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"
16 struct path_simplify
{
22 * Tells read_directory_recursive how a file or directory should be treated.
23 * Values are ordered by significance, e.g. if a directory contains both
24 * excluded and untracked files, it is listed as untracked because
25 * path_untracked > path_excluded.
34 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
35 const char *path
, int len
,
36 int check_only
, const struct path_simplify
*simplify
);
37 static int get_dtype(struct dirent
*de
, const char *path
, int len
);
39 /* helper string functions with support for the ignore_case flag */
40 int strcmp_icase(const char *a
, const char *b
)
42 return ignore_case
? strcasecmp(a
, b
) : strcmp(a
, b
);
45 int strncmp_icase(const char *a
, const char *b
, size_t count
)
47 return ignore_case
? strncasecmp(a
, b
, count
) : strncmp(a
, b
, count
);
50 int fnmatch_icase(const char *pattern
, const char *string
, int flags
)
52 return wildmatch(pattern
, string
,
53 flags
| (ignore_case
? WM_CASEFOLD
: 0),
57 int git_fnmatch(const struct pathspec_item
*item
,
58 const char *pattern
, const char *string
,
62 if (ps_strncmp(item
, pattern
, string
, prefix
))
67 if (item
->flags
& PATHSPEC_ONESTAR
) {
68 int pattern_len
= strlen(++pattern
);
69 int string_len
= strlen(string
);
70 return string_len
< pattern_len
||
71 ps_strcmp(item
, pattern
,
72 string
+ string_len
- pattern_len
);
74 if (item
->magic
& PATHSPEC_GLOB
)
75 return wildmatch(pattern
, string
,
77 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0),
80 /* wildmatch has not learned no FNM_PATHNAME mode yet */
81 return wildmatch(pattern
, string
,
82 item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0,
86 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
87 const char *string
, int stringlen
,
91 struct strbuf pat_buf
= STRBUF_INIT
;
92 struct strbuf str_buf
= STRBUF_INIT
;
93 const char *use_pat
= pattern
;
94 const char *use_str
= string
;
96 if (pattern
[patternlen
]) {
97 strbuf_add(&pat_buf
, pattern
, patternlen
);
98 use_pat
= pat_buf
.buf
;
100 if (string
[stringlen
]) {
101 strbuf_add(&str_buf
, string
, stringlen
);
102 use_str
= str_buf
.buf
;
106 flags
|= WM_CASEFOLD
;
107 match_status
= wildmatch(use_pat
, use_str
, flags
, NULL
);
109 strbuf_release(&pat_buf
);
110 strbuf_release(&str_buf
);
115 static size_t common_prefix_len(const struct pathspec
*pathspec
)
121 * ":(icase)path" is treated as a pathspec full of
122 * wildcard. In other words, only prefix is considered common
123 * prefix. If the pathspec is abc/foo abc/bar, running in
124 * subdir xyz, the common prefix is still xyz, not xuz/abc as
127 GUARD_PATHSPEC(pathspec
,
135 for (n
= 0; n
< pathspec
->nr
; n
++) {
136 size_t i
= 0, len
= 0, item_len
;
137 if (pathspec
->items
[n
].magic
& PATHSPEC_EXCLUDE
)
139 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
140 item_len
= pathspec
->items
[n
].prefix
;
142 item_len
= pathspec
->items
[n
].nowildcard_len
;
143 while (i
< item_len
&& (n
== 0 || i
< max
)) {
144 char c
= pathspec
->items
[n
].match
[i
];
145 if (c
!= pathspec
->items
[0].match
[i
])
151 if (n
== 0 || len
< max
) {
161 * Returns a copy of the longest leading path common among all
164 char *common_prefix(const struct pathspec
*pathspec
)
166 unsigned long len
= common_prefix_len(pathspec
);
168 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
171 int fill_directory(struct dir_struct
*dir
, const struct pathspec
*pathspec
)
176 * Calculate common prefix for the pathspec, and
177 * use that to optimize the directory walk
179 len
= common_prefix_len(pathspec
);
181 /* Read the directory and prune it */
182 read_directory(dir
, pathspec
->nr
? pathspec
->_raw
[0] : "", len
, pathspec
);
186 int within_depth(const char *name
, int namelen
,
187 int depth
, int max_depth
)
189 const char *cp
= name
, *cpe
= name
+ namelen
;
195 if (depth
> max_depth
)
201 #define DO_MATCH_EXCLUDE 1
202 #define DO_MATCH_DIRECTORY 2
205 * Does 'match' match the given name?
206 * A match is found if
208 * (1) the 'match' string is leading directory of 'name', or
209 * (2) the 'match' string is a wildcard and matches 'name', or
210 * (3) the 'match' string is exactly the same as 'name'.
212 * and the return value tells which case it was.
214 * It returns 0 when there is no match.
216 static int match_pathspec_item(const struct pathspec_item
*item
, int prefix
,
217 const char *name
, int namelen
, unsigned flags
)
219 /* name/namelen has prefix cut off by caller */
220 const char *match
= item
->match
+ prefix
;
221 int matchlen
= item
->len
- prefix
;
224 * The normal call pattern is:
225 * 1. prefix = common_prefix_len(ps);
226 * 2. prune something, or fill_directory
227 * 3. match_pathspec()
229 * 'prefix' at #1 may be shorter than the command's prefix and
230 * it's ok for #2 to match extra files. Those extras will be
233 * Suppose the pathspec is 'foo' and '../bar' running from
234 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
235 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
236 * user does not want XYZ/foo, only the "foo" part should be
237 * case-insensitive. We need to filter out XYZ/foo here. In
238 * other words, we do not trust the caller on comparing the
239 * prefix part when :(icase) is involved. We do exact
240 * comparison ourselves.
242 * Normally the caller (common_prefix_len() in fact) does
243 * _exact_ matching on name[-prefix+1..-1] and we do not need
244 * to check that part. Be defensive and check it anyway, in
245 * case common_prefix_len is changed, or a new caller is
246 * introduced that does not use common_prefix_len.
248 * If the penalty turns out too high when prefix is really
249 * long, maybe change it to
250 * strncmp(match, name, item->prefix - prefix)
252 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
253 strncmp(item
->match
, name
- prefix
, item
->prefix
))
256 /* If the match was just the prefix, we matched */
258 return MATCHED_RECURSIVELY
;
260 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
261 if (matchlen
== namelen
)
262 return MATCHED_EXACTLY
;
264 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
265 return MATCHED_RECURSIVELY
;
266 } else if ((flags
& DO_MATCH_DIRECTORY
) &&
267 match
[matchlen
- 1] == '/' &&
268 namelen
== matchlen
- 1 &&
269 !ps_strncmp(item
, match
, name
, namelen
))
270 return MATCHED_EXACTLY
;
272 if (item
->nowildcard_len
< item
->len
&&
273 !git_fnmatch(item
, match
, name
,
274 item
->nowildcard_len
- prefix
))
275 return MATCHED_FNMATCH
;
281 * Given a name and a list of pathspecs, returns the nature of the
282 * closest (i.e. most specific) match of the name to any of the
285 * The caller typically calls this multiple times with the same
286 * pathspec and seen[] array but with different name/namelen
287 * (e.g. entries from the index) and is interested in seeing if and
288 * how each pathspec matches all the names it calls this function
289 * with. A mark is left in the seen[] array for each pathspec element
290 * indicating the closest type of match that element achieved, so if
291 * seen[n] remains zero after multiple invocations, that means the nth
292 * pathspec did not match any names, which could indicate that the
293 * user mistyped the nth pathspec.
295 static int do_match_pathspec(const struct pathspec
*ps
,
296 const char *name
, int namelen
,
297 int prefix
, char *seen
,
300 int i
, retval
= 0, exclude
= flags
& DO_MATCH_EXCLUDE
;
311 if (!ps
->recursive
||
312 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
314 return MATCHED_RECURSIVELY
;
316 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
317 return MATCHED_EXACTLY
;
325 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
328 if ((!exclude
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
) ||
329 ( exclude
&& !(ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)))
332 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
335 * Make exclude patterns optional and never report
336 * "pathspec ':(exclude)foo' matches no files"
338 if (seen
&& ps
->items
[i
].magic
& PATHSPEC_EXCLUDE
)
339 seen
[i
] = MATCHED_FNMATCH
;
340 how
= match_pathspec_item(ps
->items
+i
, prefix
, name
,
343 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
344 ps
->max_depth
!= -1 &&
345 how
&& how
!= MATCHED_FNMATCH
) {
346 int len
= ps
->items
[i
].len
;
347 if (name
[len
] == '/')
349 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
350 how
= MATCHED_EXACTLY
;
357 if (seen
&& seen
[i
] < how
)
364 int match_pathspec(const struct pathspec
*ps
,
365 const char *name
, int namelen
,
366 int prefix
, char *seen
, int is_dir
)
368 int positive
, negative
;
369 unsigned flags
= is_dir
? DO_MATCH_DIRECTORY
: 0;
370 positive
= do_match_pathspec(ps
, name
, namelen
,
371 prefix
, seen
, flags
);
372 if (!(ps
->magic
& PATHSPEC_EXCLUDE
) || !positive
)
374 negative
= do_match_pathspec(ps
, name
, namelen
,
376 flags
| DO_MATCH_EXCLUDE
);
377 return negative
? 0 : positive
;
380 int report_path_error(const char *ps_matched
,
381 const struct pathspec
*pathspec
,
385 * Make sure all pathspec matched; otherwise it is an error.
387 struct strbuf sb
= STRBUF_INIT
;
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
);
424 * Return the length of the "simple" part of a path match limiter.
426 int simple_length(const char *match
)
431 unsigned char c
= *match
++;
433 if (c
== '\0' || is_glob_special(c
))
438 int no_wildcard(const char *string
)
440 return string
[simple_length(string
)] == '\0';
443 void parse_exclude_pattern(const char **pattern
,
448 const char *p
= *pattern
;
453 *flags
|= EXC_FLAG_NEGATIVE
;
457 if (len
&& p
[len
- 1] == '/') {
459 *flags
|= EXC_FLAG_MUSTBEDIR
;
461 for (i
= 0; i
< len
; i
++) {
466 *flags
|= EXC_FLAG_NODIR
;
467 *nowildcardlen
= simple_length(p
);
469 * we should have excluded the trailing slash from 'p' too,
470 * but that's one more allocation. Instead just make sure
471 * nowildcardlen does not exceed real patternlen
473 if (*nowildcardlen
> len
)
474 *nowildcardlen
= len
;
475 if (*p
== '*' && no_wildcard(p
+ 1))
476 *flags
|= EXC_FLAG_ENDSWITH
;
481 void add_exclude(const char *string
, const char *base
,
482 int baselen
, struct exclude_list
*el
, int srcpos
)
489 parse_exclude_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
490 if (flags
& EXC_FLAG_MUSTBEDIR
) {
492 x
= xmalloc(sizeof(*x
) + patternlen
+ 1);
494 memcpy(s
, string
, patternlen
);
495 s
[patternlen
] = '\0';
498 x
= xmalloc(sizeof(*x
));
501 x
->patternlen
= patternlen
;
502 x
->nowildcardlen
= nowildcardlen
;
504 x
->baselen
= baselen
;
507 ALLOC_GROW(el
->excludes
, el
->nr
+ 1, el
->alloc
);
508 el
->excludes
[el
->nr
++] = x
;
512 static void *read_skip_worktree_file_from_index(const char *path
, size_t *size
)
516 enum object_type type
;
520 pos
= cache_name_pos(path
, len
);
523 if (!ce_skip_worktree(active_cache
[pos
]))
525 data
= read_sha1_file(active_cache
[pos
]->sha1
, &type
, &sz
);
526 if (!data
|| type
!= OBJ_BLOB
) {
535 * Frees memory within el which was allocated for exclude patterns and
536 * the file buffer. Does not free el itself.
538 void clear_exclude_list(struct exclude_list
*el
)
542 for (i
= 0; i
< el
->nr
; i
++)
543 free(el
->excludes
[i
]);
552 static void trim_trailing_spaces(char *buf
)
554 char *p
, *last_space
= NULL
;
556 for (p
= buf
; *p
; p
++)
575 int add_excludes_from_file_to_list(const char *fname
,
578 struct exclude_list
*el
,
582 int fd
, i
, lineno
= 1;
586 fd
= open(fname
, O_RDONLY
);
587 if (fd
< 0 || fstat(fd
, &st
) < 0) {
589 warn_on_inaccessible(fname
);
593 (buf
= read_skip_worktree_file_from_index(fname
, &size
)) == NULL
)
599 if (buf
[size
-1] != '\n') {
600 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 for (i
= 0; i
< size
; i
++) {
623 if (buf
[i
] == '\n') {
624 if (entry
!= buf
+ i
&& entry
[0] != '#') {
625 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
626 trim_trailing_spaces(entry
);
627 add_exclude(entry
, base
, baselen
, el
, lineno
);
636 struct exclude_list
*add_exclude_list(struct dir_struct
*dir
,
637 int group_type
, const char *src
)
639 struct exclude_list
*el
;
640 struct exclude_list_group
*group
;
642 group
= &dir
->exclude_list_group
[group_type
];
643 ALLOC_GROW(group
->el
, group
->nr
+ 1, group
->alloc
);
644 el
= &group
->el
[group
->nr
++];
645 memset(el
, 0, sizeof(*el
));
651 * Used to set up core.excludesfile and .git/info/exclude lists.
653 void add_excludes_from_file(struct dir_struct
*dir
, const char *fname
)
655 struct exclude_list
*el
;
656 el
= add_exclude_list(dir
, EXC_FILE
, fname
);
657 if (add_excludes_from_file_to_list(fname
, "", 0, el
, 0) < 0)
658 die("cannot use %s as an exclude file", fname
);
661 int match_basename(const char *basename
, int basenamelen
,
662 const char *pattern
, int prefix
, int patternlen
,
665 if (prefix
== patternlen
) {
666 if (patternlen
== basenamelen
&&
667 !strncmp_icase(pattern
, basename
, basenamelen
))
669 } else if (flags
& EXC_FLAG_ENDSWITH
) {
670 /* "*literal" matching against "fooliteral" */
671 if (patternlen
- 1 <= basenamelen
&&
672 !strncmp_icase(pattern
+ 1,
673 basename
+ basenamelen
- (patternlen
- 1),
677 if (fnmatch_icase_mem(pattern
, patternlen
,
678 basename
, basenamelen
,
685 int match_pathname(const char *pathname
, int pathlen
,
686 const char *base
, int baselen
,
687 const char *pattern
, int prefix
, int patternlen
,
694 * match with FNM_PATHNAME; the pattern has base implicitly
697 if (*pattern
== '/') {
704 * baselen does not count the trailing slash. base[] may or
705 * may not end with a trailing slash though.
707 if (pathlen
< baselen
+ 1 ||
708 (baselen
&& pathname
[baselen
] != '/') ||
709 strncmp_icase(pathname
, base
, baselen
))
712 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
713 name
= pathname
+ pathlen
- namelen
;
717 * if the non-wildcard part is longer than the
718 * remaining pathname, surely it cannot match.
720 if (prefix
> namelen
)
723 if (strncmp_icase(pattern
, name
, prefix
))
726 patternlen
-= prefix
;
731 * If the whole pattern did not have a wildcard,
732 * then our prefix match is all we need; we
733 * do not need to call fnmatch at all.
735 if (!patternlen
&& !namelen
)
739 return fnmatch_icase_mem(pattern
, patternlen
,
745 * Scan the given exclude list in reverse to see whether pathname
746 * should be ignored. The first match (i.e. the last on the list), if
747 * any, determines the fate. Returns the exclude_list element which
748 * matched, or NULL for undecided.
750 static struct exclude
*last_exclude_matching_from_list(const char *pathname
,
752 const char *basename
,
754 struct exclude_list
*el
)
759 return NULL
; /* undefined */
761 for (i
= el
->nr
- 1; 0 <= i
; i
--) {
762 struct exclude
*x
= el
->excludes
[i
];
763 const char *exclude
= x
->pattern
;
764 int prefix
= x
->nowildcardlen
;
766 if (x
->flags
& EXC_FLAG_MUSTBEDIR
) {
767 if (*dtype
== DT_UNKNOWN
)
768 *dtype
= get_dtype(NULL
, pathname
, pathlen
);
769 if (*dtype
!= DT_DIR
)
773 if (x
->flags
& EXC_FLAG_NODIR
) {
774 if (match_basename(basename
,
775 pathlen
- (basename
- pathname
),
776 exclude
, prefix
, x
->patternlen
,
782 assert(x
->baselen
== 0 || x
->base
[x
->baselen
- 1] == '/');
783 if (match_pathname(pathname
, pathlen
,
784 x
->base
, x
->baselen
? x
->baselen
- 1 : 0,
785 exclude
, prefix
, x
->patternlen
, x
->flags
))
788 return NULL
; /* undecided */
792 * Scan the list and let the last match determine the fate.
793 * Return 1 for exclude, 0 for include and -1 for undecided.
795 int is_excluded_from_list(const char *pathname
,
796 int pathlen
, const char *basename
, int *dtype
,
797 struct exclude_list
*el
)
799 struct exclude
*exclude
;
800 exclude
= last_exclude_matching_from_list(pathname
, pathlen
, basename
, dtype
, el
);
802 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
803 return -1; /* undecided */
806 static struct exclude
*last_exclude_matching_from_lists(struct dir_struct
*dir
,
807 const char *pathname
, int pathlen
, const char *basename
,
811 struct exclude_list_group
*group
;
812 struct exclude
*exclude
;
813 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
814 group
= &dir
->exclude_list_group
[i
];
815 for (j
= group
->nr
- 1; j
>= 0; j
--) {
816 exclude
= last_exclude_matching_from_list(
817 pathname
, pathlen
, basename
, dtype_p
,
827 * Loads the per-directory exclude list for the substring of base
828 * which has a char length of baselen.
830 static void prep_exclude(struct dir_struct
*dir
, const char *base
, int baselen
)
832 struct exclude_list_group
*group
;
833 struct exclude_list
*el
;
834 struct exclude_stack
*stk
= NULL
;
837 group
= &dir
->exclude_list_group
[EXC_DIRS
];
839 /* Pop the exclude lists from the EXCL_DIRS exclude_list_group
840 * which originate from directories not in the prefix of the
841 * path being checked. */
842 while ((stk
= dir
->exclude_stack
) != NULL
) {
843 if (stk
->baselen
<= baselen
&&
844 !strncmp(dir
->basebuf
, base
, stk
->baselen
))
846 el
= &group
->el
[dir
->exclude_stack
->exclude_ix
];
847 dir
->exclude_stack
= stk
->prev
;
849 free((char *)el
->src
); /* see strdup() below */
850 clear_exclude_list(el
);
855 /* Skip traversing into sub directories if the parent is excluded */
859 /* Read from the parent directories and push them down. */
860 current
= stk
? stk
->baselen
: -1;
861 while (current
< baselen
) {
862 struct exclude_stack
*stk
= xcalloc(1, sizeof(*stk
));
870 cp
= strchr(base
+ current
+ 1, '/');
872 die("oops in prep_exclude");
875 stk
->prev
= dir
->exclude_stack
;
876 stk
->baselen
= cp
- base
;
877 stk
->exclude_ix
= group
->nr
;
878 el
= add_exclude_list(dir
, EXC_DIRS
, NULL
);
879 memcpy(dir
->basebuf
+ current
, base
+ current
,
880 stk
->baselen
- current
);
882 /* Abort if the directory is excluded */
885 dir
->basebuf
[stk
->baselen
- 1] = 0;
886 dir
->exclude
= last_exclude_matching_from_lists(dir
,
887 dir
->basebuf
, stk
->baselen
- 1,
888 dir
->basebuf
+ current
, &dt
);
889 dir
->basebuf
[stk
->baselen
- 1] = '/';
891 dir
->exclude
->flags
& EXC_FLAG_NEGATIVE
)
894 dir
->basebuf
[stk
->baselen
] = 0;
895 dir
->exclude_stack
= stk
;
900 /* Try to read per-directory file unless path is too long */
901 if (dir
->exclude_per_dir
&&
902 stk
->baselen
+ strlen(dir
->exclude_per_dir
) < PATH_MAX
) {
903 strcpy(dir
->basebuf
+ stk
->baselen
,
904 dir
->exclude_per_dir
);
906 * dir->basebuf gets reused by the traversal, but we
907 * need fname to remain unchanged to ensure the src
908 * member of each struct exclude correctly
909 * back-references its source file. Other invocations
910 * of add_exclude_list provide stable strings, so we
911 * strdup() and free() here in the caller.
913 el
->src
= strdup(dir
->basebuf
);
914 add_excludes_from_file_to_list(dir
->basebuf
,
915 dir
->basebuf
, stk
->baselen
, el
, 1);
917 dir
->exclude_stack
= stk
;
918 current
= stk
->baselen
;
920 dir
->basebuf
[baselen
] = '\0';
924 * Loads the exclude lists for the directory containing pathname, then
925 * scans all exclude lists to determine whether pathname is excluded.
926 * Returns the exclude_list element which matched, or NULL for
929 struct exclude
*last_exclude_matching(struct dir_struct
*dir
,
930 const char *pathname
,
933 int pathlen
= strlen(pathname
);
934 const char *basename
= strrchr(pathname
, '/');
935 basename
= (basename
) ? basename
+1 : pathname
;
937 prep_exclude(dir
, pathname
, basename
-pathname
);
942 return last_exclude_matching_from_lists(dir
, pathname
, pathlen
,
947 * Loads the exclude lists for the directory containing pathname, then
948 * scans all exclude lists to determine whether pathname is excluded.
949 * Returns 1 if true, otherwise 0.
951 int is_excluded(struct dir_struct
*dir
, const char *pathname
, int *dtype_p
)
953 struct exclude
*exclude
=
954 last_exclude_matching(dir
, pathname
, dtype_p
);
956 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
960 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
962 struct dir_entry
*ent
;
964 ent
= xmalloc(sizeof(*ent
) + len
+ 1);
966 memcpy(ent
->name
, pathname
, len
);
971 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
, const char *pathname
, int len
)
973 if (cache_file_exists(pathname
, len
, ignore_case
))
976 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->alloc
);
977 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
980 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
, const char *pathname
, int len
)
982 if (!cache_name_is_other(pathname
, len
))
985 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->ignored_alloc
);
986 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
990 index_nonexistent
= 0,
996 * Do not use the alphabetically sorted index to look up
997 * the directory name; instead, use the case insensitive
1000 static enum exist_status
directory_exists_in_index_icase(const char *dirname
, int len
)
1002 const struct cache_entry
*ce
= cache_dir_exists(dirname
, len
);
1003 unsigned char endchar
;
1006 return index_nonexistent
;
1007 endchar
= ce
->name
[len
];
1010 * The cache_entry structure returned will contain this dirname
1011 * and possibly additional path components.
1014 return index_directory
;
1017 * If there are no additional path components, then this cache_entry
1018 * represents a submodule. Submodules, despite being directories,
1019 * are stored in the cache without a closing slash.
1021 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1022 return index_gitdir
;
1024 /* This should never be hit, but it exists just in case. */
1025 return index_nonexistent
;
1029 * The index sorts alphabetically by entry name, which
1030 * means that a gitlink sorts as '\0' at the end, while
1031 * a directory (which is defined not as an entry, but as
1032 * the files it contains) will sort with the '/' at the
1035 static enum exist_status
directory_exists_in_index(const char *dirname
, int len
)
1040 return directory_exists_in_index_icase(dirname
, len
);
1042 pos
= cache_name_pos(dirname
, len
);
1045 while (pos
< active_nr
) {
1046 const struct cache_entry
*ce
= active_cache
[pos
++];
1047 unsigned char endchar
;
1049 if (strncmp(ce
->name
, dirname
, len
))
1051 endchar
= ce
->name
[len
];
1055 return index_directory
;
1056 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
1057 return index_gitdir
;
1059 return index_nonexistent
;
1063 * When we find a directory when traversing the filesystem, we
1064 * have three distinct cases:
1067 * - see it as a directory
1070 * and which one we choose depends on a combination of existing
1071 * git index contents and the flags passed into the directory
1072 * traversal routine.
1074 * Case 1: If we *already* have entries in the index under that
1075 * directory name, we always recurse into the directory to see
1078 * Case 2: If we *already* have that directory name as a gitlink,
1079 * we always continue to see it as a gitlink, regardless of whether
1080 * there is an actual git directory there or not (it might not
1081 * be checked out as a subproject!)
1083 * Case 3: if we didn't have it in the index previously, we
1084 * have a few sub-cases:
1086 * (a) if "show_other_directories" is true, we show it as
1087 * just a directory, unless "hide_empty_directories" is
1088 * also true, in which case we need to check if it contains any
1089 * untracked and / or ignored files.
1090 * (b) if it looks like a git directory, and we don't have
1091 * 'no_gitlinks' set we treat it as a gitlink, and show it
1093 * (c) otherwise, we recurse into it.
1095 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
1096 const char *dirname
, int len
, int exclude
,
1097 const struct path_simplify
*simplify
)
1099 /* The "len-1" is to strip the final '/' */
1100 switch (directory_exists_in_index(dirname
, len
-1)) {
1101 case index_directory
:
1102 return path_recurse
;
1107 case index_nonexistent
:
1108 if (dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)
1110 if (!(dir
->flags
& DIR_NO_GITLINKS
)) {
1111 unsigned char sha1
[20];
1112 if (resolve_gitlink_ref(dirname
, "HEAD", sha1
) == 0)
1113 return path_untracked
;
1115 return path_recurse
;
1118 /* This is the "show_other_directories" case */
1120 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1121 return exclude
? path_excluded
: path_untracked
;
1123 return read_directory_recursive(dir
, dirname
, len
, 1, simplify
);
1127 * This is an inexact early pruning of any recursive directory
1128 * reading - if the path cannot possibly be in the pathspec,
1129 * return true, and we'll skip it early.
1131 static int simplify_away(const char *path
, int pathlen
, const struct path_simplify
*simplify
)
1135 const char *match
= simplify
->path
;
1136 int len
= simplify
->len
;
1142 if (!memcmp(path
, match
, len
))
1152 * This function tells us whether an excluded path matches a
1153 * list of "interesting" pathspecs. That is, whether a path matched
1154 * by any of the pathspecs could possibly be ignored by excluding
1155 * the specified path. This can happen if:
1157 * 1. the path is mentioned explicitly in the pathspec
1159 * 2. the path is a directory prefix of some element in the
1162 static int exclude_matches_pathspec(const char *path
, int len
,
1163 const struct path_simplify
*simplify
)
1166 for (; simplify
->path
; simplify
++) {
1167 if (len
== simplify
->len
1168 && !memcmp(path
, simplify
->path
, len
))
1170 if (len
< simplify
->len
1171 && simplify
->path
[len
] == '/'
1172 && !memcmp(path
, simplify
->path
, len
))
1179 static int get_index_dtype(const char *path
, int len
)
1182 const struct cache_entry
*ce
;
1184 ce
= cache_file_exists(path
, len
, 0);
1186 if (!ce_uptodate(ce
))
1188 if (S_ISGITLINK(ce
->ce_mode
))
1191 * Nobody actually cares about the
1192 * difference between DT_LNK and DT_REG
1197 /* Try to look it up as a directory */
1198 pos
= cache_name_pos(path
, len
);
1202 while (pos
< active_nr
) {
1203 ce
= active_cache
[pos
++];
1204 if (strncmp(ce
->name
, path
, len
))
1206 if (ce
->name
[len
] > '/')
1208 if (ce
->name
[len
] < '/')
1210 if (!ce_uptodate(ce
))
1211 break; /* continue? */
1217 static int get_dtype(struct dirent
*de
, const char *path
, int len
)
1219 int dtype
= de
? DTYPE(de
) : DT_UNKNOWN
;
1222 if (dtype
!= DT_UNKNOWN
)
1224 dtype
= get_index_dtype(path
, len
);
1225 if (dtype
!= DT_UNKNOWN
)
1227 if (lstat(path
, &st
))
1229 if (S_ISREG(st
.st_mode
))
1231 if (S_ISDIR(st
.st_mode
))
1233 if (S_ISLNK(st
.st_mode
))
1238 static enum path_treatment
treat_one_path(struct dir_struct
*dir
,
1239 struct strbuf
*path
,
1240 const struct path_simplify
*simplify
,
1241 int dtype
, struct dirent
*de
)
1244 int has_path_in_index
= !!cache_file_exists(path
->buf
, path
->len
, ignore_case
);
1246 if (dtype
== DT_UNKNOWN
)
1247 dtype
= get_dtype(de
, path
->buf
, path
->len
);
1249 /* Always exclude indexed files */
1250 if (dtype
!= DT_DIR
&& has_path_in_index
)
1254 * When we are looking at a directory P in the working tree,
1255 * there are three cases:
1257 * (1) P exists in the index. Everything inside the directory P in
1258 * the working tree needs to go when P is checked out from the
1261 * (2) P does not exist in the index, but there is P/Q in the index.
1262 * We know P will stay a directory when we check out the contents
1263 * of the index, but we do not know yet if there is a directory
1264 * P/Q in the working tree to be killed, so we need to recurse.
1266 * (3) P does not exist in the index, and there is no P/Q in the index
1267 * to require P to be a directory, either. Only in this case, we
1268 * know that everything inside P will not be killed without
1271 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
1272 (dtype
== DT_DIR
) &&
1273 !has_path_in_index
&&
1274 (directory_exists_in_index(path
->buf
, path
->len
) == index_nonexistent
))
1277 exclude
= is_excluded(dir
, path
->buf
, &dtype
);
1280 * Excluded? If we don't explicitly want to show
1281 * ignored files, ignore it
1283 if (exclude
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
1284 return path_excluded
;
1290 strbuf_addch(path
, '/');
1291 return treat_directory(dir
, path
->buf
, path
->len
, exclude
,
1295 return exclude
? path_excluded
: path_untracked
;
1299 static enum path_treatment
treat_path(struct dir_struct
*dir
,
1301 struct strbuf
*path
,
1303 const struct path_simplify
*simplify
)
1307 if (is_dot_or_dotdot(de
->d_name
) || !strcmp(de
->d_name
, ".git"))
1309 strbuf_setlen(path
, baselen
);
1310 strbuf_addstr(path
, de
->d_name
);
1311 if (simplify_away(path
->buf
, path
->len
, simplify
))
1315 return treat_one_path(dir
, path
, simplify
, dtype
, de
);
1319 * Read a directory tree. We currently ignore anything but
1320 * directories, regular files and symlinks. That's because git
1321 * doesn't handle them at all yet. Maybe that will change some
1324 * Also, we ignore the name ".git" (even if it is not a directory).
1325 * That likely will not change.
1327 * Returns the most significant path_treatment value encountered in the scan.
1329 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
1330 const char *base
, int baselen
,
1332 const struct path_simplify
*simplify
)
1335 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
1337 struct strbuf path
= STRBUF_INIT
;
1339 strbuf_add(&path
, base
, baselen
);
1341 fdir
= opendir(path
.len
? path
.buf
: ".");
1345 while ((de
= readdir(fdir
)) != NULL
) {
1346 /* check how the file or directory should be treated */
1347 state
= treat_path(dir
, de
, &path
, baselen
, simplify
);
1348 if (state
> dir_state
)
1351 /* recurse into subdir if instructed by treat_path */
1352 if (state
== path_recurse
) {
1353 subdir_state
= read_directory_recursive(dir
, path
.buf
,
1354 path
.len
, check_only
, simplify
);
1355 if (subdir_state
> dir_state
)
1356 dir_state
= subdir_state
;
1360 /* abort early if maximum state has been reached */
1361 if (dir_state
== path_untracked
)
1363 /* skip the dir_add_* part */
1367 /* add the path to the appropriate result list */
1370 if (dir
->flags
& DIR_SHOW_IGNORED
)
1371 dir_add_name(dir
, path
.buf
, path
.len
);
1372 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
1373 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
1374 exclude_matches_pathspec(path
.buf
, path
.len
,
1376 dir_add_ignored(dir
, path
.buf
, path
.len
);
1379 case path_untracked
:
1380 if (!(dir
->flags
& DIR_SHOW_IGNORED
))
1381 dir_add_name(dir
, path
.buf
, path
.len
);
1390 strbuf_release(&path
);
1395 static int cmp_name(const void *p1
, const void *p2
)
1397 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
1398 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
1400 return cache_name_compare(e1
->name
, e1
->len
,
1404 static struct path_simplify
*create_simplify(const char **pathspec
)
1407 struct path_simplify
*simplify
= NULL
;
1412 for (nr
= 0 ; ; nr
++) {
1414 ALLOC_GROW(simplify
, nr
+ 1, alloc
);
1415 match
= *pathspec
++;
1418 simplify
[nr
].path
= match
;
1419 simplify
[nr
].len
= simple_length(match
);
1421 simplify
[nr
].path
= NULL
;
1422 simplify
[nr
].len
= 0;
1426 static void free_simplify(struct path_simplify
*simplify
)
1431 static int treat_leading_path(struct dir_struct
*dir
,
1432 const char *path
, int len
,
1433 const struct path_simplify
*simplify
)
1435 struct strbuf sb
= STRBUF_INIT
;
1436 int baselen
, rc
= 0;
1438 int old_flags
= dir
->flags
;
1440 while (len
&& path
[len
- 1] == '/')
1445 dir
->flags
&= ~DIR_SHOW_OTHER_DIRECTORIES
;
1447 cp
= path
+ baselen
+ !!baselen
;
1448 cp
= memchr(cp
, '/', path
+ len
- cp
);
1452 baselen
= cp
- path
;
1453 strbuf_setlen(&sb
, 0);
1454 strbuf_add(&sb
, path
, baselen
);
1455 if (!is_directory(sb
.buf
))
1457 if (simplify_away(sb
.buf
, sb
.len
, simplify
))
1459 if (treat_one_path(dir
, &sb
, simplify
,
1460 DT_DIR
, NULL
) == path_none
)
1461 break; /* do not recurse into it */
1462 if (len
<= baselen
) {
1464 break; /* finished checking */
1467 strbuf_release(&sb
);
1468 dir
->flags
= old_flags
;
1472 int read_directory(struct dir_struct
*dir
, const char *path
, int len
, const struct pathspec
*pathspec
)
1474 struct path_simplify
*simplify
;
1477 * Check out create_simplify()
1480 GUARD_PATHSPEC(pathspec
,
1488 if (has_symlink_leading_path(path
, len
))
1492 * exclude patterns are treated like positive ones in
1493 * create_simplify. Usually exclude patterns should be a
1494 * subset of positive ones, which has no impacts on
1495 * create_simplify().
1497 simplify
= create_simplify(pathspec
? pathspec
->_raw
: NULL
);
1498 if (!len
|| treat_leading_path(dir
, path
, len
, simplify
))
1499 read_directory_recursive(dir
, path
, len
, 0, simplify
);
1500 free_simplify(simplify
);
1501 qsort(dir
->entries
, dir
->nr
, sizeof(struct dir_entry
*), cmp_name
);
1502 qsort(dir
->ignored
, dir
->ignored_nr
, sizeof(struct dir_entry
*), cmp_name
);
1506 int file_exists(const char *f
)
1509 return lstat(f
, &sb
) == 0;
1513 * Given two normalized paths (a trailing slash is ok), if subdir is
1514 * outside dir, return -1. Otherwise return the offset in subdir that
1515 * can be used as relative path to dir.
1517 int dir_inside_of(const char *subdir
, const char *dir
)
1521 assert(dir
&& subdir
&& *dir
&& *subdir
);
1523 while (*dir
&& *subdir
&& *dir
== *subdir
) {
1529 /* hel[p]/me vs hel[l]/yeah */
1530 if (*dir
&& *subdir
)
1534 return !*dir
? offset
: -1; /* same dir */
1536 /* foo/[b]ar vs foo/[] */
1537 if (is_dir_sep(dir
[-1]))
1538 return is_dir_sep(subdir
[-1]) ? offset
: -1;
1540 /* foo[/]bar vs foo[] */
1541 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
1544 int is_inside_dir(const char *dir
)
1549 if (!getcwd(cwd
, sizeof(cwd
)))
1550 die_errno("can't find the current directory");
1551 return dir_inside_of(cwd
, dir
) >= 0;
1554 int is_empty_dir(const char *path
)
1556 DIR *dir
= opendir(path
);
1563 while ((e
= readdir(dir
)) != NULL
)
1564 if (!is_dot_or_dotdot(e
->d_name
)) {
1573 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
1577 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
1578 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
1579 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
1580 unsigned char submodule_head
[20];
1582 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
1583 !resolve_gitlink_ref(path
->buf
, "HEAD", submodule_head
)) {
1584 /* Do not descend and nuke a nested git work tree. */
1590 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
1591 dir
= opendir(path
->buf
);
1593 if (errno
== ENOENT
)
1594 return keep_toplevel
? -1 : 0;
1595 else if (errno
== EACCES
&& !keep_toplevel
)
1597 * An empty dir could be removable even if it
1600 return rmdir(path
->buf
);
1604 if (path
->buf
[original_len
- 1] != '/')
1605 strbuf_addch(path
, '/');
1608 while ((e
= readdir(dir
)) != NULL
) {
1610 if (is_dot_or_dotdot(e
->d_name
))
1613 strbuf_setlen(path
, len
);
1614 strbuf_addstr(path
, e
->d_name
);
1615 if (lstat(path
->buf
, &st
)) {
1616 if (errno
== ENOENT
)
1618 * file disappeared, which is what we
1623 } else if (S_ISDIR(st
.st_mode
)) {
1624 if (!remove_dir_recurse(path
, flag
, &kept_down
))
1625 continue; /* happy */
1626 } else if (!only_empty
&&
1627 (!unlink(path
->buf
) || errno
== ENOENT
)) {
1628 continue; /* happy, too */
1631 /* path too long, stat fails, or non-directory still exists */
1637 strbuf_setlen(path
, original_len
);
1638 if (!ret
&& !keep_toplevel
&& !kept_down
)
1639 ret
= (!rmdir(path
->buf
) || errno
== ENOENT
) ? 0 : -1;
1642 * report the uplevel that it is not an error that we
1643 * did not rmdir() our directory.
1649 int remove_dir_recursively(struct strbuf
*path
, int flag
)
1651 return remove_dir_recurse(path
, flag
, NULL
);
1654 void setup_standard_excludes(struct dir_struct
*dir
)
1659 dir
->exclude_per_dir
= ".gitignore";
1660 path
= git_path("info/exclude");
1661 if (!excludes_file
) {
1662 home_config_paths(NULL
, &xdg_path
, "ignore");
1663 excludes_file
= xdg_path
;
1665 if (!access_or_warn(path
, R_OK
, 0))
1666 add_excludes_from_file(dir
, path
);
1667 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
1668 add_excludes_from_file(dir
, excludes_file
);
1671 int remove_path(const char *name
)
1675 if (unlink(name
) && errno
!= ENOENT
&& errno
!= ENOTDIR
)
1678 slash
= strrchr(name
, '/');
1680 char *dirs
= xstrdup(name
);
1681 slash
= dirs
+ (slash
- name
);
1684 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
1691 * Frees memory within dir which was allocated for exclude lists and
1692 * the exclude_stack. Does not free dir itself.
1694 void clear_directory(struct dir_struct
*dir
)
1697 struct exclude_list_group
*group
;
1698 struct exclude_list
*el
;
1699 struct exclude_stack
*stk
;
1701 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1702 group
= &dir
->exclude_list_group
[i
];
1703 for (j
= 0; j
< group
->nr
; j
++) {
1706 free((char *)el
->src
);
1707 clear_exclude_list(el
);
1712 stk
= dir
->exclude_stack
;
1714 struct exclude_stack
*prev
= stk
->prev
;