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 "wildmatch.h"
17 struct path_simplify
{
23 * Tells read_directory_recursive how a file or directory should be treated.
24 * Values are ordered by significance, e.g. if a directory contains both
25 * excluded and untracked files, it is listed as untracked because
26 * path_untracked > path_excluded.
35 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
36 const char *path
, int len
,
37 int check_only
, const struct path_simplify
*simplify
);
38 static int get_dtype(struct dirent
*de
, const char *path
, int len
);
40 /* helper string functions with support for the ignore_case flag */
41 int strcmp_icase(const char *a
, const char *b
)
43 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
46 int strncmp_icase(const char *a
, const char *b
, size_t count
)
48 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
51 int fnmatch_icase(const char *pattern
, const char *string
, int flags
)
53 return wildmatch(pattern
, string
,
54 flags
| (ignore_case
? WM_CASEFOLD
: 0),
58 int git_fnmatch(const struct pathspec_item
*item
,
59 const char *pattern
, const char *string
,
63 if (ps_strncmp(item
, pattern
, string
, prefix
))
68 if (item
->flags
& PATHSPEC_ONESTAR
) {
69 int pattern_len
= strlen(++pattern
);
70 int string_len
= strlen(string
);
71 return string_len
< pattern_len
||
72 ps_strcmp(item
, pattern
,
73 string
+ string_len
- pattern_len
);
75 if (item
->magic
& PATHSPEC_GLOB
)
76 return wildmatch(pattern
, string
,
78 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0),
81 /* wildmatch has not learned no FNM_PATHNAME mode yet */
82 return wildmatch(pattern
, string
,
83 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0,
87 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
88 const char *string
, int stringlen
,
92 struct strbuf pat_buf
= STRBUF_INIT
;
93 struct strbuf str_buf
= STRBUF_INIT
;
94 const char *use_pat
= pattern
;
95 const char *use_str
= string
;
97 if (pattern
[patternlen
]) {
98 strbuf_add(&pat_buf
, pattern
, patternlen
);
99 use_pat
= pat_buf
.buf
;
101 if (string
[stringlen
]) {
102 strbuf_add(&str_buf
, string
, stringlen
);
103 use_str
= str_buf
.buf
;
107 flags
|= WM_CASEFOLD
;
108 match_status
= wildmatch(use_pat
, use_str
, flags
, NULL
);
110 strbuf_release(&pat_buf
);
111 strbuf_release(&str_buf
);
116 static size_t common_prefix_len(const struct pathspec
*pathspec
)
122 * ":(icase)path" is treated as a pathspec full of
123 * wildcard. In other words, only prefix is considered common
124 * prefix. If the pathspec is abc/foo abc/bar, running in
125 * subdir xyz, the common prefix is still xyz, not xuz/abc as
128 GUARD_PATHSPEC(pathspec
,
136 for (n
= 0; n
< pathspec
->nr
; n
++) {
137 size_t i
= 0, len
= 0, item_len
;
138 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
140 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
141 item_len
= pathspec
->items
[n
].prefix
;
143 item_len
= pathspec
->items
[n
].nowildcard_len
;
144 while (i
< item_len
&& (n
== 0 || i
< max
)) {
145 char c
= pathspec
->items
[n
].match
[i
];
146 if (c
!= pathspec
->items
[0].match
[i
])
152 if (n
== 0 || len
< max
) {
162 * Returns a copy of the longest leading path common among all
165 char *common_prefix(const struct pathspec
*pathspec
)
167 unsigned long len
= common_prefix_len(pathspec
);
169 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
172 int fill_directory(struct dir_struct
*dir
, const struct pathspec
*pathspec
)
177 * Calculate common prefix for the pathspec, and
178 * use that to optimize the directory walk
180 len
= common_prefix_len(pathspec
);
182 /* Read the directory and prune it */
183 read_directory(dir
, pathspec
->nr
? pathspec
->_raw
[0] : "", len
, pathspec
);
187 int within_depth(const char *name
, int namelen
,
188 int depth
, int max_depth
)
190 const char *cp
= name
, *cpe
= name
+ namelen
;
196 if (depth
> max_depth
)
202 #define DO_MATCH_EXCLUDE 1
203 #define DO_MATCH_DIRECTORY 2
206 * Does 'match' match the given name?
207 * A match is found if
209 * (1) the 'match' string is leading directory of 'name', or
210 * (2) the 'match' string is a wildcard and matches 'name', or
211 * (3) the 'match' string is exactly the same as 'name'.
213 * and the return value tells which case it was.
215 * It returns 0 when there is no match.
217 static int match_pathspec_item(const struct pathspec_item
*item
, int prefix
,
218 const char *name
, int namelen
, unsigned flags
)
220 /* name/namelen has prefix cut off by caller */
221 const char *match
= item
->match
+ prefix
;
222 int matchlen
= item
->len
- prefix
;
225 * The normal call pattern is:
226 * 1. prefix = common_prefix_len(ps);
227 * 2. prune something, or fill_directory
228 * 3. match_pathspec()
230 * 'prefix' at #1 may be shorter than the command's prefix and
231 * it's ok for #2 to match extra files. Those extras will be
234 * Suppose the pathspec is 'foo' and '../bar' running from
235 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
236 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
237 * user does not want XYZ/foo, only the "foo" part should be
238 * case-insensitive. We need to filter out XYZ/foo here. In
239 * other words, we do not trust the caller on comparing the
240 * prefix part when :(icase) is involved. We do exact
241 * comparison ourselves.
243 * Normally the caller (common_prefix_len() in fact) does
244 * _exact_ matching on name[-prefix+1..-1] and we do not need
245 * to check that part. Be defensive and check it anyway, in
246 * case common_prefix_len is changed, or a new caller is
247 * introduced that does not use common_prefix_len.
249 * If the penalty turns out too high when prefix is really
250 * long, maybe change it to
251 * strncmp(match, name, item->prefix - prefix)
253 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
254 strncmp(item
->match
, name
- prefix
, item
->prefix
))
257 /* If the match was just the prefix, we matched */
259 return MATCHED_RECURSIVELY
;
261 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
262 if (matchlen
== namelen
)
263 return MATCHED_EXACTLY
;
265 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
266 return MATCHED_RECURSIVELY
;
267 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
268 match
[matchlen
- 1] == '/' &&
269 namelen
== matchlen
- 1 &&
270 !ps_strncmp(item
, match
, name
, namelen
))
271 return MATCHED_EXACTLY
;
273 if (item
->nowildcard_len
< item
->len
&&
274 !git_fnmatch(item
, match
, name
,
275 item
->nowildcard_len
- prefix
))
276 return MATCHED_FNMATCH
;
282 * Given a name and a list of pathspecs, returns the nature of the
283 * closest (i.e. most specific) match of the name to any of the
286 * The caller typically calls this multiple times with the same
287 * pathspec and seen[] array but with different name/namelen
288 * (e.g. entries from the index) and is interested in seeing if and
289 * how each pathspec matches all the names it calls this function
290 * with. A mark is left in the seen[] array for each pathspec element
291 * indicating the closest type of match that element achieved, so if
292 * seen[n] remains zero after multiple invocations, that means the nth
293 * pathspec did not match any names, which could indicate that the
294 * user mistyped the nth pathspec.
296 static int do_match_pathspec(const struct pathspec
*ps
,
297 const char *name
, int namelen
,
298 int prefix
, char *seen
,
301 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
312 if (!ps
->recursive
||
313 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
315 return MATCHED_RECURSIVELY
;
317 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
318 return MATCHED_EXACTLY
;
326 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
329 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
330 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
333 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
336 * Make exclude patterns optional and never report
337 * "pathspec ':(exclude)foo' matches no files"
339 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
340 seen
[i
] = MATCHED_FNMATCH
;
341 how
= match_pathspec_item(ps
->items
+i
, prefix
, name
,
344 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
345 ps
->max_depth
!= -1 &&
346 how
&& how
!= MATCHED_FNMATCH
) {
347 int len
= ps
->items
[i
].len
;
348 if (name
[len
] == '/')
350 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
351 how
= MATCHED_EXACTLY
;
358 if (seen
&& seen
[i
] < how
)
365 int match_pathspec(const struct pathspec
*ps
,
366 const char *name
, int namelen
,
367 int prefix
, char *seen
, int is_dir
)
369 int positive
, negative
;
370 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
371 positive
= do_match_pathspec(ps
, name
, namelen
,
372 prefix
, seen
, flags
);
373 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
375 negative
= do_match_pathspec(ps
, name
, namelen
,
377 flags
| DO_MATCH_EXCLUDE
);
378 return negative
? 0 : positive
;
381 int report_path_error(const char *ps_matched
,
382 const struct pathspec
*pathspec
,
386 * Make sure all pathspec matched; otherwise it is an error.
389 for (num
= 0; num
< pathspec
->nr
; num
++) {
390 int other
, found_dup
;
395 * The caller might have fed identical pathspec
396 * twice. Do not barf on such a mistake.
397 * FIXME: parse_pathspec should have eliminated
398 * duplicate pathspec.
400 for (found_dup
= other
= 0;
401 !found_dup
&& other
< pathspec
->nr
;
403 if (other
== num
|| !ps_matched
[other
])
405 if (!strcmp(pathspec
->items
[other
].original
,
406 pathspec
->items
[num
].original
))
408 * Ok, we have a match already.
415 error("pathspec '%s' did not match any file(s) known to git.",
416 pathspec
->items
[num
].original
);
423 * Return the length of the "simple" part of a path match limiter.
425 int simple_length(const char *match
)
430 unsigned char c
= *match
++;
432 if (c
== '\0' || is_glob_special(c
))
437 int no_wildcard(const char *string
)
439 return string
[simple_length(string
)] == '\0';
442 void parse_exclude_pattern(const char **pattern
,
447 const char *p
= *pattern
;
452 *flags
|= EXC_FLAG_NEGATIVE
;
456 if (len
&& p
[len
- 1] == '/') {
458 *flags
|= EXC_FLAG_MUSTBEDIR
;
460 for (i
= 0; i
< len
; i
++) {
465 *flags
|= EXC_FLAG_NODIR
;
466 *nowildcardlen
= simple_length(p
);
468 * we should have excluded the trailing slash from 'p' too,
469 * but that's one more allocation. Instead just make sure
470 * nowildcardlen does not exceed real patternlen
472 if (*nowildcardlen
> len
)
473 *nowildcardlen
= len
;
474 if (*p
== '*' && no_wildcard(p
+ 1))
475 *flags
|= EXC_FLAG_ENDSWITH
;
480 void add_exclude(const char *string
, const char *base
,
481 int baselen
, struct exclude_list
*el
, int srcpos
)
488 parse_exclude_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
489 if (flags
& EXC_FLAG_MUSTBEDIR
) {
491 x
= xmalloc(sizeof(*x
) + patternlen
+ 1);
493 memcpy(s
, string
, patternlen
);
494 s
[patternlen
] = '\0';
497 x
= xmalloc(sizeof(*x
));
500 x
->patternlen
= patternlen
;
501 x
->nowildcardlen
= nowildcardlen
;
503 x
->baselen
= baselen
;
506 ALLOC_GROW(el
->excludes
, el
->nr
+ 1, el
->alloc
);
507 el
->excludes
[el
->nr
++] = x
;
511 static void *read_skip_worktree_file_from_index(const char *path
, size_t *size
)
515 enum object_type type
;
519 pos
= cache_name_pos(path
, len
);
522 if (!ce_skip_worktree(active_cache
[pos
]))
524 data
= read_sha1_file(active_cache
[pos
]->sha1
, &type
, &sz
);
525 if (!data
|| type
!= OBJ_BLOB
) {
534 * Frees memory within el which was allocated for exclude patterns and
535 * the file buffer. Does not free el itself.
537 void clear_exclude_list(struct exclude_list
*el
)
541 for (i
= 0; i
< el
->nr
; i
++)
542 free(el
->excludes
[i
]);
551 static void trim_trailing_spaces(char *buf
)
553 char *p
, *last_space
= NULL
;
555 for (p
= buf
; *p
; p
++)
574 int add_excludes_from_file_to_list(const char *fname
,
577 struct exclude_list
*el
,
581 int fd
, i
, lineno
= 1;
585 fd
= open(fname
, O_RDONLY
);
586 if (fd
< 0 || fstat(fd
, &st
) < 0) {
588 warn_on_inaccessible(fname
);
592 (buf
= read_skip_worktree_file_from_index(fname
, &size
)) == NULL
)
598 if (buf
[size
-1] != '\n') {
599 buf
= xrealloc(buf
, size
+1);
603 size
= xsize_t(st
.st_size
);
608 buf
= xmalloc(size
+1);
609 if (read_in_full(fd
, buf
, size
) != size
) {
620 if (skip_utf8_bom(&buf
, size
))
621 size
-= buf
- el
->filebuf
;
625 for (i
= 0; i
< size
; i
++) {
626 if (buf
[i
] == '\n') {
627 if (entry
!= buf
+ i
&& entry
[0] != '#') {
628 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
629 trim_trailing_spaces(entry
);
630 add_exclude(entry
, base
, baselen
, el
, lineno
);
639 struct exclude_list
*add_exclude_list(struct dir_struct
*dir
,
640 int group_type
, const char *src
)
642 struct exclude_list
*el
;
643 struct exclude_list_group
*group
;
645 group
= &dir
->exclude_list_group
[group_type
];
646 ALLOC_GROW(group
->el
, group
->nr
+ 1, group
->alloc
);
647 el
= &group
->el
[group
->nr
++];
648 memset(el
, 0, sizeof(*el
));
654 * Used to set up core.excludesfile and .git/info/exclude lists.
656 void add_excludes_from_file(struct dir_struct
*dir
, const char *fname
)
658 struct exclude_list
*el
;
659 el
= add_exclude_list(dir
, EXC_FILE
, fname
);
660 if (add_excludes_from_file_to_list(fname
, "", 0, el
, 0) < 0)
661 die("cannot use %s as an exclude file", fname
);
664 int match_basename(const char *basename
, int basenamelen
,
665 const char *pattern
, int prefix
, int patternlen
,
668 if (prefix
== patternlen
) {
669 if (patternlen
== basenamelen
&&
670 !strncmp_icase(pattern
, basename
, basenamelen
))
672 } else if (flags
& EXC_FLAG_ENDSWITH
) {
673 /* "*literal" matching against "fooliteral" */
674 if (patternlen
- 1 <= basenamelen
&&
675 !strncmp_icase(pattern
+ 1,
676 basename
+ basenamelen
- (patternlen
- 1),
680 if (fnmatch_icase_mem(pattern
, patternlen
,
681 basename
, basenamelen
,
688 int match_pathname(const char *pathname
, int pathlen
,
689 const char *base
, int baselen
,
690 const char *pattern
, int prefix
, int patternlen
,
697 * match with FNM_PATHNAME; the pattern has base implicitly
700 if (*pattern
== '/') {
707 * baselen does not count the trailing slash. base[] may or
708 * may not end with a trailing slash though.
710 if (pathlen
< baselen
+ 1 ||
711 (baselen
&& pathname
[baselen
] != '/') ||
712 strncmp_icase(pathname
, base
, baselen
))
715 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
716 name
= pathname
+ pathlen
- namelen
;
720 * if the non-wildcard part is longer than the
721 * remaining pathname, surely it cannot match.
723 if (prefix
> namelen
)
726 if (strncmp_icase(pattern
, name
, prefix
))
729 patternlen
-= prefix
;
734 * If the whole pattern did not have a wildcard,
735 * then our prefix match is all we need; we
736 * do not need to call fnmatch at all.
738 if (!patternlen
&& !namelen
)
742 return fnmatch_icase_mem(pattern
, patternlen
,
748 * Scan the given exclude list in reverse to see whether pathname
749 * should be ignored. The first match (i.e. the last on the list), if
750 * any, determines the fate. Returns the exclude_list element which
751 * matched, or NULL for undecided.
753 static struct exclude
*last_exclude_matching_from_list(const char *pathname
,
755 const char *basename
,
757 struct exclude_list
*el
)
762 return NULL
; /* undefined */
764 for (i
= el
->nr
- 1; 0 <= i
; i
--) {
765 struct exclude
*x
= el
->excludes
[i
];
766 const char *exclude
= x
->pattern
;
767 int prefix
= x
->nowildcardlen
;
769 if (x
->flags
& EXC_FLAG_MUSTBEDIR
) {
770 if (*dtype
== DT_UNKNOWN
)
771 *dtype
= get_dtype(NULL
, pathname
, pathlen
);
772 if (*dtype
!= DT_DIR
)
776 if (x
->flags
& EXC_FLAG_NODIR
) {
777 if (match_basename(basename
,
778 pathlen
- (basename
- pathname
),
779 exclude
, prefix
, x
->patternlen
,
785 assert(x
->baselen
== 0 || x
->base
[x
->baselen
- 1] == '/');
786 if (match_pathname(pathname
, pathlen
,
787 x
->base
, x
->baselen
? x
->baselen
- 1 : 0,
788 exclude
, prefix
, x
->patternlen
, x
->flags
))
791 return NULL
; /* undecided */
795 * Scan the list and let the last match determine the fate.
796 * Return 1 for exclude, 0 for include and -1 for undecided.
798 int is_excluded_from_list(const char *pathname
,
799 int pathlen
, const char *basename
, int *dtype
,
800 struct exclude_list
*el
)
802 struct exclude
*exclude
;
803 exclude
= last_exclude_matching_from_list(pathname
, pathlen
, basename
, dtype
, el
);
805 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
806 return -1; /* undecided */
809 static struct exclude
*last_exclude_matching_from_lists(struct dir_struct
*dir
,
810 const char *pathname
, int pathlen
, const char *basename
,
814 struct exclude_list_group
*group
;
815 struct exclude
*exclude
;
816 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
817 group
= &dir
->exclude_list_group
[i
];
818 for (j
= group
->nr
- 1; j
>= 0; j
--) {
819 exclude
= last_exclude_matching_from_list(
820 pathname
, pathlen
, basename
, dtype_p
,
830 * Loads the per-directory exclude list for the substring of base
831 * which has a char length of baselen.
833 static void prep_exclude(struct dir_struct
*dir
, const char *base
, int baselen
)
835 struct exclude_list_group
*group
;
836 struct exclude_list
*el
;
837 struct exclude_stack
*stk
= NULL
;
840 group
= &dir
->exclude_list_group
[EXC_DIRS
];
843 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
844 * which originate from directories not in the prefix of the
845 * path being checked.
847 while ((stk
= dir
->exclude_stack
) != NULL
) {
848 if (stk
->baselen
<= baselen
&&
849 !strncmp(dir
->basebuf
.buf
, base
, stk
->baselen
))
851 el
= &group
->el
[dir
->exclude_stack
->exclude_ix
];
852 dir
->exclude_stack
= stk
->prev
;
854 free((char *)el
->src
); /* see strbuf_detach() below */
855 clear_exclude_list(el
);
860 /* Skip traversing into sub directories if the parent is excluded */
865 * Lazy initialization. All call sites currently just
866 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
867 * them seems lots of work for little benefit.
869 if (!dir
->basebuf
.buf
)
870 strbuf_init(&dir
->basebuf
, PATH_MAX
);
872 /* Read from the parent directories and push them down. */
873 current
= stk
? stk
->baselen
: -1;
874 strbuf_setlen(&dir
->basebuf
, current
< 0 ? 0 : current
);
875 while (current
< baselen
) {
878 stk
= xcalloc(1, sizeof(*stk
));
883 cp
= strchr(base
+ current
+ 1, '/');
885 die("oops in prep_exclude");
888 stk
->prev
= dir
->exclude_stack
;
889 stk
->baselen
= cp
- base
;
890 stk
->exclude_ix
= group
->nr
;
891 el
= add_exclude_list(dir
, EXC_DIRS
, NULL
);
892 strbuf_add(&dir
->basebuf
, base
+ current
, stk
->baselen
- current
);
893 assert(stk
->baselen
== dir
->basebuf
.len
);
895 /* Abort if the directory is excluded */
898 dir
->basebuf
.buf
[stk
->baselen
- 1] = 0;
899 dir
->exclude
= last_exclude_matching_from_lists(dir
,
900 dir
->basebuf
.buf
, stk
->baselen
- 1,
901 dir
->basebuf
.buf
+ current
, &dt
);
902 dir
->basebuf
.buf
[stk
->baselen
- 1] = '/';
904 dir
->exclude
->flags
& EXC_FLAG_NEGATIVE
)
907 dir
->exclude_stack
= stk
;
912 /* Try to read per-directory file */
913 if (dir
->exclude_per_dir
) {
915 * dir->basebuf gets reused by the traversal, but we
916 * need fname to remain unchanged to ensure the src
917 * member of each struct exclude correctly
918 * back-references its source file. Other invocations
919 * of add_exclude_list provide stable strings, so we
920 * strbuf_detach() and free() here in the caller.
922 struct strbuf sb
= STRBUF_INIT
;
923 strbuf_addbuf(&sb
, &dir
->basebuf
);
924 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
925 el
->src
= strbuf_detach(&sb
, NULL
);
926 add_excludes_from_file_to_list(el
->src
, el
->src
,
927 stk
->baselen
, el
, 1);
929 dir
->exclude_stack
= stk
;
930 current
= stk
->baselen
;
932 strbuf_setlen(&dir
->basebuf
, baselen
);
936 * Loads the exclude lists for the directory containing pathname, then
937 * scans all exclude lists to determine whether pathname is excluded.
938 * Returns the exclude_list element which matched, or NULL for
941 struct exclude
*last_exclude_matching(struct dir_struct
*dir
,
942 const char *pathname
,
945 int pathlen
= strlen(pathname
);
946 const char *basename
= strrchr(pathname
, '/');
947 basename
= (basename
) ? basename
+1 : pathname
;
949 prep_exclude(dir
, pathname
, basename
-pathname
);
954 return last_exclude_matching_from_lists(dir
, pathname
, pathlen
,
959 * Loads the exclude lists for the directory containing pathname, then
960 * scans all exclude lists to determine whether pathname is excluded.
961 * Returns 1 if true, otherwise 0.
963 int is_excluded(struct dir_struct
*dir
, const char *pathname
, int *dtype_p
)
965 struct exclude
*exclude
=
966 last_exclude_matching(dir
, pathname
, dtype_p
);
968 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
972 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
974 struct dir_entry
*ent
;
976 ent
= xmalloc(sizeof(*ent
) + len
+ 1);
978 memcpy(ent
->name
, pathname
, len
);
983 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
, const char *pathname
, int len
)
985 if (cache_file_exists(pathname
, len
, ignore_case
))
988 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->alloc
);
989 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
992 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
, const char *pathname
, int len
)
994 if (!cache_name_is_other(pathname
, len
))
997 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->ignored_alloc
);
998 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1002 index_nonexistent
= 0,
1008 * Do not use the alphabetically sorted index to look up
1009 * the directory name; instead, use the case insensitive
1012 static enum exist_status
directory_exists_in_index_icase(const char *dirname
, int len
)
1014 const struct cache_entry
*ce
= cache_dir_exists(dirname
, len
);
1015 unsigned char endchar
;
1018 return index_nonexistent
;
1019 endchar
= ce
->name
[len
];
1022 * The cache_entry structure returned will contain this dirname
1023 * and possibly additional path components.
1026 return index_directory
;
1029 * If there are no additional path components, then this cache_entry
1030 * represents a submodule. Submodules, despite being directories,
1031 * are stored in the cache without a closing slash.
1033 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1034 return index_gitdir
;
1036 /* This should never be hit, but it exists just in case. */
1037 return index_nonexistent
;
1041 * The index sorts alphabetically by entry name, which
1042 * means that a gitlink sorts as '\0' at the end, while
1043 * a directory (which is defined not as an entry, but as
1044 * the files it contains) will sort with the '/' at the
1047 static enum exist_status
directory_exists_in_index(const char *dirname
, int len
)
1052 return directory_exists_in_index_icase(dirname
, len
);
1054 pos
= cache_name_pos(dirname
, len
);
1057 while (pos
< active_nr
) {
1058 const struct cache_entry
*ce
= active_cache
[pos
++];
1059 unsigned char endchar
;
1061 if (strncmp(ce
->name
, dirname
, len
))
1063 endchar
= ce
->name
[len
];
1067 return index_directory
;
1068 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1069 return index_gitdir
;
1071 return index_nonexistent
;
1075 * When we find a directory when traversing the filesystem, we
1076 * have three distinct cases:
1079 * - see it as a directory
1082 * and which one we choose depends on a combination of existing
1083 * git index contents and the flags passed into the directory
1084 * traversal routine.
1086 * Case 1: If we *already* have entries in the index under that
1087 * directory name, we always recurse into the directory to see
1090 * Case 2: If we *already* have that directory name as a gitlink,
1091 * we always continue to see it as a gitlink, regardless of whether
1092 * there is an actual git directory there or not (it might not
1093 * be checked out as a subproject!)
1095 * Case 3: if we didn't have it in the index previously, we
1096 * have a few sub-cases:
1098 * (a) if "show_other_directories" is true, we show it as
1099 * just a directory, unless "hide_empty_directories" is
1100 * also true, in which case we need to check if it contains any
1101 * untracked and / or ignored files.
1102 * (b) if it looks like a git directory, and we don't have
1103 * 'no_gitlinks' set we treat it as a gitlink, and show it
1105 * (c) otherwise, we recurse into it.
1107 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1108 const char *dirname
, int len
, int exclude
,
1109 const struct path_simplify
*simplify
)
1111 /* The "len-1" is to strip the final '/' */
1112 switch (directory_exists_in_index(dirname
, len
-1)) {
1113 case index_directory
:
1114 return path_recurse
;
1119 case index_nonexistent
:
1120 if (dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)
1122 if (!(dir
->flags
& DIR_NO_GITLINKS
)) {
1123 unsigned char sha1
[20];
1124 if (resolve_gitlink_ref(dirname
, "HEAD", sha1
) == 0)
1125 return path_untracked
;
1127 return path_recurse
;
1130 /* This is the "show_other_directories" case */
1132 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1133 return exclude
? path_excluded
: path_untracked
;
1135 return read_directory_recursive(dir
, dirname
, len
, 1, simplify
);
1139 * This is an inexact early pruning of any recursive directory
1140 * reading - if the path cannot possibly be in the pathspec,
1141 * return true, and we'll skip it early.
1143 static int simplify_away(const char *path
, int pathlen
, const struct path_simplify
*simplify
)
1147 const char *match
= simplify
->path
;
1148 int len
= simplify
->len
;
1154 if (!memcmp(path
, match
, len
))
1164 * This function tells us whether an excluded path matches a
1165 * list of "interesting" pathspecs. That is, whether a path matched
1166 * by any of the pathspecs could possibly be ignored by excluding
1167 * the specified path. This can happen if:
1169 * 1. the path is mentioned explicitly in the pathspec
1171 * 2. the path is a directory prefix of some element in the
1174 static int exclude_matches_pathspec(const char *path
, int len
,
1175 const struct path_simplify
*simplify
)
1178 for (; simplify
->path
; simplify
++) {
1179 if (len
== simplify
->len
1180 && !memcmp(path
, simplify
->path
, len
))
1182 if (len
< simplify
->len
1183 && simplify
->path
[len
] == '/'
1184 && !memcmp(path
, simplify
->path
, len
))
1191 static int get_index_dtype(const char *path
, int len
)
1194 const struct cache_entry
*ce
;
1196 ce
= cache_file_exists(path
, len
, 0);
1198 if (!ce_uptodate(ce
))
1200 if (S_ISGITLINK(ce
->ce_mode
))
1203 * Nobody actually cares about the
1204 * difference between DT_LNK and DT_REG
1209 /* Try to look it up as a directory */
1210 pos
= cache_name_pos(path
, len
);
1214 while (pos
< active_nr
) {
1215 ce
= active_cache
[pos
++];
1216 if (strncmp(ce
->name
, path
, len
))
1218 if (ce
->name
[len
] > '/')
1220 if (ce
->name
[len
] < '/')
1222 if (!ce_uptodate(ce
))
1223 break; /* continue? */
1229 static int get_dtype(struct dirent
*de
, const char *path
, int len
)
1231 int dtype
= de
? DTYPE(de
) : DT_UNKNOWN
;
1234 if (dtype
!= DT_UNKNOWN
)
1236 dtype
= get_index_dtype(path
, len
);
1237 if (dtype
!= DT_UNKNOWN
)
1239 if (lstat(path
, &st
))
1241 if (S_ISREG(st
.st_mode
))
1243 if (S_ISDIR(st
.st_mode
))
1245 if (S_ISLNK(st
.st_mode
))
1250 static enum path_treatment
treat_one_path(struct dir_struct
*dir
,
1251 struct strbuf
*path
,
1252 const struct path_simplify
*simplify
,
1253 int dtype
, struct dirent
*de
)
1256 int has_path_in_index
= !!cache_file_exists(path
->buf
, path
->len
, ignore_case
);
1258 if (dtype
== DT_UNKNOWN
)
1259 dtype
= get_dtype(de
, path
->buf
, path
->len
);
1261 /* Always exclude indexed files */
1262 if (dtype
!= DT_DIR
&& has_path_in_index
)
1266 * When we are looking at a directory P in the working tree,
1267 * there are three cases:
1269 * (1) P exists in the index. Everything inside the directory P in
1270 * the working tree needs to go when P is checked out from the
1273 * (2) P does not exist in the index, but there is P/Q in the index.
1274 * We know P will stay a directory when we check out the contents
1275 * of the index, but we do not know yet if there is a directory
1276 * P/Q in the working tree to be killed, so we need to recurse.
1278 * (3) P does not exist in the index, and there is no P/Q in the index
1279 * to require P to be a directory, either. Only in this case, we
1280 * know that everything inside P will not be killed without
1283 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
1284 (dtype
== DT_DIR
) &&
1285 !has_path_in_index
&&
1286 (directory_exists_in_index(path
->buf
, path
->len
) == index_nonexistent
))
1289 exclude
= is_excluded(dir
, path
->buf
, &dtype
);
1292 * Excluded? If we don't explicitly want to show
1293 * ignored files, ignore it
1295 if (exclude
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
1296 return path_excluded
;
1302 strbuf_addch(path
, '/');
1303 return treat_directory(dir
, path
->buf
, path
->len
, exclude
,
1307 return exclude
? path_excluded
: path_untracked
;
1311 static enum path_treatment
treat_path(struct dir_struct
*dir
,
1313 struct strbuf
*path
,
1315 const struct path_simplify
*simplify
)
1319 if (is_dot_or_dotdot(de
->d_name
) || !strcmp(de
->d_name
, ".git"))
1321 strbuf_setlen(path
, baselen
);
1322 strbuf_addstr(path
, de
->d_name
);
1323 if (simplify_away(path
->buf
, path
->len
, simplify
))
1327 return treat_one_path(dir
, path
, simplify
, dtype
, de
);
1331 * Read a directory tree. We currently ignore anything but
1332 * directories, regular files and symlinks. That's because git
1333 * doesn't handle them at all yet. Maybe that will change some
1336 * Also, we ignore the name ".git" (even if it is not a directory).
1337 * That likely will not change.
1339 * Returns the most significant path_treatment value encountered in the scan.
1341 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
1342 const char *base
, int baselen
,
1344 const struct path_simplify
*simplify
)
1347 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
1349 struct strbuf path
= STRBUF_INIT
;
1351 strbuf_add(&path
, base
, baselen
);
1353 fdir
= opendir(path
.len
? path
.buf
: ".");
1357 while ((de
= readdir(fdir
)) != NULL
) {
1358 /* check how the file or directory should be treated */
1359 state
= treat_path(dir
, de
, &path
, baselen
, simplify
);
1360 if (state
> dir_state
)
1363 /* recurse into subdir if instructed by treat_path */
1364 if (state
== path_recurse
) {
1365 subdir_state
= read_directory_recursive(dir
, path
.buf
,
1366 path
.len
, check_only
, simplify
);
1367 if (subdir_state
> dir_state
)
1368 dir_state
= subdir_state
;
1372 /* abort early if maximum state has been reached */
1373 if (dir_state
== path_untracked
)
1375 /* skip the dir_add_* part */
1379 /* add the path to the appropriate result list */
1382 if (dir
->flags
& DIR_SHOW_IGNORED
)
1383 dir_add_name(dir
, path
.buf
, path
.len
);
1384 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
1385 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
1386 exclude_matches_pathspec(path
.buf
, path
.len
,
1388 dir_add_ignored(dir
, path
.buf
, path
.len
);
1391 case path_untracked
:
1392 if (!(dir
->flags
& DIR_SHOW_IGNORED
))
1393 dir_add_name(dir
, path
.buf
, path
.len
);
1402 strbuf_release(&path
);
1407 static int cmp_name(const void *p1
, const void *p2
)
1409 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
1410 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
1412 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
1415 static struct path_simplify
*create_simplify(const char **pathspec
)
1418 struct path_simplify
*simplify
= NULL
;
1423 for (nr
= 0 ; ; nr
++) {
1425 ALLOC_GROW(simplify
, nr
+ 1, alloc
);
1426 match
= *pathspec
++;
1429 simplify
[nr
].path
= match
;
1430 simplify
[nr
].len
= simple_length(match
);
1432 simplify
[nr
].path
= NULL
;
1433 simplify
[nr
].len
= 0;
1437 static void free_simplify(struct path_simplify
*simplify
)
1442 static int treat_leading_path(struct dir_struct
*dir
,
1443 const char *path
, int len
,
1444 const struct path_simplify
*simplify
)
1446 struct strbuf sb
= STRBUF_INIT
;
1447 int baselen
, rc
= 0;
1449 int old_flags
= dir
->flags
;
1451 while (len
&& path
[len
- 1] == '/')
1456 dir
->flags
&= ~DIR_SHOW_OTHER_DIRECTORIES
;
1458 cp
= path
+ baselen
+ !!baselen
;
1459 cp
= memchr(cp
, '/', path
+ len
- cp
);
1463 baselen
= cp
- path
;
1464 strbuf_setlen(&sb
, 0);
1465 strbuf_add(&sb
, path
, baselen
);
1466 if (!is_directory(sb
.buf
))
1468 if (simplify_away(sb
.buf
, sb
.len
, simplify
))
1470 if (treat_one_path(dir
, &sb
, simplify
,
1471 DT_DIR
, NULL
) == path_none
)
1472 break; /* do not recurse into it */
1473 if (len
<= baselen
) {
1475 break; /* finished checking */
1478 strbuf_release(&sb
);
1479 dir
->flags
= old_flags
;
1483 int read_directory(struct dir_struct
*dir
, const char *path
, int len
, const struct pathspec
*pathspec
)
1485 struct path_simplify
*simplify
;
1488 * Check out create_simplify()
1491 GUARD_PATHSPEC(pathspec
,
1499 if (has_symlink_leading_path(path
, len
))
1503 * exclude patterns are treated like positive ones in
1504 * create_simplify. Usually exclude patterns should be a
1505 * subset of positive ones, which has no impacts on
1506 * create_simplify().
1508 simplify
= create_simplify(pathspec
? pathspec
->_raw
: NULL
);
1509 if (!len
|| treat_leading_path(dir
, path
, len
, simplify
))
1510 read_directory_recursive(dir
, path
, len
, 0, simplify
);
1511 free_simplify(simplify
);
1512 qsort(dir
->entries
, dir
->nr
, sizeof(struct dir_entry
*), cmp_name
);
1513 qsort(dir
->ignored
, dir
->ignored_nr
, sizeof(struct dir_entry
*), cmp_name
);
1517 int file_exists(const char *f
)
1520 return lstat(f
, &sb
) == 0;
1524 * Given two normalized paths (a trailing slash is ok), if subdir is
1525 * outside dir, return -1. Otherwise return the offset in subdir that
1526 * can be used as relative path to dir.
1528 int dir_inside_of(const char *subdir
, const char *dir
)
1532 assert(dir
&& subdir
&& *dir
&& *subdir
);
1534 while (*dir
&& *subdir
&& *dir
== *subdir
) {
1540 /* hel[p]/me vs hel[l]/yeah */
1541 if (*dir
&& *subdir
)
1545 return !*dir
? offset
: -1; /* same dir */
1547 /* foo/[b]ar vs foo/[] */
1548 if (is_dir_sep(dir
[-1]))
1549 return is_dir_sep(subdir
[-1]) ? offset
: -1;
1551 /* foo[/]bar vs foo[] */
1552 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
1555 int is_inside_dir(const char *dir
)
1564 rc
= (dir_inside_of(cwd
, dir
) >= 0);
1569 int is_empty_dir(const char *path
)
1571 DIR *dir
= opendir(path
);
1578 while ((e
= readdir(dir
)) != NULL
)
1579 if (!is_dot_or_dotdot(e
->d_name
)) {
1588 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
1592 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
1593 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
1594 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
1595 unsigned char submodule_head
[20];
1597 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
1598 !resolve_gitlink_ref(path
->buf
, "HEAD", submodule_head
)) {
1599 /* Do not descend and nuke a nested git work tree. */
1605 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
1606 dir
= opendir(path
->buf
);
1608 if (errno
== ENOENT
)
1609 return keep_toplevel
? -1 : 0;
1610 else if (errno
== EACCES
&& !keep_toplevel
)
1612 * An empty dir could be removable even if it
1615 return rmdir(path
->buf
);
1619 if (path
->buf
[original_len
- 1] != '/')
1620 strbuf_addch(path
, '/');
1623 while ((e
= readdir(dir
)) != NULL
) {
1625 if (is_dot_or_dotdot(e
->d_name
))
1628 strbuf_setlen(path
, len
);
1629 strbuf_addstr(path
, e
->d_name
);
1630 if (lstat(path
->buf
, &st
)) {
1631 if (errno
== ENOENT
)
1633 * file disappeared, which is what we
1638 } else if (S_ISDIR(st
.st_mode
)) {
1639 if (!remove_dir_recurse(path
, flag
, &kept_down
))
1640 continue; /* happy */
1641 } else if (!only_empty
&&
1642 (!unlink(path
->buf
) || errno
== ENOENT
)) {
1643 continue; /* happy, too */
1646 /* path too long, stat fails, or non-directory still exists */
1652 strbuf_setlen(path
, original_len
);
1653 if (!ret
&& !keep_toplevel
&& !kept_down
)
1654 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
1657 * report the uplevel that it is not an error that we
1658 * did not rmdir() our directory.
1664 int remove_dir_recursively(struct strbuf
*path
, int flag
)
1666 return remove_dir_recurse(path
, flag
, NULL
);
1669 void setup_standard_excludes(struct dir_struct
*dir
)
1673 dir
->exclude_per_dir
= ".gitignore";
1675 /* core.excludefile defaulting to $XDG_HOME/git/ignore */
1677 excludes_file
= xdg_config_home("ignore");
1678 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
1679 add_excludes_from_file(dir
, excludes_file
);
1681 /* per repository user preference */
1682 path
= git_path("info/exclude");
1683 if (!access_or_warn(path
, R_OK
, 0))
1684 add_excludes_from_file(dir
, path
);
1687 int remove_path(const char *name
)
1691 if (unlink(name
) && errno
!= ENOENT
&& errno
!= ENOTDIR
)
1694 slash
= strrchr(name
, '/');
1696 char *dirs
= xstrdup(name
);
1697 slash
= dirs
+ (slash
- name
);
1700 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
1707 * Frees memory within dir which was allocated for exclude lists and
1708 * the exclude_stack. Does not free dir itself.
1710 void clear_directory(struct dir_struct
*dir
)
1713 struct exclude_list_group
*group
;
1714 struct exclude_list
*el
;
1715 struct exclude_stack
*stk
;
1717 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1718 group
= &dir
->exclude_list_group
[i
];
1719 for (j
= 0; j
< group
->nr
; j
++) {
1722 free((char *)el
->src
);
1723 clear_exclude_list(el
);
1728 stk
= dir
->exclude_stack
;
1730 struct exclude_stack
*prev
= stk
->prev
;
1734 strbuf_release(&dir
->basebuf
);