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.
388 struct strbuf sb
= STRBUF_INIT
;
390 for (num
= 0; num
< pathspec
->nr
; num
++) {
391 int other
, found_dup
;
396 * The caller might have fed identical pathspec
397 * twice. Do not barf on such a mistake.
398 * FIXME: parse_pathspec should have eliminated
399 * duplicate pathspec.
401 for (found_dup
= other
= 0;
402 !found_dup
&& other
< pathspec
->nr
;
404 if (other
== num
|| !ps_matched
[other
])
406 if (!strcmp(pathspec
->items
[other
].original
,
407 pathspec
->items
[num
].original
))
409 * Ok, we have a match already.
416 error("pathspec '%s' did not match any file(s) known to git.",
417 pathspec
->items
[num
].original
);
425 * Return the length of the "simple" part of a path match limiter.
427 int simple_length(const char *match
)
432 unsigned char c
= *match
++;
434 if (c
== '\0' || is_glob_special(c
))
439 int no_wildcard(const char *string
)
441 return string
[simple_length(string
)] == '\0';
444 void parse_exclude_pattern(const char **pattern
,
449 const char *p
= *pattern
;
454 *flags
|= EXC_FLAG_NEGATIVE
;
458 if (len
&& p
[len
- 1] == '/') {
460 *flags
|= EXC_FLAG_MUSTBEDIR
;
462 for (i
= 0; i
< len
; i
++) {
467 *flags
|= EXC_FLAG_NODIR
;
468 *nowildcardlen
= simple_length(p
);
470 * we should have excluded the trailing slash from 'p' too,
471 * but that's one more allocation. Instead just make sure
472 * nowildcardlen does not exceed real patternlen
474 if (*nowildcardlen
> len
)
475 *nowildcardlen
= len
;
476 if (*p
== '*' && no_wildcard(p
+ 1))
477 *flags
|= EXC_FLAG_ENDSWITH
;
482 void add_exclude(const char *string
, const char *base
,
483 int baselen
, struct exclude_list
*el
, int srcpos
)
490 parse_exclude_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
491 if (flags
& EXC_FLAG_MUSTBEDIR
) {
493 x
= xmalloc(sizeof(*x
) + patternlen
+ 1);
495 memcpy(s
, string
, patternlen
);
496 s
[patternlen
] = '\0';
499 x
= xmalloc(sizeof(*x
));
502 x
->patternlen
= patternlen
;
503 x
->nowildcardlen
= nowildcardlen
;
505 x
->baselen
= baselen
;
508 ALLOC_GROW(el
->excludes
, el
->nr
+ 1, el
->alloc
);
509 el
->excludes
[el
->nr
++] = x
;
513 static void *read_skip_worktree_file_from_index(const char *path
, size_t *size
)
517 enum object_type type
;
521 pos
= cache_name_pos(path
, len
);
524 if (!ce_skip_worktree(active_cache
[pos
]))
526 data
= read_sha1_file(active_cache
[pos
]->sha1
, &type
, &sz
);
527 if (!data
|| type
!= OBJ_BLOB
) {
536 * Frees memory within el which was allocated for exclude patterns and
537 * the file buffer. Does not free el itself.
539 void clear_exclude_list(struct exclude_list
*el
)
543 for (i
= 0; i
< el
->nr
; i
++)
544 free(el
->excludes
[i
]);
553 static void trim_trailing_spaces(char *buf
)
555 char *p
, *last_space
= NULL
;
557 for (p
= buf
; *p
; p
++)
576 int add_excludes_from_file_to_list(const char *fname
,
579 struct exclude_list
*el
,
583 int fd
, i
, lineno
= 1;
587 fd
= open(fname
, O_RDONLY
);
588 if (fd
< 0 || fstat(fd
, &st
) < 0) {
590 warn_on_inaccessible(fname
);
594 (buf
= read_skip_worktree_file_from_index(fname
, &size
)) == NULL
)
600 if (buf
[size
-1] != '\n') {
601 buf
= xrealloc(buf
, size
+1);
605 size
= xsize_t(st
.st_size
);
610 buf
= xmalloc(size
+1);
611 if (read_in_full(fd
, buf
, size
) != size
) {
622 if (skip_utf8_bom(&buf
, size
))
623 size
-= buf
- el
->filebuf
;
627 for (i
= 0; i
< size
; i
++) {
628 if (buf
[i
] == '\n') {
629 if (entry
!= buf
+ i
&& entry
[0] != '#') {
630 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
631 trim_trailing_spaces(entry
);
632 add_exclude(entry
, base
, baselen
, el
, lineno
);
641 struct exclude_list
*add_exclude_list(struct dir_struct
*dir
,
642 int group_type
, const char *src
)
644 struct exclude_list
*el
;
645 struct exclude_list_group
*group
;
647 group
= &dir
->exclude_list_group
[group_type
];
648 ALLOC_GROW(group
->el
, group
->nr
+ 1, group
->alloc
);
649 el
= &group
->el
[group
->nr
++];
650 memset(el
, 0, sizeof(*el
));
656 * Used to set up core.excludesfile and .git/info/exclude lists.
658 void add_excludes_from_file(struct dir_struct
*dir
, const char *fname
)
660 struct exclude_list
*el
;
661 el
= add_exclude_list(dir
, EXC_FILE
, fname
);
662 if (add_excludes_from_file_to_list(fname
, "", 0, el
, 0) < 0)
663 die("cannot use %s as an exclude file", fname
);
666 int match_basename(const char *basename
, int basenamelen
,
667 const char *pattern
, int prefix
, int patternlen
,
670 if (prefix
== patternlen
) {
671 if (patternlen
== basenamelen
&&
672 !strncmp_icase(pattern
, basename
, basenamelen
))
674 } else if (flags
& EXC_FLAG_ENDSWITH
) {
675 /* "*literal" matching against "fooliteral" */
676 if (patternlen
- 1 <= basenamelen
&&
677 !strncmp_icase(pattern
+ 1,
678 basename
+ basenamelen
- (patternlen
- 1),
682 if (fnmatch_icase_mem(pattern
, patternlen
,
683 basename
, basenamelen
,
690 int match_pathname(const char *pathname
, int pathlen
,
691 const char *base
, int baselen
,
692 const char *pattern
, int prefix
, int patternlen
,
699 * match with FNM_PATHNAME; the pattern has base implicitly
702 if (*pattern
== '/') {
709 * baselen does not count the trailing slash. base[] may or
710 * may not end with a trailing slash though.
712 if (pathlen
< baselen
+ 1 ||
713 (baselen
&& pathname
[baselen
] != '/') ||
714 strncmp_icase(pathname
, base
, baselen
))
717 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
718 name
= pathname
+ pathlen
- namelen
;
722 * if the non-wildcard part is longer than the
723 * remaining pathname, surely it cannot match.
725 if (prefix
> namelen
)
728 if (strncmp_icase(pattern
, name
, prefix
))
731 patternlen
-= prefix
;
736 * If the whole pattern did not have a wildcard,
737 * then our prefix match is all we need; we
738 * do not need to call fnmatch at all.
740 if (!patternlen
&& !namelen
)
744 return fnmatch_icase_mem(pattern
, patternlen
,
750 * Scan the given exclude list in reverse to see whether pathname
751 * should be ignored. The first match (i.e. the last on the list), if
752 * any, determines the fate. Returns the exclude_list element which
753 * matched, or NULL for undecided.
755 static struct exclude
*last_exclude_matching_from_list(const char *pathname
,
757 const char *basename
,
759 struct exclude_list
*el
)
764 return NULL
; /* undefined */
766 for (i
= el
->nr
- 1; 0 <= i
; i
--) {
767 struct exclude
*x
= el
->excludes
[i
];
768 const char *exclude
= x
->pattern
;
769 int prefix
= x
->nowildcardlen
;
771 if (x
->flags
& EXC_FLAG_MUSTBEDIR
) {
772 if (*dtype
== DT_UNKNOWN
)
773 *dtype
= get_dtype(NULL
, pathname
, pathlen
);
774 if (*dtype
!= DT_DIR
)
778 if (x
->flags
& EXC_FLAG_NODIR
) {
779 if (match_basename(basename
,
780 pathlen
- (basename
- pathname
),
781 exclude
, prefix
, x
->patternlen
,
787 assert(x
->baselen
== 0 || x
->base
[x
->baselen
- 1] == '/');
788 if (match_pathname(pathname
, pathlen
,
789 x
->base
, x
->baselen
? x
->baselen
- 1 : 0,
790 exclude
, prefix
, x
->patternlen
, x
->flags
))
793 return NULL
; /* undecided */
797 * Scan the list and let the last match determine the fate.
798 * Return 1 for exclude, 0 for include and -1 for undecided.
800 int is_excluded_from_list(const char *pathname
,
801 int pathlen
, const char *basename
, int *dtype
,
802 struct exclude_list
*el
)
804 struct exclude
*exclude
;
805 exclude
= last_exclude_matching_from_list(pathname
, pathlen
, basename
, dtype
, el
);
807 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
808 return -1; /* undecided */
811 static struct exclude
*last_exclude_matching_from_lists(struct dir_struct
*dir
,
812 const char *pathname
, int pathlen
, const char *basename
,
816 struct exclude_list_group
*group
;
817 struct exclude
*exclude
;
818 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
819 group
= &dir
->exclude_list_group
[i
];
820 for (j
= group
->nr
- 1; j
>= 0; j
--) {
821 exclude
= last_exclude_matching_from_list(
822 pathname
, pathlen
, basename
, dtype_p
,
832 * Loads the per-directory exclude list for the substring of base
833 * which has a char length of baselen.
835 static void prep_exclude(struct dir_struct
*dir
, const char *base
, int baselen
)
837 struct exclude_list_group
*group
;
838 struct exclude_list
*el
;
839 struct exclude_stack
*stk
= NULL
;
842 group
= &dir
->exclude_list_group
[EXC_DIRS
];
845 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
846 * which originate from directories not in the prefix of the
847 * path being checked.
849 while ((stk
= dir
->exclude_stack
) != NULL
) {
850 if (stk
->baselen
<= baselen
&&
851 !strncmp(dir
->basebuf
.buf
, base
, stk
->baselen
))
853 el
= &group
->el
[dir
->exclude_stack
->exclude_ix
];
854 dir
->exclude_stack
= stk
->prev
;
856 free((char *)el
->src
); /* see strbuf_detach() below */
857 clear_exclude_list(el
);
862 /* Skip traversing into sub directories if the parent is excluded */
867 * Lazy initialization. All call sites currently just
868 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
869 * them seems lots of work for little benefit.
871 if (!dir
->basebuf
.buf
)
872 strbuf_init(&dir
->basebuf
, PATH_MAX
);
874 /* Read from the parent directories and push them down. */
875 current
= stk
? stk
->baselen
: -1;
876 strbuf_setlen(&dir
->basebuf
, current
< 0 ? 0 : current
);
877 while (current
< baselen
) {
880 stk
= xcalloc(1, sizeof(*stk
));
885 cp
= strchr(base
+ current
+ 1, '/');
887 die("oops in prep_exclude");
890 stk
->prev
= dir
->exclude_stack
;
891 stk
->baselen
= cp
- base
;
892 stk
->exclude_ix
= group
->nr
;
893 el
= add_exclude_list(dir
, EXC_DIRS
, NULL
);
894 strbuf_add(&dir
->basebuf
, base
+ current
, stk
->baselen
- current
);
895 assert(stk
->baselen
== dir
->basebuf
.len
);
897 /* Abort if the directory is excluded */
900 dir
->basebuf
.buf
[stk
->baselen
- 1] = 0;
901 dir
->exclude
= last_exclude_matching_from_lists(dir
,
902 dir
->basebuf
.buf
, stk
->baselen
- 1,
903 dir
->basebuf
.buf
+ current
, &dt
);
904 dir
->basebuf
.buf
[stk
->baselen
- 1] = '/';
906 dir
->exclude
->flags
& EXC_FLAG_NEGATIVE
)
909 dir
->exclude_stack
= stk
;
914 /* Try to read per-directory file */
915 if (dir
->exclude_per_dir
) {
917 * dir->basebuf gets reused by the traversal, but we
918 * need fname to remain unchanged to ensure the src
919 * member of each struct exclude correctly
920 * back-references its source file. Other invocations
921 * of add_exclude_list provide stable strings, so we
922 * strbuf_detach() and free() here in the caller.
924 struct strbuf sb
= STRBUF_INIT
;
925 strbuf_addbuf(&sb
, &dir
->basebuf
);
926 strbuf_addstr(&sb
, dir
->exclude_per_dir
);
927 el
->src
= strbuf_detach(&sb
, NULL
);
928 add_excludes_from_file_to_list(el
->src
, el
->src
,
929 stk
->baselen
, el
, 1);
931 dir
->exclude_stack
= stk
;
932 current
= stk
->baselen
;
934 strbuf_setlen(&dir
->basebuf
, baselen
);
938 * Loads the exclude lists for the directory containing pathname, then
939 * scans all exclude lists to determine whether pathname is excluded.
940 * Returns the exclude_list element which matched, or NULL for
943 struct exclude
*last_exclude_matching(struct dir_struct
*dir
,
944 const char *pathname
,
947 int pathlen
= strlen(pathname
);
948 const char *basename
= strrchr(pathname
, '/');
949 basename
= (basename
) ? basename
+1 : pathname
;
951 prep_exclude(dir
, pathname
, basename
-pathname
);
956 return last_exclude_matching_from_lists(dir
, pathname
, pathlen
,
961 * Loads the exclude lists for the directory containing pathname, then
962 * scans all exclude lists to determine whether pathname is excluded.
963 * Returns 1 if true, otherwise 0.
965 int is_excluded(struct dir_struct
*dir
, const char *pathname
, int *dtype_p
)
967 struct exclude
*exclude
=
968 last_exclude_matching(dir
, pathname
, dtype_p
);
970 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
974 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
976 struct dir_entry
*ent
;
978 ent
= xmalloc(sizeof(*ent
) + len
+ 1);
980 memcpy(ent
->name
, pathname
, len
);
985 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
, const char *pathname
, int len
)
987 if (cache_file_exists(pathname
, len
, ignore_case
))
990 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->alloc
);
991 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
994 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
, const char *pathname
, int len
)
996 if (!cache_name_is_other(pathname
, len
))
999 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->ignored_alloc
);
1000 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
1004 index_nonexistent
= 0,
1010 * Do not use the alphabetically sorted index to look up
1011 * the directory name; instead, use the case insensitive
1014 static enum exist_status
directory_exists_in_index_icase(const char *dirname
, int len
)
1016 const struct cache_entry
*ce
= cache_dir_exists(dirname
, len
);
1017 unsigned char endchar
;
1020 return index_nonexistent
;
1021 endchar
= ce
->name
[len
];
1024 * The cache_entry structure returned will contain this dirname
1025 * and possibly additional path components.
1028 return index_directory
;
1031 * If there are no additional path components, then this cache_entry
1032 * represents a submodule. Submodules, despite being directories,
1033 * are stored in the cache without a closing slash.
1035 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1036 return index_gitdir
;
1038 /* This should never be hit, but it exists just in case. */
1039 return index_nonexistent
;
1043 * The index sorts alphabetically by entry name, which
1044 * means that a gitlink sorts as '\0' at the end, while
1045 * a directory (which is defined not as an entry, but as
1046 * the files it contains) will sort with the '/' at the
1049 static enum exist_status
directory_exists_in_index(const char *dirname
, int len
)
1054 return directory_exists_in_index_icase(dirname
, len
);
1056 pos
= cache_name_pos(dirname
, len
);
1059 while (pos
< active_nr
) {
1060 const struct cache_entry
*ce
= active_cache
[pos
++];
1061 unsigned char endchar
;
1063 if (strncmp(ce
->name
, dirname
, len
))
1065 endchar
= ce
->name
[len
];
1069 return index_directory
;
1070 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1071 return index_gitdir
;
1073 return index_nonexistent
;
1077 * When we find a directory when traversing the filesystem, we
1078 * have three distinct cases:
1081 * - see it as a directory
1084 * and which one we choose depends on a combination of existing
1085 * git index contents and the flags passed into the directory
1086 * traversal routine.
1088 * Case 1: If we *already* have entries in the index under that
1089 * directory name, we always recurse into the directory to see
1092 * Case 2: If we *already* have that directory name as a gitlink,
1093 * we always continue to see it as a gitlink, regardless of whether
1094 * there is an actual git directory there or not (it might not
1095 * be checked out as a subproject!)
1097 * Case 3: if we didn't have it in the index previously, we
1098 * have a few sub-cases:
1100 * (a) if "show_other_directories" is true, we show it as
1101 * just a directory, unless "hide_empty_directories" is
1102 * also true, in which case we need to check if it contains any
1103 * untracked and / or ignored files.
1104 * (b) if it looks like a git directory, and we don't have
1105 * 'no_gitlinks' set we treat it as a gitlink, and show it
1107 * (c) otherwise, we recurse into it.
1109 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1110 const char *dirname
, int len
, int exclude
,
1111 const struct path_simplify
*simplify
)
1113 /* The "len-1" is to strip the final '/' */
1114 switch (directory_exists_in_index(dirname
, len
-1)) {
1115 case index_directory
:
1116 return path_recurse
;
1121 case index_nonexistent
:
1122 if (dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)
1124 if (!(dir
->flags
& DIR_NO_GITLINKS
)) {
1125 unsigned char sha1
[20];
1126 if (resolve_gitlink_ref(dirname
, "HEAD", sha1
) == 0)
1127 return path_untracked
;
1129 return path_recurse
;
1132 /* This is the "show_other_directories" case */
1134 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1135 return exclude
? path_excluded
: path_untracked
;
1137 return read_directory_recursive(dir
, dirname
, len
, 1, simplify
);
1141 * This is an inexact early pruning of any recursive directory
1142 * reading - if the path cannot possibly be in the pathspec,
1143 * return true, and we'll skip it early.
1145 static int simplify_away(const char *path
, int pathlen
, const struct path_simplify
*simplify
)
1149 const char *match
= simplify
->path
;
1150 int len
= simplify
->len
;
1156 if (!memcmp(path
, match
, len
))
1166 * This function tells us whether an excluded path matches a
1167 * list of "interesting" pathspecs. That is, whether a path matched
1168 * by any of the pathspecs could possibly be ignored by excluding
1169 * the specified path. This can happen if:
1171 * 1. the path is mentioned explicitly in the pathspec
1173 * 2. the path is a directory prefix of some element in the
1176 static int exclude_matches_pathspec(const char *path
, int len
,
1177 const struct path_simplify
*simplify
)
1180 for (; simplify
->path
; simplify
++) {
1181 if (len
== simplify
->len
1182 && !memcmp(path
, simplify
->path
, len
))
1184 if (len
< simplify
->len
1185 && simplify
->path
[len
] == '/'
1186 && !memcmp(path
, simplify
->path
, len
))
1193 static int get_index_dtype(const char *path
, int len
)
1196 const struct cache_entry
*ce
;
1198 ce
= cache_file_exists(path
, len
, 0);
1200 if (!ce_uptodate(ce
))
1202 if (S_ISGITLINK(ce
->ce_mode
))
1205 * Nobody actually cares about the
1206 * difference between DT_LNK and DT_REG
1211 /* Try to look it up as a directory */
1212 pos
= cache_name_pos(path
, len
);
1216 while (pos
< active_nr
) {
1217 ce
= active_cache
[pos
++];
1218 if (strncmp(ce
->name
, path
, len
))
1220 if (ce
->name
[len
] > '/')
1222 if (ce
->name
[len
] < '/')
1224 if (!ce_uptodate(ce
))
1225 break; /* continue? */
1231 static int get_dtype(struct dirent
*de
, const char *path
, int len
)
1233 int dtype
= de
? DTYPE(de
) : DT_UNKNOWN
;
1236 if (dtype
!= DT_UNKNOWN
)
1238 dtype
= get_index_dtype(path
, len
);
1239 if (dtype
!= DT_UNKNOWN
)
1241 if (lstat(path
, &st
))
1243 if (S_ISREG(st
.st_mode
))
1245 if (S_ISDIR(st
.st_mode
))
1247 if (S_ISLNK(st
.st_mode
))
1252 static enum path_treatment
treat_one_path(struct dir_struct
*dir
,
1253 struct strbuf
*path
,
1254 const struct path_simplify
*simplify
,
1255 int dtype
, struct dirent
*de
)
1258 int has_path_in_index
= !!cache_file_exists(path
->buf
, path
->len
, ignore_case
);
1260 if (dtype
== DT_UNKNOWN
)
1261 dtype
= get_dtype(de
, path
->buf
, path
->len
);
1263 /* Always exclude indexed files */
1264 if (dtype
!= DT_DIR
&& has_path_in_index
)
1268 * When we are looking at a directory P in the working tree,
1269 * there are three cases:
1271 * (1) P exists in the index. Everything inside the directory P in
1272 * the working tree needs to go when P is checked out from the
1275 * (2) P does not exist in the index, but there is P/Q in the index.
1276 * We know P will stay a directory when we check out the contents
1277 * of the index, but we do not know yet if there is a directory
1278 * P/Q in the working tree to be killed, so we need to recurse.
1280 * (3) P does not exist in the index, and there is no P/Q in the index
1281 * to require P to be a directory, either. Only in this case, we
1282 * know that everything inside P will not be killed without
1285 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
1286 (dtype
== DT_DIR
) &&
1287 !has_path_in_index
&&
1288 (directory_exists_in_index(path
->buf
, path
->len
) == index_nonexistent
))
1291 exclude
= is_excluded(dir
, path
->buf
, &dtype
);
1294 * Excluded? If we don't explicitly want to show
1295 * ignored files, ignore it
1297 if (exclude
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
1298 return path_excluded
;
1304 strbuf_addch(path
, '/');
1305 return treat_directory(dir
, path
->buf
, path
->len
, exclude
,
1309 return exclude
? path_excluded
: path_untracked
;
1313 static enum path_treatment
treat_path(struct dir_struct
*dir
,
1315 struct strbuf
*path
,
1317 const struct path_simplify
*simplify
)
1321 if (is_dot_or_dotdot(de
->d_name
) || !strcmp(de
->d_name
, ".git"))
1323 strbuf_setlen(path
, baselen
);
1324 strbuf_addstr(path
, de
->d_name
);
1325 if (simplify_away(path
->buf
, path
->len
, simplify
))
1329 return treat_one_path(dir
, path
, simplify
, dtype
, de
);
1333 * Read a directory tree. We currently ignore anything but
1334 * directories, regular files and symlinks. That's because git
1335 * doesn't handle them at all yet. Maybe that will change some
1338 * Also, we ignore the name ".git" (even if it is not a directory).
1339 * That likely will not change.
1341 * Returns the most significant path_treatment value encountered in the scan.
1343 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
1344 const char *base
, int baselen
,
1346 const struct path_simplify
*simplify
)
1349 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
1351 struct strbuf path
= STRBUF_INIT
;
1353 strbuf_add(&path
, base
, baselen
);
1355 fdir
= opendir(path
.len
? path
.buf
: ".");
1359 while ((de
= readdir(fdir
)) != NULL
) {
1360 /* check how the file or directory should be treated */
1361 state
= treat_path(dir
, de
, &path
, baselen
, simplify
);
1362 if (state
> dir_state
)
1365 /* recurse into subdir if instructed by treat_path */
1366 if (state
== path_recurse
) {
1367 subdir_state
= read_directory_recursive(dir
, path
.buf
,
1368 path
.len
, check_only
, simplify
);
1369 if (subdir_state
> dir_state
)
1370 dir_state
= subdir_state
;
1374 /* abort early if maximum state has been reached */
1375 if (dir_state
== path_untracked
)
1377 /* skip the dir_add_* part */
1381 /* add the path to the appropriate result list */
1384 if (dir
->flags
& DIR_SHOW_IGNORED
)
1385 dir_add_name(dir
, path
.buf
, path
.len
);
1386 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
1387 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
1388 exclude_matches_pathspec(path
.buf
, path
.len
,
1390 dir_add_ignored(dir
, path
.buf
, path
.len
);
1393 case path_untracked
:
1394 if (!(dir
->flags
& DIR_SHOW_IGNORED
))
1395 dir_add_name(dir
, path
.buf
, path
.len
);
1404 strbuf_release(&path
);
1409 static int cmp_name(const void *p1
, const void *p2
)
1411 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
1412 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
1414 return name_compare(e1
->name
, e1
->len
, e2
->name
, e2
->len
);
1417 static struct path_simplify
*create_simplify(const char **pathspec
)
1420 struct path_simplify
*simplify
= NULL
;
1425 for (nr
= 0 ; ; nr
++) {
1427 ALLOC_GROW(simplify
, nr
+ 1, alloc
);
1428 match
= *pathspec
++;
1431 simplify
[nr
].path
= match
;
1432 simplify
[nr
].len
= simple_length(match
);
1434 simplify
[nr
].path
= NULL
;
1435 simplify
[nr
].len
= 0;
1439 static void free_simplify(struct path_simplify
*simplify
)
1444 static int treat_leading_path(struct dir_struct
*dir
,
1445 const char *path
, int len
,
1446 const struct path_simplify
*simplify
)
1448 struct strbuf sb
= STRBUF_INIT
;
1449 int baselen
, rc
= 0;
1451 int old_flags
= dir
->flags
;
1453 while (len
&& path
[len
- 1] == '/')
1458 dir
->flags
&= ~DIR_SHOW_OTHER_DIRECTORIES
;
1460 cp
= path
+ baselen
+ !!baselen
;
1461 cp
= memchr(cp
, '/', path
+ len
- cp
);
1465 baselen
= cp
- path
;
1466 strbuf_setlen(&sb
, 0);
1467 strbuf_add(&sb
, path
, baselen
);
1468 if (!is_directory(sb
.buf
))
1470 if (simplify_away(sb
.buf
, sb
.len
, simplify
))
1472 if (treat_one_path(dir
, &sb
, simplify
,
1473 DT_DIR
, NULL
) == path_none
)
1474 break; /* do not recurse into it */
1475 if (len
<= baselen
) {
1477 break; /* finished checking */
1480 strbuf_release(&sb
);
1481 dir
->flags
= old_flags
;
1485 int read_directory(struct dir_struct
*dir
, const char *path
, int len
, const struct pathspec
*pathspec
)
1487 struct path_simplify
*simplify
;
1490 * Check out create_simplify()
1493 GUARD_PATHSPEC(pathspec
,
1501 if (has_symlink_leading_path(path
, len
))
1505 * exclude patterns are treated like positive ones in
1506 * create_simplify. Usually exclude patterns should be a
1507 * subset of positive ones, which has no impacts on
1508 * create_simplify().
1510 simplify
= create_simplify(pathspec
? pathspec
->_raw
: NULL
);
1511 if (!len
|| treat_leading_path(dir
, path
, len
, simplify
))
1512 read_directory_recursive(dir
, path
, len
, 0, simplify
);
1513 free_simplify(simplify
);
1514 qsort(dir
->entries
, dir
->nr
, sizeof(struct dir_entry
*), cmp_name
);
1515 qsort(dir
->ignored
, dir
->ignored_nr
, sizeof(struct dir_entry
*), cmp_name
);
1519 int file_exists(const char *f
)
1522 return lstat(f
, &sb
) == 0;
1526 * Given two normalized paths (a trailing slash is ok), if subdir is
1527 * outside dir, return -1. Otherwise return the offset in subdir that
1528 * can be used as relative path to dir.
1530 int dir_inside_of(const char *subdir
, const char *dir
)
1534 assert(dir
&& subdir
&& *dir
&& *subdir
);
1536 while (*dir
&& *subdir
&& *dir
== *subdir
) {
1542 /* hel[p]/me vs hel[l]/yeah */
1543 if (*dir
&& *subdir
)
1547 return !*dir
? offset
: -1; /* same dir */
1549 /* foo/[b]ar vs foo/[] */
1550 if (is_dir_sep(dir
[-1]))
1551 return is_dir_sep(subdir
[-1]) ? offset
: -1;
1553 /* foo[/]bar vs foo[] */
1554 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
1557 int is_inside_dir(const char *dir
)
1566 rc
= (dir_inside_of(cwd
, dir
) >= 0);
1571 int is_empty_dir(const char *path
)
1573 DIR *dir
= opendir(path
);
1580 while ((e
= readdir(dir
)) != NULL
)
1581 if (!is_dot_or_dotdot(e
->d_name
)) {
1590 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
1594 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
1595 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
1596 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
1597 unsigned char submodule_head
[20];
1599 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
1600 !resolve_gitlink_ref(path
->buf
, "HEAD", submodule_head
)) {
1601 /* Do not descend and nuke a nested git work tree. */
1607 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
1608 dir
= opendir(path
->buf
);
1610 if (errno
== ENOENT
)
1611 return keep_toplevel
? -1 : 0;
1612 else if (errno
== EACCES
&& !keep_toplevel
)
1614 * An empty dir could be removable even if it
1617 return rmdir(path
->buf
);
1621 if (path
->buf
[original_len
- 1] != '/')
1622 strbuf_addch(path
, '/');
1625 while ((e
= readdir(dir
)) != NULL
) {
1627 if (is_dot_or_dotdot(e
->d_name
))
1630 strbuf_setlen(path
, len
);
1631 strbuf_addstr(path
, e
->d_name
);
1632 if (lstat(path
->buf
, &st
)) {
1633 if (errno
== ENOENT
)
1635 * file disappeared, which is what we
1640 } else if (S_ISDIR(st
.st_mode
)) {
1641 if (!remove_dir_recurse(path
, flag
, &kept_down
))
1642 continue; /* happy */
1643 } else if (!only_empty
&&
1644 (!unlink(path
->buf
) || errno
== ENOENT
)) {
1645 continue; /* happy, too */
1648 /* path too long, stat fails, or non-directory still exists */
1654 strbuf_setlen(path
, original_len
);
1655 if (!ret
&& !keep_toplevel
&& !kept_down
)
1656 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
1659 * report the uplevel that it is not an error that we
1660 * did not rmdir() our directory.
1666 int remove_dir_recursively(struct strbuf
*path
, int flag
)
1668 return remove_dir_recurse(path
, flag
, NULL
);
1671 void setup_standard_excludes(struct dir_struct
*dir
)
1675 dir
->exclude_per_dir
= ".gitignore";
1676 path
= git_path("info/exclude");
1678 excludes_file
= xdg_config_home("ignore");
1679 if (!access_or_warn(path
, R_OK
, 0))
1680 add_excludes_from_file(dir
, path
);
1681 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
1682 add_excludes_from_file(dir
, excludes_file
);
1685 int remove_path(const char *name
)
1689 if (unlink(name
) && errno
!= ENOENT
&& errno
!= ENOTDIR
)
1692 slash
= strrchr(name
, '/');
1694 char *dirs
= xstrdup(name
);
1695 slash
= dirs
+ (slash
- name
);
1698 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
1705 * Frees memory within dir which was allocated for exclude lists and
1706 * the exclude_stack. Does not free dir itself.
1708 void clear_directory(struct dir_struct
*dir
)
1711 struct exclude_list_group
*group
;
1712 struct exclude_list
*el
;
1713 struct exclude_stack
*stk
;
1715 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1716 group
= &dir
->exclude_list_group
[i
];
1717 for (j
= 0; j
< group
->nr
; j
++) {
1720 free((char *)el
->src
);
1721 clear_exclude_list(el
);
1726 stk
= dir
->exclude_stack
;
1728 struct exclude_stack
*prev
= stk
->prev
;
1732 strbuf_release(&dir
->basebuf
);