
= STRBUF_INIT
;
1492 strbuf_addstr(&sb
, dirname
);
1493 if (is_nonbare_repository_dir(&sb
))
1494 return exclude
? path_excluded
: path_untracked
;
1495 strbuf_release(&sb
);
1497 return path_recurse
;
1500 /* This is the "show_other_directories" case */
1502 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1503 return exclude
? path_excluded
: path_untracked
;
1505 untracked
= lookup_untracked(dir
->untracked
, untracked
,
1506 dirname
+ baselen
, len
- baselen
);
1509 * If this is an excluded directory, then we only need to check if
1510 * the directory contains any files.
1512 return read_directory_recursive(dir
, istate
, dirname
, len
,
1513 untracked
, 1, exclude
, pathspec
);
1517 * This is an inexact early pruning of any recursive directory
1518 * reading - if the path cannot possibly be in the pathspec,
1519 * return true, and we'll skip it early.
1521 static int simplify_away(const char *path
, int pathlen
,
1522 const struct pathspec
*pathspec
)
1526 if (!pathspec
|| !pathspec
->nr
)
1529 GUARD_PATHSPEC(pathspec
,
1538 for (i
= 0; i
< pathspec
->nr
; i
++) {
1539 const struct pathspec_item
*item
= &pathspec
->items
[i
];
1540 int len
= item
->nowildcard_len
;
1544 if (!ps_strncmp(item
, item
->match
, path
, len
))
1552 * This function tells us whether an excluded path matches a
1553 * list of "interesting" pathspecs. That is, whether a path matched
1554 * by any of the pathspecs could possibly be ignored by excluding
1555 * the specified path. This can happen if:
1557 * 1. the path is mentioned explicitly in the pathspec
1559 * 2. the path is a directory prefix of some element in the
1562 static int exclude_matches_pathspec(const char *path
, int pathlen
,
1563 const struct pathspec
*pathspec
)
1567 if (!pathspec
|| !pathspec
->nr
)
1570 GUARD_PATHSPEC(pathspec
,
1578 for (i
= 0; i
< pathspec
->nr
; i
++) {
1579 const struct pathspec_item
*item
= &pathspec
->items
[i
];
1580 int len
= item
->nowildcard_len
;
1582 if (len
== pathlen
&&
1583 !ps_strncmp(item
, item
->match
, path
, pathlen
))
1585 if (len
> pathlen
&&
1586 item
->match
[pathlen
] == '/' &&
1587 !ps_strncmp(item
, item
->match
, path
, pathlen
))
1593 static int get_index_dtype(struct index_state
*istate
,
1594 const char *path
, int len
)
1597 const struct cache_entry
*ce
;
1599 ce
= index_file_exists(istate
, path
, len
, 0);
1601 if (!ce_uptodate(ce
))
1603 if (S_ISGITLINK(ce
->ce_mode
))
1606 * Nobody actually cares about the
1607 * difference between DT_LNK and DT_REG
1612 /* Try to look it up as a directory */
1613 pos
= index_name_pos(istate
, path
, len
);
1617 while (pos
< istate
->cache_nr
) {
1618 ce
= istate
->cache
[pos
++];
1619 if (strncmp(ce
->name
, path
, len
))
1621 if (ce
->name
[len
] > '/')
1623 if (ce
->name
[len
] < '/')
1625 if (!ce_uptodate(ce
))
1626 break; /* continue? */
1632 static int get_dtype(struct dirent
*de
, struct index_state
*istate
,
1633 const char *path
, int len
)
1635 int dtype
= de
? DTYPE(de
) : DT_UNKNOWN
;
1638 if (dtype
!= DT_UNKNOWN
)
1640 dtype
= get_index_dtype(istate
, path
, len
);
1641 if (dtype
!= DT_UNKNOWN
)
1643 if (lstat(path
, &st
))
1645 if (S_ISREG(st
.st_mode
))
1647 if (S_ISDIR(st
.st_mode
))
1649 if (S_ISLNK(st
.st_mode
))
1654 static enum path_treatment
treat_one_path(struct dir_struct
*dir
,
1655 struct untracked_cache_dir
*untracked
,
1656 struct index_state
*istate
,
1657 struct strbuf
*path
,
1659 const struct pathspec
*pathspec
,
1660 int dtype
, struct dirent
*de
)
1663 int has_path_in_index
= !!index_file_exists(istate
, path
->buf
, path
->len
, ignore_case
);
1664 enum path_treatment path_treatment
;
1666 if (dtype
== DT_UNKNOWN
)
1667 dtype
= get_dtype(de
, istate
, path
->buf
, path
->len
);
1669 /* Always exclude indexed files */
1670 if (dtype
!= DT_DIR
&& has_path_in_index
)
1674 * When we are looking at a directory P in the working tree,
1675 * there are three cases:
1677 * (1) P exists in the index. Everything inside the directory P in
1678 * the working tree needs to go when P is checked out from the
1681 * (2) P does not exist in the index, but there is P/Q in the index.
1682 * We know P will stay a directory when we check out the contents
1683 * of the index, but we do not know yet if there is a directory
1684 * P/Q in the working tree to be killed, so we need to recurse.
1686 * (3) P does not exist in the index, and there is no P/Q in the index
1687 * to require P to be a directory, either. Only in this case, we
1688 * know that everything inside P will not be killed without
1691 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
1692 (dtype
== DT_DIR
) &&
1693 !has_path_in_index
&&
1694 (directory_exists_in_index(istate
, path
->buf
, path
->len
) == index_nonexistent
))
1697 exclude
= is_excluded(dir
, istate
, path
->buf
, &dtype
);
1700 * Excluded? If we don't explicitly want to show
1701 * ignored files, ignore it
1703 if (exclude
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
1704 return path_excluded
;
1710 strbuf_addch(path
, '/');
1711 path_treatment
= treat_directory(dir
, istate
, untracked
,
1712 path
->buf
, path
->len
,
1713 baselen
, exclude
, pathspec
);
1715 * If 1) we only want to return directories that
1716 * match an exclude pattern and 2) this directory does
1717 * not match an exclude pattern but all of its
1718 * contents are excluded, then indicate that we should
1719 * recurse into this directory (instead of marking the
1720 * directory itself as an ignored path).
1723 path_treatment
== path_excluded
&&
1724 (dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
1725 (dir
->flags
& DIR_SHOW_IGNORED_TOO_MODE_MATCHING
))
1726 return path_recurse
;
1727 return path_treatment
;
1730 return exclude
? path_excluded
: path_untracked
;
1734 static enum path_treatment
treat_path_fast(struct dir_struct
*dir
,
1735 struct untracked_cache_dir
*untracked
,
1736 struct cached_dir
*cdir
,
1737 struct index_state
*istate
,
1738 struct strbuf
*path
,
1740 const struct pathspec
*pathspec
)
1742 strbuf_setlen(path
, baselen
);
1744 strbuf_addstr(path
, cdir
->file
);
1745 return path_untracked
;
1747 strbuf_addstr(path
, cdir
->ucd
->name
);
1748 /* treat_one_path() does this before it calls treat_directory() */
1749 strbuf_complete(path
, '/');
1750 if (cdir
->ucd
->check_only
)
1752 * check_only is set as a result of treat_directory() getting
1753 * to its bottom. Verify again the same set of directories
1754 * with check_only set.
1756 return read_directory_recursive(dir
, istate
, path
->buf
, path
->len
,
1757 cdir
->ucd
, 1, 0, pathspec
);
1759 * We get path_recurse in the first run when
1760 * directory_exists_in_index() returns index_nonexistent. We
1761 * are sure that new changes in the index does not impact the
1762 * outcome. Return now.
1764 return path_recurse
;
1767 static enum path_treatment
treat_path(struct dir_struct
*dir
,
1768 struct untracked_cache_dir
*untracked
,
1769 struct cached_dir
*cdir
,
1770 struct index_state
*istate
,
1771 struct strbuf
*path
,
1773 const struct pathspec
*pathspec
)
1776 struct dirent
*de
= cdir
->de
;
1779 return treat_path_fast(dir
, untracked
, cdir
, istate
, path
,
1781 if (is_dot_or_dotdot(de
->d_name
) || !fspathcmp(de
->d_name
, ".git"))
1783 strbuf_setlen(path
, baselen
);
1784 strbuf_addstr(path
, de
->d_name
);
1785 if (simplify_away(path
->buf
, path
->len
, pathspec
))
1789 return treat_one_path(dir
, untracked
, istate
, path
, baselen
, pathspec
, dtype
, de
);
1792 static void add_untracked(struct untracked_cache_dir
*dir
, const char *name
)
1796 ALLOC_GROW(dir
->untracked
, dir
->untracked_nr
+ 1,
1797 dir
->untracked_alloc
);
1798 dir
->untracked
[dir
->untracked_nr
++] = xstrdup(name
);
1801 static int valid_cached_dir(struct dir_struct
*dir
,
1802 struct untracked_cache_dir
*untracked
,
1803 struct index_state
*istate
,
1804 struct strbuf
*path
,
1813 * With fsmonitor, we can trust the untracked cache's valid field.
1815 refresh_fsmonitor(istate
);
1816 if (!(dir
->untracked
->use_fsmonitor
&& untracked
->valid
)) {
1817 if (lstat(path
->len
? path
->buf
: ".", &st
)) {
1818 memset(&untracked
->stat_data
, 0, sizeof(untracked
->stat_data
));
1821 if (!untracked
->valid
||
1822 match_stat_data_racy(istate
, &untracked
->stat_data
, &st
)) {
1823 fill_stat_data(&untracked
->stat_data
, &st
);
1828 if (untracked
->check_only
!= !!check_only
)
1832 * prep_exclude will be called eventually on this directory,
1833 * but it's called much later in last_exclude_matching(). We
1834 * need it now to determine the validity of the cache for this
1835 * path. The next calls will be nearly no-op, the way
1836 * prep_exclude() is designed.
1838 if (path
->len
&& path
->buf
[path
->len
- 1] != '/') {
1839 strbuf_addch(path
, '/');
1840 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
1841 strbuf_setlen(path
, path
->len
- 1);
1843 prep_exclude(dir
, istate
, path
->buf
, path
->len
);
1845 /* hopefully prep_exclude() haven't invalidated this entry... */
1846 return untracked
->valid
;
1849 static int open_cached_dir(struct cached_dir
*cdir
,
1850 struct dir_struct
*dir
,
1851 struct untracked_cache_dir
*untracked
,
1852 struct index_state
*istate
,
1853 struct strbuf
*path
,
1858 memset(cdir
, 0, sizeof(*cdir
));
1859 cdir
->untracked
= untracked
;
1860 if (valid_cached_dir(dir
, untracked
, istate
, path
, check_only
))
1862 c_path
= path
->len
? path
->buf
: ".";
1863 cdir
->fdir
= opendir(c_path
);
1865 warning_errno(_("could not open directory '%s'"), c_path
);
1866 if (dir
->untracked
) {
1867 invalidate_directory(dir
->untracked
, untracked
);
1868 dir
->untracked
->dir_opened
++;
1875 static int read_cached_dir(struct cached_dir
*cdir
)
1878 cdir
->de
= readdir(cdir
->fdir
);
1883 while (cdir
->nr_dirs
< cdir
->untracked
->dirs_nr
) {
1884 struct untracked_cache_dir
*d
= cdir
->untracked
->dirs
[cdir
->nr_dirs
];
1894 if (cdir
->nr_files
< cdir
->untracked
->untracked_nr
) {
1895 struct untracked_cache_dir
*d
= cdir
->untracked
;
1896 cdir
->file
= d
->untracked
[cdir
->nr_files
++];
1902 static void close_cached_dir(struct cached_dir
*cdir
)
1905 closedir(cdir
->fdir
);
1907 * We have gone through this directory and found no untracked
1908 * entries. Mark it valid.
1910 if (cdir
->untracked
) {
1911 cdir
->untracked
->valid
= 1;
1912 cdir
->untracked
->recurse
= 1;
1917 * Read a directory tree. We currently ignore anything but
1918 * directories, regular files and symlinks. That's because git
1919 * doesn't handle them at all yet. Maybe that will change some
1922 * Also, we ignore the name ".git" (even if it is not a directory).
1923 * That likely will not change.
1925 * If 'stop_at_first_file' is specified, 'path_excluded' is returned
1926 * to signal that a file was found. This is the least significant value that
1927 * indicates that a file was encountered that does not depend on the order of
1928 * whether an untracked or exluded path was encountered first.
1930 * Returns the most significant path_treatment value encountered in the scan.
1931 * If 'stop_at_first_file' is specified, `path_excluded` is the most
1932 * significant path_treatment value that will be returned.
1935 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
1936 struct index_state
*istate
, const char *base
, int baselen
,
1937 struct untracked_cache_dir
*untracked
, int check_only
,
1938 int stop_at_first_file
, const struct pathspec
*pathspec
)
1940 struct cached_dir cdir
;
1941 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
1942 struct strbuf path
= STRBUF_INIT
;
1944 strbuf_add(&path
, base
, baselen
);
1946 if (open_cached_dir(&cdir
, dir
, untracked
, istate
, &path
, check_only
))
1950 untracked
->check_only
= !!check_only
;
1952 while (!read_cached_dir(&cdir
)) {
1953 /* check how the file or directory should be treated */
1954 state
= treat_path(dir
, untracked
, &cdir
, istate
, &path
,
1957 if (state
> dir_state
)
1960 /* recurse into subdir if instructed by treat_path */
1961 if ((state
== path_recurse
) ||
1962 ((state
== path_untracked
) &&
1963 (get_dtype(cdir
.de
, istate
, path
.buf
, path
.len
) == DT_DIR
) &&
1964 ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
1966 do_match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
1967 baselen
, NULL
, DO_MATCH_LEADING_PATHSPEC
) == MATCHED_RECURSIVELY_LEADING_PATHSPEC
)))) {
1968 struct untracked_cache_dir
*ud
;
1969 ud
= lookup_untracked(dir
->untracked
, untracked
,
1971 path
.len
- baselen
);
1973 read_directory_recursive(dir
, istate
, path
.buf
,
1975 check_only
, stop_at_first_file
, pathspec
);
1976 if (subdir_state
> dir_state
)
1977 dir_state
= subdir_state
;
1980 !match_pathspec(istate
, pathspec
, path
.buf
, path
.len
,
1981 0 /* prefix */, NULL
,
1982 0 /* do NOT special case dirs */))
1987 if (stop_at_first_file
) {
1989 * If stopping at first file, then
1990 * signal that a file was found by
1991 * returning `path_excluded`. This is
1992 * to return a consistent value
1993 * regardless of whether an ignored or
1994 * excluded file happened to be
1997 * In current usage, the
1998 * `stop_at_first_file` is passed when
1999 * an ancestor directory has matched
2000 * an exclude pattern, so any found
2001 * files will be excluded.
2003 if (dir_state
>= path_excluded
) {
2004 dir_state
= path_excluded
;
2009 /* abort early if maximum state has been reached */
2010 if (dir_state
== path_untracked
) {
2012 add_untracked(untracked
, path
.buf
+ baselen
);
2015 /* skip the dir_add_* part */
2019 /* add the path to the appropriate result list */
2022 if (dir
->flags
& DIR_SHOW_IGNORED
)
2023 dir_add_name(dir
, istate
, path
.buf
, path
.len
);
2024 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
2025 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
2026 exclude_matches_pathspec(path
.buf
, path
.len
,
2028 dir_add_ignored(dir
, istate
, path
.buf
, path
.len
);
2031 case path_untracked
:
2032 if (dir
->flags
& DIR_SHOW_IGNORED
)
2034 dir_add_name(dir
, istate
, path
.buf
, path
.len
);
2036 add_untracked(untracked
, path
.buf
+ baselen
);
2043 close_cached_dir(&cdir
);
2045 strbuf_release(&path
);
2050 int cmp_dir_entry(const void *p1
, const void *p2
)
2052 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
2053 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
2055 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
2058 /* check if *out lexically strictly contains *in */
2059 int check_dir_entry_contains(const struct dir_entry
*out
, const struct dir_entry
*in
)
2061 return (out
->len
< in
->len
) &&
2062 (out
->name
[out
->len
- 1] == '/') &&
2063 !memcmp(out
->name
, in
->name
, out
->len
);
2066 static int treat_leading_path(struct dir_struct
*dir
,
2067 struct index_state
*istate
,
2068 const char *path
, int len
,
2069 const struct pathspec
*pathspec
)
2071 struct strbuf sb
= STRBUF_INIT
;
2072 int baselen
, rc
= 0;
2074 int old_flags
= dir
->flags
;
2076 while (len
&& path
[len
- 1] == '/')
2081 dir
->flags
&= ~DIR_SHOW_OTHER_DIRECTORIES
;
2083 cp
= path
+ baselen
+ !!baselen
;
2084 cp
= memchr(cp
, '/', path
+ len
- cp
);
2088 baselen
= cp
- path
;
2089 strbuf_setlen(&sb
, 0);
2090 strbuf_add(&sb
, path
, baselen
);
2091 if (!is_directory(sb
.buf
))
2093 if (simplify_away(sb
.buf
, sb
.len
, pathspec
))
2095 if (treat_one_path(dir
, NULL
, istate
, &sb
, baselen
, pathspec
,
2096 DT_DIR
, NULL
) == path_none
)
2097 break; /* do not recurse into it */
2098 if (len
<= baselen
) {
2100 break; /* finished checking */
2103 strbuf_release(&sb
);
2104 dir
->flags
= old_flags
;
2108 static const char *get_ident_string(void)
2110 static struct strbuf sb
= STRBUF_INIT
;
2115 if (uname(&uts
) < 0)
2116 die_errno(_("failed to get kernel name and information"));
2117 strbuf_addf(&sb
, "Location %s, system %s", get_git_work_tree(),
2122 static int ident_in_untracked(const struct untracked_cache
*uc
)
2125 * Previous git versions may have saved many NUL separated
2126 * strings in the "ident" field, but it is insane to manage
2127 * many locations, so just take care of the first one.
2130 return !strcmp(uc
->ident
.buf
, get_ident_string());
2133 static void set_untracked_ident(struct untracked_cache
*uc
)
2135 strbuf_reset(&uc
->ident
);
2136 strbuf_addstr(&uc
->ident
, get_ident_string());
2139 * This strbuf used to contain a list of NUL separated
2140 * strings, so save NUL too for backward compatibility.
2142 strbuf_addch(&uc
->ident
, 0);
2145 static void new_untracked_cache(struct index_state
*istate
)
2147 struct untracked_cache
*uc
= xcalloc(1, sizeof(*uc
));
2148 strbuf_init(&uc
->ident
, 100);
2149 uc
->exclude_per_dir
= ".gitignore";
2150 /* should be the same flags used by git-status */
2151 uc
->dir_flags
= DIR_SHOW_OTHER_DIRECTORIES
| DIR_HIDE_EMPTY_DIRECTORIES
;
2152 set_untracked_ident(uc
);
2153 istate
->untracked
= uc
;
2154 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2157 void add_untracked_cache(struct index_state
*istate
)
2159 if (!istate
->untracked
) {
2160 new_untracked_cache(istate
);
2162 if (!ident_in_untracked(istate
->untracked
)) {
2163 free_untracked_cache(istate
->untracked
);
2164 new_untracked_cache(istate
);
2169 void remove_untracked_cache(struct index_state
*istate
)
2171 if (istate
->untracked
) {
2172 free_untracked_cache(istate
->untracked
);
2173 istate
->untracked
= NULL
;
2174 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2178 static struct untracked_cache_dir
*validate_untracked_cache(struct dir_struct
*dir
,
2180 const struct pathspec
*pathspec
)
2182 struct untracked_cache_dir
*root
;
2183 static int untracked_cache_disabled
= -1;
2185 if (!dir
->untracked
)
2187 if (untracked_cache_disabled
< 0)
2188 untracked_cache_disabled
= git_env_bool("GIT_DISABLE_UNTRACKED_CACHE", 0);
2189 if (untracked_cache_disabled
)
2193 * We only support $GIT_DIR/info/exclude and core.excludesfile
2194 * as the global ignore rule files. Any other additions
2195 * (e.g. from command line) invalidate the cache. This
2196 * condition also catches running setup_standard_excludes()
2197 * before setting dir->untracked!
2199 if (dir
->unmanaged_exclude_files
)
2203 * Optimize for the main use case only: whole-tree git
2204 * status. More work involved in treat_leading_path() if we
2205 * use cache on just a subset of the worktree. pathspec
2206 * support could make the matter even worse.
2208 if (base_len
|| (pathspec
&& pathspec
->nr
))
2211 /* Different set of flags may produce different results */
2212 if (dir
->flags
!= dir
->untracked
->dir_flags
||
2214 * See treat_directory(), case index_nonexistent. Without
2215 * this flag, we may need to also cache .git file content
2216 * for the resolve_gitlink_ref() call, which we don't.
2218 !(dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
) ||
2219 /* We don't support collecting ignore files */
2220 (dir
->flags
& (DIR_SHOW_IGNORED
| DIR_SHOW_IGNORED_TOO
|
2221 DIR_COLLECT_IGNORED
)))
2225 * If we use .gitignore in the cache and now you change it to
2226 * .gitexclude, everything will go wrong.
2228 if (dir
->exclude_per_dir
!= dir
->untracked
->exclude_per_dir
&&
2229 strcmp(dir
->exclude_per_dir
, dir
->untracked
->exclude_per_dir
))
2233 * EXC_CMDL is not considered in the cache. If people set it,
2236 if (dir
->exclude_list_group
[EXC_CMDL
].nr
)
2239 if (!ident_in_untracked(dir
->untracked
)) {
2240 warning(_("untracked cache is disabled on this system or location"));
2244 if (!dir
->untracked
->root
) {
2245 const int len
= sizeof(*dir
->untracked
->root
);
2246 dir
->untracked
->root
= xmalloc(len
);
2247 memset(dir
->untracked
->root
, 0, len
);
2250 /* Validate $GIT_DIR/info/exclude and core.excludesfile */
2251 root
= dir
->untracked
->root
;
2252 if (!oideq(&dir
->ss_info_exclude
.oid
,
2253 &dir
->untracked
->ss_info_exclude
.oid
)) {
2254 invalidate_gitignore(dir
->untracked
, root
);
2255 dir
->untracked
->ss_info_exclude
= dir
->ss_info_exclude
;
2257 if (!oideq(&dir
->ss_excludes_file
.oid
,
2258 &dir
->untracked
->ss_excludes_file
.oid
)) {
2259 invalidate_gitignore(dir
->untracked
, root
);
2260 dir
->untracked
->ss_excludes_file
= dir
->ss_excludes_file
;
2263 /* Make sure this directory is not dropped out at saving phase */
2268 int read_directory(struct dir_struct
*dir
, struct index_state
*istate
,
2269 const char *path
, int len
, const struct pathspec
*pathspec
)
2271 struct untracked_cache_dir
*untracked
;
2273 trace_performance_enter();
2275 if (has_symlink_leading_path(path
, len
)) {
2276 trace_performance_leave("read directory %.*s", len
, path
);
2280 untracked
= validate_untracked_cache(dir
, len
, pathspec
);
2283 * make sure untracked cache code path is disabled,
2284 * e.g. prep_exclude()
2286 dir
->untracked
= NULL
;
2287 if (!len
|| treat_leading_path(dir
, istate
, path
, len
, pathspec
))
2288 read_directory_recursive(dir
, istate
, path
, len
, untracked
, 0, 0, pathspec
);
2289 QSORT(dir
->entries
, dir
->nr
, cmp_dir_entry
);
2290 QSORT(dir
->ignored
, dir
->ignored_nr
, cmp_dir_entry
);
2293 * If DIR_SHOW_IGNORED_TOO is set, read_directory_recursive() will
2294 * also pick up untracked contents of untracked dirs; by default
2295 * we discard these, but given DIR_KEEP_UNTRACKED_CONTENTS we do not.
2297 if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) &&
2298 !(dir
->flags
& DIR_KEEP_UNTRACKED_CONTENTS
)) {
2301 /* remove from dir->entries untracked contents of untracked dirs */
2302 for (i
= j
= 0; j
< dir
->nr
; j
++) {
2304 check_dir_entry_contains(dir
->entries
[i
- 1], dir
->entries
[j
])) {
2305 FREE_AND_NULL(dir
->entries
[j
]);
2307 dir
->entries
[i
++] = dir
->entries
[j
];
2314 trace_performance_leave("read directory %.*s", len
, path
);
2315 if (dir
->untracked
) {
2316 static int force_untracked_cache
= -1;
2317 static struct trace_key trace_untracked_stats
= TRACE_KEY_INIT(UNTRACKED_STATS
);
2319 if (force_untracked_cache
< 0)
2320 force_untracked_cache
=
2321 git_env_bool("GIT_FORCE_UNTRACKED_CACHE", 0);
2322 trace_printf_key(&trace_untracked_stats
,
2323 "node creation: %u\n"
2324 "gitignore invalidation: %u\n"
2325 "directory invalidation: %u\n"
2327 dir
->untracked
->dir_created
,
2328 dir
->untracked
->gitignore_invalidated
,
2329 dir
->untracked
->dir_invalidated
,
2330 dir
->untracked
->dir_opened
);
2331 if (force_untracked_cache
&&
2332 dir
->untracked
== istate
->untracked
&&
2333 (dir
->untracked
->dir_opened
||
2334 dir
->untracked
->gitignore_invalidated
||
2335 dir
->untracked
->dir_invalidated
))
2336 istate
->cache_changed
|= UNTRACKED_CHANGED
;
2337 if (dir
->untracked
!= istate
->untracked
) {
2338 FREE_AND_NULL(dir
->untracked
);
2344 int file_exists(const char *f
)
2347 return lstat(f
, &sb
) == 0;
2350 int repo_file_exists(struct repository
*repo
, const char *path
)
2352 if (repo
!= the_repository
)
2353 BUG("do not know how to check file existence in arbitrary repo");
2355 return file_exists(path
);
2358 static int cmp_icase(char a
, char b
)
2363 return toupper(a
) - toupper(b
);
2368 * Given two normalized paths (a trailing slash is ok), if subdir is
2369 * outside dir, return -1. Otherwise return the offset in subdir that
2370 * can be used as relative path to dir.
2372 int dir_inside_of(const char *subdir
, const char *dir
)
2376 assert(dir
&& subdir
&& *dir
&& *subdir
);
2378 while (*dir
&& *subdir
&& !cmp_icase(*dir
, *subdir
)) {
2384 /* hel[p]/me vs hel[l]/yeah */
2385 if (*dir
&& *subdir
)
2389 return !*dir
? offset
: -1; /* same dir */
2391 /* foo/[b]ar vs foo/[] */
2392 if (is_dir_sep(dir
[-1]))
2393 return is_dir_sep(subdir
[-1]) ? offset
: -1;
2395 /* foo[/]bar vs foo[] */
2396 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
2399 int is_inside_dir(const char *dir
)
2408 rc
= (dir_inside_of(cwd
, dir
) >= 0);
2413 int is_empty_dir(const char *path
)
2415 DIR *dir
= opendir(path
);
2422 while ((e
= readdir(dir
)) != NULL
)
2423 if (!is_dot_or_dotdot(e
->d_name
)) {
2432 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
2436 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
2437 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
2438 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
2439 struct object_id submodule_head
;
2441 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
2442 !resolve_gitlink_ref(path
->buf
, "HEAD", &submodule_head
)) {
2443 /* Do not descend and nuke a nested git work tree. */
2449 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
2450 dir
= opendir(path
->buf
);
2452 if (errno
== ENOENT
)
2453 return keep_toplevel
? -1 : 0;
2454 else if (errno
== EACCES
&& !keep_toplevel
)
2456 * An empty dir could be removable even if it
2459 return rmdir(path
->buf
);
2463 strbuf_complete(path
, '/');
2466 while ((e
= readdir(dir
)) != NULL
) {
2468 if (is_dot_or_dotdot(e
->d_name
))
2471 strbuf_setlen(path
, len
);
2472 strbuf_addstr(path
, e
->d_name
);
2473 if (lstat(path
->buf
, &st
)) {
2474 if (errno
== ENOENT
)
2476 * file disappeared, which is what we
2481 } else if (S_ISDIR(st
.st_mode
)) {
2482 if (!remove_dir_recurse(path
, flag
, &kept_down
))
2483 continue; /* happy */
2484 } else if (!only_empty
&&
2485 (!unlink(path
->buf
) || errno
== ENOENT
)) {
2486 continue; /* happy, too */
2489 /* path too long, stat fails, or non-directory still exists */
2495 strbuf_setlen(path
, original_len
);
2496 if (!ret
&& !keep_toplevel
&& !kept_down
)
2497 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
2500 * report the uplevel that it is not an error that we
2501 * did not rmdir() our directory.
2507 int remove_dir_recursively(struct strbuf
*path
, int flag
)
2509 return remove_dir_recurse(path
, flag
, NULL
);
2512 static GIT_PATH_FUNC(git_path_info_exclude
, "info/exclude")
2514 void setup_standard_excludes(struct dir_struct
*dir
)
2516 dir
->exclude_per_dir
= ".gitignore";
2518 /* core.excludesfile defaulting to $XDG_CONFIG_HOME/git/ignore */
2520 excludes_file
= xdg_config_home("ignore");
2521 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
2522 add_excludes_from_file_1(dir
, excludes_file
,
2523 dir
->untracked
? &dir
->ss_excludes_file
: NULL
);
2525 /* per repository user preference */
2526 if (startup_info
->have_repository
) {
2527 const char *path
= git_path_info_exclude();
2528 if (!access_or_warn(path
, R_OK
, 0))
2529 add_excludes_from_file_1(dir
, path
,
2530 dir
->untracked
? &dir
->ss_info_exclude
: NULL
);
2534 int remove_path(const char *name
)
2538 if (unlink(name
) && !is_missing_file_error(errno
))
2541 slash
= strrchr(name
, '/');
2543 char *dirs
= xstrdup(name
);
2544 slash
= dirs
+ (slash
- name
);
2547 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
2554 * Frees memory within dir which was allocated for exclude lists and
2555 * the exclude_stack. Does not free dir itself.
2557 void clear_directory(struct dir_struct
*dir
)
2560 struct exclude_list_group
*group
;
2561 struct exclude_list
*el
;
2562 struct exclude_stack
*stk
;
2564 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
2565 group
= &dir
->exclude_list_group
[i
];
2566 for (j
= 0; j
< group
->nr
; j
++) {
2569 free((char *)el
->src
);
2570 clear_exclude_list(el
);
2575 stk
= dir
->exclude_stack
;
2577 struct exclude_stack
*prev
= stk
->prev
;
2581 strbuf_release(&dir
->basebuf
);
2584 struct ondisk_untracked_cache
{
2585 struct stat_data info_exclude_stat
;
2586 struct stat_data excludes_file_stat
;
2590 #define ouc_offset(x) offsetof(struct ondisk_untracked_cache, x)
2593 int index
; /* number of written untracked_cache_dir */
2594 struct ewah_bitmap
*check_only
; /* from untracked_cache_dir */
2595 struct ewah_bitmap
*valid
; /* from untracked_cache_dir */
2596 struct ewah_bitmap
*sha1_valid
; /* set if exclude_sha1 is not null */
2598 struct strbuf sb_stat
;
2599 struct strbuf sb_sha1
;
2602 static void stat_data_to_disk(struct stat_data
*to
, const struct stat_data
*from
)
2604 to
->sd_ctime
.sec
= htonl(from
->sd_ctime
.sec
);
2605 to
->sd_ctime
.nsec
= htonl(from
->sd_ctime
.nsec
);
2606 to
->sd_mtime
.sec
= htonl(from
->sd_mtime
.sec
);
2607 to
->sd_mtime
.nsec
= htonl(from
->sd_mtime
.nsec
);
2608 to
->sd_dev
= htonl(from
->sd_dev
);
2609 to
->sd_ino
= htonl(from
->sd_ino
);
2610 to
->sd_uid
= htonl(from
->sd_uid
);
2611 to
->sd_gid
= htonl(from
->sd_gid
);
2612 to
->sd_size
= htonl(from
->sd_size
);
2615 static void write_one_dir(struct untracked_cache_dir
*untracked
,
2616 struct write_data
*wd
)
2618 struct stat_data stat_data
;
2619 struct strbuf
*out
= &wd
->out
;
2620 unsigned char intbuf
[16];
2621 unsigned int intlen
, value
;
2622 int i
= wd
->index
++;
2625 * untracked_nr should be reset whenever valid is clear, but
2628 if (!untracked
->valid
) {
2629 untracked
->untracked_nr
= 0;
2630 untracked
->check_only
= 0;
2633 if (untracked
->check_only
)
2634 ewah_set(wd
->check_only
, i
);
2635 if (untracked
->valid
) {
2636 ewah_set(wd
->valid
, i
);
2637 stat_data_to_disk(&stat_data
, &untracked
->stat_data
);
2638 strbuf_add(&wd
->sb_stat
, &stat_data
, sizeof(stat_data
));
2640 if (!is_null_oid(&untracked
->exclude_oid
)) {
2641 ewah_set(wd
->sha1_valid
, i
);
2642 strbuf_add(&wd
->sb_sha1
, untracked
->exclude_oid
.hash
,
2643 the_hash_algo
->rawsz
);
2646 intlen
= encode_varint(untracked
->untracked_nr
, intbuf
);
2647 strbuf_add(out
, intbuf
, intlen
);
2649 /* skip non-recurse directories */
2650 for (i
= 0, value
= 0; i
< untracked
->dirs_nr
; i
++)
2651 if (untracked
->dirs
[i
]->recurse
)
2653 intlen
= encode_varint(value
, intbuf
);
2654 strbuf_add(out
, intbuf
, intlen
);
2656 strbuf_add(out
, untracked
->name
, strlen(untracked
->name
) + 1);
2658 for (i
= 0; i
< untracked
->untracked_nr
; i
++)
2659 strbuf_add(out
, untracked
->untracked
[i
],
2660 strlen(untracked
->untracked
[i
]) + 1);
2662 for (i
= 0; i
< untracked
->dirs_nr
; i
++)
2663 if (untracked
->dirs
[i
]->recurse
)
2664 write_one_dir(untracked
->dirs
[i
], wd
);
2667 void write_untracked_extension(struct strbuf
*out
, struct untracked_cache
*untracked
)
2669 struct ondisk_untracked_cache
*ouc
;
2670 struct write_data wd
;
2671 unsigned char varbuf
[16];
2673 const unsigned hashsz
= the_hash_algo
->rawsz
;
2675 ouc
= xcalloc(1, sizeof(*ouc
));
2676 stat_data_to_disk(&ouc
->info_exclude_stat
, &untracked
->ss_info_exclude
.stat
);
2677 stat_data_to_disk(&ouc
->excludes_file_stat
, &untracked
->ss_excludes_file
.stat
);
2678 ouc
->dir_flags
= htonl(untracked
->dir_flags
);
2680 varint_len
= encode_varint(untracked
->ident
.len
, varbuf
);
2681 strbuf_add(out
, varbuf
, varint_len
);
2682 strbuf_addbuf(out
, &untracked
->ident
);
2684 strbuf_add(out
, ouc
, sizeof(*ouc
));
2685 strbuf_add(out
, untracked
->ss_info_exclude
.oid
.hash
, hashsz
);
2686 strbuf_add(out
, untracked
->ss_excludes_file
.oid
.hash
, hashsz
);
2687 strbuf_add(out
, untracked
->exclude_per_dir
, strlen(untracked
->exclude_per_dir
) + 1);
2690 if (!untracked
->root
) {
2691 varint_len
= encode_varint(0, varbuf
);
2692 strbuf_add(out
, varbuf
, varint_len
);
2697 wd
.check_only
= ewah_new();
2698 wd
.valid
= ewah_new();
2699 wd
.sha1_valid
= ewah_new();
2700 strbuf_init(&wd
.out
, 1024);
2701 strbuf_init(&wd
.sb_stat
, 1024);
2702 strbuf_init(&wd
.sb_sha1
, 1024);
2703 write_one_dir(untracked
->root
, &wd
);
2705 varint_len
= encode_varint(wd
.index
, varbuf
);
2706 strbuf_add(out
, varbuf
, varint_len
);
2707 strbuf_addbuf(out
, &wd
.out
);
2708 ewah_serialize_strbuf(wd
.valid
, out
);
2709 ewah_serialize_strbuf(wd
.check_only
, out
);
2710 ewah_serialize_strbuf(wd
.sha1_valid
, out
);
2711 strbuf_addbuf(out
, &wd
.sb_stat
);
2712 strbuf_addbuf(out
, &wd
.sb_sha1
);
2713 strbuf_addch(out
, '\0'); /* safe guard for string lists */
2715 ewah_free(wd
.valid
);
2716 ewah_free(wd
.check_only
);
2717 ewah_free(wd
.sha1_valid
);
2718 strbuf_release(&wd
.out
);
2719 strbuf_release(&wd
.sb_stat
);
2720 strbuf_release(&wd
.sb_sha1
);
2723 static void free_untracked(struct untracked_cache_dir
*ucd
)
2728 for (i
= 0; i
< ucd
->dirs_nr
; i
++)
2729 free_untracked(ucd
->dirs
[i
]);
2730 for (i
= 0; i
< ucd
->untracked_nr
; i
++)
2731 free(ucd
->untracked
[i
]);
2732 free(ucd
->untracked
);
2737 void free_untracked_cache(struct untracked_cache
*uc
)
2740 free_untracked(uc
->root
);
2746 struct untracked_cache_dir
**ucd
;
2747 struct ewah_bitmap
*check_only
;
2748 struct ewah_bitmap
*valid
;
2749 struct ewah_bitmap
*sha1_valid
;
2750 const unsigned char *data
;
2751 const unsigned char *end
;
2754 static void stat_data_from_disk(struct stat_data
*to
, const unsigned char *data
)
2756 memcpy(to
, data
, sizeof(*to
));
2757 to
->sd_ctime
.sec
= ntohl(to
->sd_ctime
.sec
);
2758 to
->sd_ctime
.nsec
= ntohl(to
->sd_ctime
.nsec
);
2759 to
->sd_mtime
.sec
= ntohl(to
->sd_mtime
.sec
);
2760 to
->sd_mtime
.nsec
= ntohl(to
->sd_mtime
.nsec
);
2761 to
->sd_dev
= ntohl(to
->sd_dev
);
2762 to
->sd_ino
= ntohl(to
->sd_ino
);
2763 to
->sd_uid
= ntohl(to
->sd_uid
);
2764 to
->sd_gid
= ntohl(to
->sd_gid
);
2765 to
->sd_size
= ntohl(to
->sd_size
);
2768 static int read_one_dir(struct untracked_cache_dir
**untracked_
,
2769 struct read_data
*rd
)
2771 struct untracked_cache_dir ud
, *untracked
;
2772 const unsigned char *data
= rd
->data
, *end
= rd
->end
;
2773 const unsigned char *eos
;
2777 memset(&ud
, 0, sizeof(ud
));
2779 value
= decode_varint(&data
);
2783 ud
.untracked_alloc
= value
;
2784 ud
.untracked_nr
= value
;
2785 if (ud
.untracked_nr
)
2786 ALLOC_ARRAY(ud
.untracked
, ud
.untracked_nr
);
2788 ud
.dirs_alloc
= ud
.dirs_nr
= decode_varint(&data
);
2791 ALLOC_ARRAY(ud
.dirs
, ud
.dirs_nr
);
2793 eos
= memchr(data
, '\0', end
- data
);
2794 if (!eos
|| eos
== end
)
2797 *untracked_
= untracked
= xmalloc(st_add3(sizeof(*untracked
), eos
- data
, 1));
2798 memcpy(untracked
, &ud
, sizeof(ud
));
2799 memcpy(untracked
->name
, data
, eos
- data
+ 1);
2802 for (i
= 0; i
< untracked
->untracked_nr
; i
++) {
2803 eos
= memchr(data
, '\0', end
- data
);
2804 if (!eos
|| eos
== end
)
2806 untracked
->untracked
[i
] = xmemdupz(data
, eos
- data
);
2810 rd
->ucd
[rd
->index
++] = untracked
;
2813 for (i
= 0; i
< untracked
->dirs_nr
; i
++) {
2814 if (read_one_dir(untracked
->dirs
+ i
, rd
) < 0)
2820 static void set_check_only(size_t pos
, void *cb
)
2822 struct read_data
*rd
= cb
;
2823 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
2827 static void read_stat(size_t pos
, void *cb
)
2829 struct read_data
*rd
= cb
;
2830 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
2831 if (rd
->data
+ sizeof(struct stat_data
) > rd
->end
) {
2832 rd
->data
= rd
->end
+ 1;
2835 stat_data_from_disk(&ud
->stat_data
, rd
->data
);
2836 rd
->data
+= sizeof(struct stat_data
);
2840 static void read_oid(size_t pos
, void *cb
)
2842 struct read_data
*rd
= cb
;
2843 struct untracked_cache_dir
*ud
= rd
->ucd
[pos
];
2844 if (rd
->data
+ the_hash_algo
->rawsz
> rd
->end
) {
2845 rd
->data
= rd
->end
+ 1;
2848 hashcpy(ud
->exclude_oid
.hash
, rd
->data
);
2849 rd
->data
+= the_hash_algo
->rawsz
;
2852 static void load_oid_stat(struct oid_stat
*oid_stat
, const unsigned char *data
,
2853 const unsigned char *sha1
)
2855 stat_data_from_disk(&oid_stat
->stat
, data
);
2856 hashcpy(oid_stat
->oid
.hash
, sha1
);
2857 oid_stat
->valid
= 1;
2860 struct untracked_cache
*read_untracked_extension(const void *data
, unsigned long sz
)
2862 struct untracked_cache
*uc
;
2863 struct read_data rd
;
2864 const unsigned char *next
= data
, *end
= (const unsigned char *)data
+ sz
;
2868 const char *exclude_per_dir
;
2869 const unsigned hashsz
= the_hash_algo
->rawsz
;
2870 const unsigned offset
= sizeof(struct ondisk_untracked_cache
);
2871 const unsigned exclude_per_dir_offset
= offset
+ 2 * hashsz
;
2873 if (sz
<= 1 || end
[-1] != '\0')
2877 ident_len
= decode_varint(&next
);
2878 if (next
+ ident_len
> end
)
2880 ident
= (const char *)next
;
2883 if (next
+ exclude_per_dir_offset
+ 1 > end
)
2886 uc
= xcalloc(1, sizeof(*uc
));
2887 strbuf_init(&uc
->ident
, ident_len
);
2888 strbuf_add(&uc
->ident
, ident
, ident_len
);
2889 load_oid_stat(&uc
->ss_info_exclude
,
2890 next
+ ouc_offset(info_exclude_stat
),
2892 load_oid_stat(&uc
->ss_excludes_file
,
2893 next
+ ouc_offset(excludes_file_stat
),
2894 next
+ offset
+ hashsz
);
2895 uc
->dir_flags
= get_be32(next
+ ouc_offset(dir_flags
));
2896 exclude_per_dir
= (const char *)next
+ exclude_per_dir_offset
;
2897 uc
->exclude_per_dir
= xstrdup(exclude_per_dir
);
2898 /* NUL after exclude_per_dir is covered by sizeof(*ouc) */
2899 next
+= exclude_per_dir_offset
+ strlen(exclude_per_dir
) + 1;
2903 len
= decode_varint(&next
);
2904 if (next
> end
|| len
== 0)
2907 rd
.valid
= ewah_new();
2908 rd
.check_only
= ewah_new();
2909 rd
.sha1_valid
= ewah_new();
2913 ALLOC_ARRAY(rd
.ucd
, len
);
2915 if (read_one_dir(&uc
->root
, &rd
) || rd
.index
!= len
)
2919 len
= ewah_read_mmap(rd
.valid
, next
, end
- next
);
2924 len
= ewah_read_mmap(rd
.check_only
, next
, end
- next
);
2929 len
= ewah_read_mmap(rd
.sha1_valid
, next
, end
- next
);
2933 ewah_each_bit(rd
.check_only
, set_check_only
, &rd
);
2934 rd
.data
= next
+ len
;
2935 ewah_each_bit(rd
.valid
, read_stat
, &rd
);
2936 ewah_each_bit(rd
.sha1_valid
, read_oid
, &rd
);
2941 ewah_free(rd
.valid
);
2942 ewah_free(rd
.check_only
);
2943 ewah_free(rd
.sha1_valid
);
2946 free_untracked_cache(uc
);
2952 static void invalidate_one_directory(struct untracked_cache
*uc
,
2953 struct untracked_cache_dir
*ucd
)
2955 uc
->dir_invalidated
++;
2957 ucd
->untracked_nr
= 0;
2961 * Normally when an entry is added or removed from a directory,
2962 * invalidating that directory is enough. No need to touch its
2963 * ancestors. When a directory is shown as "foo/bar/" in git-status
2964 * however, deleting or adding an entry may have cascading effect.
2966 * Say the "foo/bar/file" has become untracked, we need to tell the
2967 * untracked_cache_dir of "foo" that "bar/" is not an untracked
2968 * directory any more (because "bar" is managed by foo as an untracked
2971 * Similarly, if "foo/bar/file" moves from untracked to tracked and it
2972 * was the last untracked entry in the entire "foo", we should show
2973 * "foo/" instead. Which means we have to invalidate past "bar" up to
2976 * This function traverses all directories from root to leaf. If there
2977 * is a chance of one of the above cases happening, we invalidate back
2978 * to root. Otherwise we just invalidate the leaf. There may be a more
2979 * sophisticated way than checking for SHOW_OTHER_DIRECTORIES to
2980 * detect these cases and avoid unnecessary invalidation, for example,
2981 * checking for the untracked entry named "bar/" in "foo", but for now
2982 * stick to something safe and simple.
2984 static int invalidate_one_component(struct untracked_cache
*uc
,
2985 struct untracked_cache_dir
*dir
,
2986 const char *path
, int len
)
2988 const char *rest
= strchr(path
, '/');
2991 int component_len
= rest
- path
;
2992 struct untracked_cache_dir
*d
=
2993 lookup_untracked(uc
, dir
, path
, component_len
);
2995 invalidate_one_component(uc
, d
, rest
+ 1,
2996 len
- (component_len
+ 1));
2998 invalidate_one_directory(uc
, dir
);
3002 invalidate_one_directory(uc
, dir
);
3003 return uc
->dir_flags
& DIR_SHOW_OTHER_DIRECTORIES
;
3006 void untracked_cache_invalidate_path(struct index_state
*istate
,
3007 const char *path
, int safe_path
)
3009 if (!istate
->untracked
|| !istate
->untracked
->root
)
3011 if (!safe_path
&& !verify_path(path
, 0))
3013 invalidate_one_component(istate
->untracked
, istate
->untracked
->root
,
3014 path
, strlen(path
));
3017 void untracked_cache_remove_from_index(struct index_state
*istate
,
3020 untracked_cache_invalidate_path(istate
, path
, 1);
3023 void untracked_cache_add_to_index(struct index_state
*istate
,
3026 untracked_cache_invalidate_path(istate
, path
, 1);
3029 static void connect_wt_gitdir_in_nested(const char *sub_worktree
,
3030 const char *sub_gitdir
)
3033 struct repository subrepo
;
3034 struct strbuf sub_wt
= STRBUF_INIT
;
3035 struct strbuf sub_gd
= STRBUF_INIT
;
3037 const struct submodule
*sub
;
3039 /* If the submodule has no working tree, we can ignore it. */
3040 if (repo_init(&subrepo
, sub_gitdir
, sub_worktree
))
3043 if (repo_read_index(&subrepo
) < 0)
3044 die(_("index file corrupt in repo %s"), subrepo
.gitdir
);
3046 for (i
= 0; i
< subrepo
.index
->cache_nr
; i
++) {
3047 const struct cache_entry
*ce
= subrepo
.index
->cache
[i
];
3049 if (!S_ISGITLINK(ce
->ce_mode
))
3052 while (i
+ 1 < subrepo
.index
->cache_nr
&&
3053 !strcmp(ce
->name
, subrepo
.index
->cache
[i
+ 1]->name
))
3055 * Skip entries with the same name in different stages
3056 * to make sure an entry is returned only once.
3060 sub
= submodule_from_path(&subrepo
, &null_oid
, ce
->name
);
3061 if (!sub
|| !is_submodule_active(&subrepo
, ce
->name
))
3062 /* .gitmodules broken or inactive sub */
3065 strbuf_reset(&sub_wt
);
3066 strbuf_reset(&sub_gd
);
3067 strbuf_addf(&sub_wt
, "%s/%s", sub_worktree
, sub
->path
);
3068 strbuf_addf(&sub_gd
, "%s/modules/%s", sub_gitdir
, sub
->name
);
3070 connect_work_tree_and_git_dir(sub_wt
.buf
, sub_gd
.buf
, 1);
3072 strbuf_release(&sub_wt
);
3073 strbuf_release(&sub_gd
);
3074 repo_clear(&subrepo
);
3077 void connect_work_tree_and_git_dir(const char *work_tree_
,
3078 const char *git_dir_
,
3079 int recurse_into_nested
)
3081 struct strbuf gitfile_sb
= STRBUF_INIT
;
3082 struct strbuf cfg_sb
= STRBUF_INIT
;
3083 struct strbuf rel_path
= STRBUF_INIT
;
3084 char *git_dir
, *work_tree
;
3086 /* Prepare .git file */
3087 strbuf_addf(&gitfile_sb
, "%s/.git", work_tree_
);
3088 if (safe_create_leading_directories_const(gitfile_sb
.buf
))
3089 die(_("could not create directories for %s"), gitfile_sb
.buf
);
3091 /* Prepare config file */
3092 strbuf_addf(&cfg_sb
, "%s/config", git_dir_
);
3093 if (safe_create_leading_directories_const(cfg_sb
.buf
))
3094 die(_("could not create directories for %s"), cfg_sb
.buf
);
3096 git_dir
= real_pathdup(git_dir_
, 1);
3097 work_tree
= real_pathdup(work_tree_
, 1);
3099 /* Write .git file */
3100 write_file(gitfile_sb
.buf
, "gitdir: %s",
3101 relative_path(git_dir
, work_tree
, &rel_path
));
3102 /* Update core.worktree setting */
3103 git_config_set_in_file(cfg_sb
.buf
, "core.worktree",
3104 relative_path(work_tree
, git_dir
, &rel_path
));
3106 strbuf_release(&gitfile_sb
);
3107 strbuf_release(&cfg_sb
);
3108 strbuf_release(&rel_path
);
3110 if (recurse_into_nested
)
3111 connect_wt_gitdir_in_nested(work_tree
, git_dir
);
3118 * Migrate the git directory of the given path from old_git_dir to new_git_dir.
3120 void relocate_gitdir(const char *path
, const char *old_git_dir
, const char *new_git_dir
)
3122 if (rename(old_git_dir
, new_git_dir
) < 0)
3123 die_errno(_("could not migrate git directory from '%s' to '%s'"),
3124 old_git_dir
, new_git_dir
);
3126 connect_work_tree_and_git_dir(path
, new_git_dir
, 0);