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 fnmatch(pattern
, string
, flags
| (ignore_case
? FNM_CASEFOLD
: 0));
55 inline int git_fnmatch(const struct pathspec_item
*item
,
56 const char *pattern
, const char *string
,
60 if (ps_strncmp(item
, pattern
, string
, prefix
))
65 if (item
->flags
& PATHSPEC_ONESTAR
) {
66 int pattern_len
= strlen(++pattern
);
67 int string_len
= strlen(string
);
68 return string_len
< pattern_len
||
69 ps_strcmp(item
, pattern
,
70 string
+ string_len
- pattern_len
);
72 if (item
->magic
& PATHSPEC_GLOB
)
73 return wildmatch(pattern
, string
,
75 (item
->magic
& PATHSPEC_ICASE
? WM_CASEFOLD
: 0),
78 /* wildmatch has not learned no FNM_PATHNAME mode yet */
79 return fnmatch(pattern
, string
,
80 item
->magic
& PATHSPEC_ICASE
? FNM_CASEFOLD
: 0);
83 static int fnmatch_icase_mem(const char *pattern
, int patternlen
,
84 const char *string
, int stringlen
,
88 struct strbuf pat_buf
= STRBUF_INIT
;
89 struct strbuf str_buf
= STRBUF_INIT
;
90 const char *use_pat
= pattern
;
91 const char *use_str
= string
;
93 if (pattern
[patternlen
]) {
94 strbuf_add(&pat_buf
, pattern
, patternlen
);
95 use_pat
= pat_buf
.buf
;
97 if (string
[stringlen
]) {
98 strbuf_add(&str_buf
, string
, stringlen
);
99 use_str
= str_buf
.buf
;
103 flags
|= WM_CASEFOLD
;
104 match_status
= wildmatch(use_pat
, use_str
, flags
, NULL
);
106 strbuf_release(&pat_buf
);
107 strbuf_release(&str_buf
);
112 static size_t common_prefix_len(const struct pathspec
*pathspec
)
118 * ":(icase)path" is treated as a pathspec full of
119 * wildcard. In other words, only prefix is considered common
120 * prefix. If the pathspec is abc/foo abc/bar, running in
121 * subdir xyz, the common prefix is still xyz, not xuz/abc as
124 GUARD_PATHSPEC(pathspec
,
131 for (n
= 0; n
< pathspec
->nr
; n
++) {
132 size_t i
= 0, len
= 0, item_len
;
133 if (pathspec
->items
[n
].magic
& PATHSPEC_ICASE
)
134 item_len
= pathspec
->items
[n
].prefix
;
136 item_len
= pathspec
->items
[n
].nowildcard_len
;
137 while (i
< item_len
&& (n
== 0 || i
< max
)) {
138 char c
= pathspec
->items
[n
].match
[i
];
139 if (c
!= pathspec
->items
[0].match
[i
])
145 if (n
== 0 || len
< max
) {
155 * Returns a copy of the longest leading path common among all
158 char *common_prefix(const struct pathspec
*pathspec
)
160 unsigned long len
= common_prefix_len(pathspec
);
162 return len
? xmemdupz(pathspec
->items
[0].match
, len
) : NULL
;
165 int fill_directory(struct dir_struct
*dir
, const struct pathspec
*pathspec
)
170 * Calculate common prefix for the pathspec, and
171 * use that to optimize the directory walk
173 len
= common_prefix_len(pathspec
);
175 /* Read the directory and prune it */
176 read_directory(dir
, pathspec
->nr
? pathspec
->_raw
[0] : "", len
, pathspec
);
180 int within_depth(const char *name
, int namelen
,
181 int depth
, int max_depth
)
183 const char *cp
= name
, *cpe
= name
+ namelen
;
189 if (depth
> max_depth
)
196 * Does 'match' match the given name?
197 * A match is found if
199 * (1) the 'match' string is leading directory of 'name', or
200 * (2) the 'match' string is a wildcard and matches 'name', or
201 * (3) the 'match' string is exactly the same as 'name'.
203 * and the return value tells which case it was.
205 * It returns 0 when there is no match.
207 static int match_pathspec_item(const struct pathspec_item
*item
, int prefix
,
208 const char *name
, int namelen
)
210 /* name/namelen has prefix cut off by caller */
211 const char *match
= item
->match
+ prefix
;
212 int matchlen
= item
->len
- prefix
;
215 * The normal call pattern is:
216 * 1. prefix = common_prefix_len(ps);
217 * 2. prune something, or fill_directory
218 * 3. match_pathspec_depth()
220 * 'prefix' at #1 may be shorter than the command's prefix and
221 * it's ok for #2 to match extra files. Those extras will be
224 * Suppose the pathspec is 'foo' and '../bar' running from
225 * subdir 'xyz'. The common prefix at #1 will be empty, thanks
226 * to "../". We may have xyz/foo _and_ XYZ/foo after #2. The
227 * user does not want XYZ/foo, only the "foo" part should be
228 * case-insensitive. We need to filter out XYZ/foo here. In
229 * other words, we do not trust the caller on comparing the
230 * prefix part when :(icase) is involved. We do exact
231 * comparison ourselves.
233 * Normally the caller (common_prefix_len() in fact) does
234 * _exact_ matching on name[-prefix+1..-1] and we do not need
235 * to check that part. Be defensive and check it anyway, in
236 * case common_prefix_len is changed, or a new caller is
237 * introduced that does not use common_prefix_len.
239 * If the penalty turns out too high when prefix is really
240 * long, maybe change it to
241 * strncmp(match, name, item->prefix - prefix)
243 if (item
->prefix
&& (item
->magic
& PATHSPEC_ICASE
) &&
244 strncmp(item
->match
, name
- prefix
, item
->prefix
))
247 /* If the match was just the prefix, we matched */
249 return MATCHED_RECURSIVELY
;
251 if (matchlen
<= namelen
&& !ps_strncmp(item
, match
, name
, matchlen
)) {
252 if (matchlen
== namelen
)
253 return MATCHED_EXACTLY
;
255 if (match
[matchlen
-1] == '/' || name
[matchlen
] == '/')
256 return MATCHED_RECURSIVELY
;
259 if (item
->nowildcard_len
< item
->len
&&
260 !git_fnmatch(item
, match
, name
,
261 item
->nowildcard_len
- prefix
))
262 return MATCHED_FNMATCH
;
268 * Given a name and a list of pathspecs, returns the nature of the
269 * closest (i.e. most specific) match of the name to any of the
272 * The caller typically calls this multiple times with the same
273 * pathspec and seen[] array but with different name/namelen
274 * (e.g. entries from the index) and is interested in seeing if and
275 * how each pathspec matches all the names it calls this function
276 * with. A mark is left in the seen[] array for each pathspec element
277 * indicating the closest type of match that element achieved, so if
278 * seen[n] remains zero after multiple invocations, that means the nth
279 * pathspec did not match any names, which could indicate that the
280 * user mistyped the nth pathspec.
282 int match_pathspec_depth(const struct pathspec
*ps
,
283 const char *name
, int namelen
,
284 int prefix
, char *seen
)
296 if (!ps
->recursive
||
297 !(ps
->magic
& PATHSPEC_MAXDEPTH
) ||
299 return MATCHED_RECURSIVELY
;
301 if (within_depth(name
, namelen
, 0, ps
->max_depth
))
302 return MATCHED_EXACTLY
;
310 for (i
= ps
->nr
- 1; i
>= 0; i
--) {
312 if (seen
&& seen
[i
] == MATCHED_EXACTLY
)
314 how
= match_pathspec_item(ps
->items
+i
, prefix
, name
, namelen
);
316 (ps
->magic
& PATHSPEC_MAXDEPTH
) &&
317 ps
->max_depth
!= -1 &&
318 how
&& how
!= MATCHED_FNMATCH
) {
319 int len
= ps
->items
[i
].len
;
320 if (name
[len
] == '/')
322 if (within_depth(name
+len
, namelen
-len
, 0, ps
->max_depth
))
323 how
= MATCHED_EXACTLY
;
330 if (seen
&& seen
[i
] < how
)
338 * Return the length of the "simple" part of a path match limiter.
340 int simple_length(const char *match
)
345 unsigned char c
= *match
++;
347 if (c
== '\0' || is_glob_special(c
))
352 int no_wildcard(const char *string
)
354 return string
[simple_length(string
)] == '\0';
357 void parse_exclude_pattern(const char **pattern
,
362 const char *p
= *pattern
;
367 *flags
|= EXC_FLAG_NEGATIVE
;
371 if (len
&& p
[len
- 1] == '/') {
373 *flags
|= EXC_FLAG_MUSTBEDIR
;
375 for (i
= 0; i
< len
; i
++) {
380 *flags
|= EXC_FLAG_NODIR
;
381 *nowildcardlen
= simple_length(p
);
383 * we should have excluded the trailing slash from 'p' too,
384 * but that's one more allocation. Instead just make sure
385 * nowildcardlen does not exceed real patternlen
387 if (*nowildcardlen
> len
)
388 *nowildcardlen
= len
;
389 if (*p
== '*' && no_wildcard(p
+ 1))
390 *flags
|= EXC_FLAG_ENDSWITH
;
395 void add_exclude(const char *string
, const char *base
,
396 int baselen
, struct exclude_list
*el
, int srcpos
)
403 parse_exclude_pattern(&string
, &patternlen
, &flags
, &nowildcardlen
);
404 if (flags
& EXC_FLAG_MUSTBEDIR
) {
406 x
= xmalloc(sizeof(*x
) + patternlen
+ 1);
408 memcpy(s
, string
, patternlen
);
409 s
[patternlen
] = '\0';
412 x
= xmalloc(sizeof(*x
));
415 x
->patternlen
= patternlen
;
416 x
->nowildcardlen
= nowildcardlen
;
418 x
->baselen
= baselen
;
421 ALLOC_GROW(el
->excludes
, el
->nr
+ 1, el
->alloc
);
422 el
->excludes
[el
->nr
++] = x
;
426 static void *read_skip_worktree_file_from_index(const char *path
, size_t *size
)
430 enum object_type type
;
434 pos
= cache_name_pos(path
, len
);
437 if (!ce_skip_worktree(active_cache
[pos
]))
439 data
= read_sha1_file(active_cache
[pos
]->sha1
, &type
, &sz
);
440 if (!data
|| type
!= OBJ_BLOB
) {
449 * Frees memory within el which was allocated for exclude patterns and
450 * the file buffer. Does not free el itself.
452 void clear_exclude_list(struct exclude_list
*el
)
456 for (i
= 0; i
< el
->nr
; i
++)
457 free(el
->excludes
[i
]);
466 int add_excludes_from_file_to_list(const char *fname
,
469 struct exclude_list
*el
,
473 int fd
, i
, lineno
= 1;
477 fd
= open(fname
, O_RDONLY
);
478 if (fd
< 0 || fstat(fd
, &st
) < 0) {
480 warn_on_inaccessible(fname
);
484 (buf
= read_skip_worktree_file_from_index(fname
, &size
)) == NULL
)
490 if (buf
[size
-1] != '\n') {
491 buf
= xrealloc(buf
, size
+1);
496 size
= xsize_t(st
.st_size
);
501 buf
= xmalloc(size
+1);
502 if (read_in_full(fd
, buf
, size
) != size
) {
513 for (i
= 0; i
< size
; i
++) {
514 if (buf
[i
] == '\n') {
515 if (entry
!= buf
+ i
&& entry
[0] != '#') {
516 buf
[i
- (i
&& buf
[i
-1] == '\r')] = 0;
517 add_exclude(entry
, base
, baselen
, el
, lineno
);
526 struct exclude_list
*add_exclude_list(struct dir_struct
*dir
,
527 int group_type
, const char *src
)
529 struct exclude_list
*el
;
530 struct exclude_list_group
*group
;
532 group
= &dir
->exclude_list_group
[group_type
];
533 ALLOC_GROW(group
->el
, group
->nr
+ 1, group
->alloc
);
534 el
= &group
->el
[group
->nr
++];
535 memset(el
, 0, sizeof(*el
));
541 * Used to set up core.excludesfile and .git/info/exclude lists.
543 void add_excludes_from_file(struct dir_struct
*dir
, const char *fname
)
545 struct exclude_list
*el
;
546 el
= add_exclude_list(dir
, EXC_FILE
, fname
);
547 if (add_excludes_from_file_to_list(fname
, "", 0, el
, 0) < 0)
548 die("cannot use %s as an exclude file", fname
);
551 int match_basename(const char *basename
, int basenamelen
,
552 const char *pattern
, int prefix
, int patternlen
,
555 if (prefix
== patternlen
) {
556 if (patternlen
== basenamelen
&&
557 !strncmp_icase(pattern
, basename
, basenamelen
))
559 } else if (flags
& EXC_FLAG_ENDSWITH
) {
560 /* "*literal" matching against "fooliteral" */
561 if (patternlen
- 1 <= basenamelen
&&
562 !strncmp_icase(pattern
+ 1,
563 basename
+ basenamelen
- (patternlen
- 1),
567 if (fnmatch_icase_mem(pattern
, patternlen
,
568 basename
, basenamelen
,
575 int match_pathname(const char *pathname
, int pathlen
,
576 const char *base
, int baselen
,
577 const char *pattern
, int prefix
, int patternlen
,
584 * match with FNM_PATHNAME; the pattern has base implicitly
587 if (*pattern
== '/') {
594 * baselen does not count the trailing slash. base[] may or
595 * may not end with a trailing slash though.
597 if (pathlen
< baselen
+ 1 ||
598 (baselen
&& pathname
[baselen
] != '/') ||
599 strncmp_icase(pathname
, base
, baselen
))
602 namelen
= baselen
? pathlen
- baselen
- 1 : pathlen
;
603 name
= pathname
+ pathlen
- namelen
;
607 * if the non-wildcard part is longer than the
608 * remaining pathname, surely it cannot match.
610 if (prefix
> namelen
)
613 if (strncmp_icase(pattern
, name
, prefix
))
616 patternlen
-= prefix
;
621 * If the whole pattern did not have a wildcard,
622 * then our prefix match is all we need; we
623 * do not need to call fnmatch at all.
625 if (!patternlen
&& !namelen
)
629 return fnmatch_icase_mem(pattern
, patternlen
,
635 * Scan the given exclude list in reverse to see whether pathname
636 * should be ignored. The first match (i.e. the last on the list), if
637 * any, determines the fate. Returns the exclude_list element which
638 * matched, or NULL for undecided.
640 static struct exclude
*last_exclude_matching_from_list(const char *pathname
,
642 const char *basename
,
644 struct exclude_list
*el
)
649 return NULL
; /* undefined */
651 for (i
= el
->nr
- 1; 0 <= i
; i
--) {
652 struct exclude
*x
= el
->excludes
[i
];
653 const char *exclude
= x
->pattern
;
654 int prefix
= x
->nowildcardlen
;
656 if (x
->flags
& EXC_FLAG_MUSTBEDIR
) {
657 if (*dtype
== DT_UNKNOWN
)
658 *dtype
= get_dtype(NULL
, pathname
, pathlen
);
659 if (*dtype
!= DT_DIR
)
663 if (x
->flags
& EXC_FLAG_NODIR
) {
664 if (match_basename(basename
,
665 pathlen
- (basename
- pathname
),
666 exclude
, prefix
, x
->patternlen
,
672 assert(x
->baselen
== 0 || x
->base
[x
->baselen
- 1] == '/');
673 if (match_pathname(pathname
, pathlen
,
674 x
->base
, x
->baselen
? x
->baselen
- 1 : 0,
675 exclude
, prefix
, x
->patternlen
, x
->flags
))
678 return NULL
; /* undecided */
682 * Scan the list and let the last match determine the fate.
683 * Return 1 for exclude, 0 for include and -1 for undecided.
685 int is_excluded_from_list(const char *pathname
,
686 int pathlen
, const char *basename
, int *dtype
,
687 struct exclude_list
*el
)
689 struct exclude
*exclude
;
690 exclude
= last_exclude_matching_from_list(pathname
, pathlen
, basename
, dtype
, el
);
692 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
693 return -1; /* undecided */
696 static struct exclude
*last_exclude_matching_from_lists(struct dir_struct
*dir
,
697 const char *pathname
, int pathlen
, const char *basename
,
701 struct exclude_list_group
*group
;
702 struct exclude
*exclude
;
703 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
704 group
= &dir
->exclude_list_group
[i
];
705 for (j
= group
->nr
- 1; j
>= 0; j
--) {
706 exclude
= last_exclude_matching_from_list(
707 pathname
, pathlen
, basename
, dtype_p
,
717 * Loads the per-directory exclude list for the substring of base
718 * which has a char length of baselen.
720 static void prep_exclude(struct dir_struct
*dir
, const char *base
, int baselen
)
722 struct exclude_list_group
*group
;
723 struct exclude_list
*el
;
724 struct exclude_stack
*stk
= NULL
;
727 group
= &dir
->exclude_list_group
[EXC_DIRS
];
729 /* Pop the exclude lists from the EXCL_DIRS exclude_list_group
730 * which originate from directories not in the prefix of the
731 * path being checked. */
732 while ((stk
= dir
->exclude_stack
) != NULL
) {
733 if (stk
->baselen
<= baselen
&&
734 !strncmp(dir
->basebuf
, base
, stk
->baselen
))
736 el
= &group
->el
[dir
->exclude_stack
->exclude_ix
];
737 dir
->exclude_stack
= stk
->prev
;
739 free((char *)el
->src
); /* see strdup() below */
740 clear_exclude_list(el
);
745 /* Skip traversing into sub directories if the parent is excluded */
749 /* Read from the parent directories and push them down. */
750 current
= stk
? stk
->baselen
: -1;
751 while (current
< baselen
) {
752 struct exclude_stack
*stk
= xcalloc(1, sizeof(*stk
));
760 cp
= strchr(base
+ current
+ 1, '/');
762 die("oops in prep_exclude");
765 stk
->prev
= dir
->exclude_stack
;
766 stk
->baselen
= cp
- base
;
767 stk
->exclude_ix
= group
->nr
;
768 el
= add_exclude_list(dir
, EXC_DIRS
, NULL
);
769 memcpy(dir
->basebuf
+ current
, base
+ current
,
770 stk
->baselen
- current
);
772 /* Abort if the directory is excluded */
775 dir
->basebuf
[stk
->baselen
- 1] = 0;
776 dir
->exclude
= last_exclude_matching_from_lists(dir
,
777 dir
->basebuf
, stk
->baselen
- 1,
778 dir
->basebuf
+ current
, &dt
);
779 dir
->basebuf
[stk
->baselen
- 1] = '/';
781 dir
->exclude
->flags
& EXC_FLAG_NEGATIVE
)
784 dir
->basebuf
[stk
->baselen
] = 0;
785 dir
->exclude_stack
= stk
;
790 /* Try to read per-directory file unless path is too long */
791 if (dir
->exclude_per_dir
&&
792 stk
->baselen
+ strlen(dir
->exclude_per_dir
) < PATH_MAX
) {
793 strcpy(dir
->basebuf
+ stk
->baselen
,
794 dir
->exclude_per_dir
);
796 * dir->basebuf gets reused by the traversal, but we
797 * need fname to remain unchanged to ensure the src
798 * member of each struct exclude correctly
799 * back-references its source file. Other invocations
800 * of add_exclude_list provide stable strings, so we
801 * strdup() and free() here in the caller.
803 el
->src
= strdup(dir
->basebuf
);
804 add_excludes_from_file_to_list(dir
->basebuf
,
805 dir
->basebuf
, stk
->baselen
, el
, 1);
807 dir
->exclude_stack
= stk
;
808 current
= stk
->baselen
;
810 dir
->basebuf
[baselen
] = '\0';
814 * Loads the exclude lists for the directory containing pathname, then
815 * scans all exclude lists to determine whether pathname is excluded.
816 * Returns the exclude_list element which matched, or NULL for
819 struct exclude
*last_exclude_matching(struct dir_struct
*dir
,
820 const char *pathname
,
823 int pathlen
= strlen(pathname
);
824 const char *basename
= strrchr(pathname
, '/');
825 basename
= (basename
) ? basename
+1 : pathname
;
827 prep_exclude(dir
, pathname
, basename
-pathname
);
832 return last_exclude_matching_from_lists(dir
, pathname
, pathlen
,
837 * Loads the exclude lists for the directory containing pathname, then
838 * scans all exclude lists to determine whether pathname is excluded.
839 * Returns 1 if true, otherwise 0.
841 int is_excluded(struct dir_struct
*dir
, const char *pathname
, int *dtype_p
)
843 struct exclude
*exclude
=
844 last_exclude_matching(dir
, pathname
, dtype_p
);
846 return exclude
->flags
& EXC_FLAG_NEGATIVE
? 0 : 1;
850 static struct dir_entry
*dir_entry_new(const char *pathname
, int len
)
852 struct dir_entry
*ent
;
854 ent
= xmalloc(sizeof(*ent
) + len
+ 1);
856 memcpy(ent
->name
, pathname
, len
);
861 static struct dir_entry
*dir_add_name(struct dir_struct
*dir
, const char *pathname
, int len
)
863 if (cache_name_exists(pathname
, len
, ignore_case
))
866 ALLOC_GROW(dir
->entries
, dir
->nr
+1, dir
->alloc
);
867 return dir
->entries
[dir
->nr
++] = dir_entry_new(pathname
, len
);
870 struct dir_entry
*dir_add_ignored(struct dir_struct
*dir
, const char *pathname
, int len
)
872 if (!cache_name_is_other(pathname
, len
))
875 ALLOC_GROW(dir
->ignored
, dir
->ignored_nr
+1, dir
->ignored_alloc
);
876 return dir
->ignored
[dir
->ignored_nr
++] = dir_entry_new(pathname
, len
);
880 index_nonexistent
= 0,
886 * Do not use the alphabetically sorted index to look up
887 * the directory name; instead, use the case insensitive
890 static enum exist_status
directory_exists_in_index_icase(const char *dirname
, int len
)
892 const struct cache_entry
*ce
= cache_name_exists(dirname
, len
+ 1, ignore_case
);
893 unsigned char endchar
;
896 return index_nonexistent
;
897 endchar
= ce
->name
[len
];
900 * The cache_entry structure returned will contain this dirname
901 * and possibly additional path components.
904 return index_directory
;
907 * If there are no additional path components, then this cache_entry
908 * represents a submodule. Submodules, despite being directories,
909 * are stored in the cache without a closing slash.
911 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
914 /* This should never be hit, but it exists just in case. */
915 return index_nonexistent
;
919 * The index sorts alphabetically by entry name, which
920 * means that a gitlink sorts as '\0' at the end, while
921 * a directory (which is defined not as an entry, but as
922 * the files it contains) will sort with the '/' at the
925 static enum exist_status
directory_exists_in_index(const char *dirname
, int len
)
930 return directory_exists_in_index_icase(dirname
, len
);
932 pos
= cache_name_pos(dirname
, len
);
935 while (pos
< active_nr
) {
936 const struct cache_entry
*ce
= active_cache
[pos
++];
937 unsigned char endchar
;
939 if (strncmp(ce
->name
, dirname
, len
))
941 endchar
= ce
->name
[len
];
945 return index_directory
;
946 if (!endchar
&& S_ISGITLINK(ce
->ce_mode
))
949 return index_nonexistent
;
953 * When we find a directory when traversing the filesystem, we
954 * have three distinct cases:
957 * - see it as a directory
960 * and which one we choose depends on a combination of existing
961 * git index contents and the flags passed into the directory
964 * Case 1: If we *already* have entries in the index under that
965 * directory name, we always recurse into the directory to see
968 * Case 2: If we *already* have that directory name as a gitlink,
969 * we always continue to see it as a gitlink, regardless of whether
970 * there is an actual git directory there or not (it might not
971 * be checked out as a subproject!)
973 * Case 3: if we didn't have it in the index previously, we
974 * have a few sub-cases:
976 * (a) if "show_other_directories" is true, we show it as
977 * just a directory, unless "hide_empty_directories" is
978 * also true, in which case we need to check if it contains any
979 * untracked and / or ignored files.
980 * (b) if it looks like a git directory, and we don't have
981 * 'no_gitlinks' set we treat it as a gitlink, and show it
983 * (c) otherwise, we recurse into it.
985 static enum path_treatment
treat_directory(struct dir_struct
*dir
,
986 const char *dirname
, int len
, int exclude
,
987 const struct path_simplify
*simplify
)
989 /* The "len-1" is to strip the final '/' */
990 switch (directory_exists_in_index(dirname
, len
-1)) {
991 case index_directory
:
997 case index_nonexistent
:
998 if (dir
->flags
& DIR_SHOW_OTHER_DIRECTORIES
)
1000 if (!(dir
->flags
& DIR_NO_GITLINKS
)) {
1001 unsigned char sha1
[20];
1002 if (resolve_gitlink_ref(dirname
, "HEAD", sha1
) == 0)
1003 return path_untracked
;
1005 return path_recurse
;
1008 /* This is the "show_other_directories" case */
1010 if (!(dir
->flags
& DIR_HIDE_EMPTY_DIRECTORIES
))
1011 return exclude
? path_excluded
: path_untracked
;
1013 return read_directory_recursive(dir
, dirname
, len
, 1, simplify
);
1017 * This is an inexact early pruning of any recursive directory
1018 * reading - if the path cannot possibly be in the pathspec,
1019 * return true, and we'll skip it early.
1021 static int simplify_away(const char *path
, int pathlen
, const struct path_simplify
*simplify
)
1025 const char *match
= simplify
->path
;
1026 int len
= simplify
->len
;
1032 if (!memcmp(path
, match
, len
))
1042 * This function tells us whether an excluded path matches a
1043 * list of "interesting" pathspecs. That is, whether a path matched
1044 * by any of the pathspecs could possibly be ignored by excluding
1045 * the specified path. This can happen if:
1047 * 1. the path is mentioned explicitly in the pathspec
1049 * 2. the path is a directory prefix of some element in the
1052 static int exclude_matches_pathspec(const char *path
, int len
,
1053 const struct path_simplify
*simplify
)
1056 for (; simplify
->path
; simplify
++) {
1057 if (len
== simplify
->len
1058 && !memcmp(path
, simplify
->path
, len
))
1060 if (len
< simplify
->len
1061 && simplify
->path
[len
] == '/'
1062 && !memcmp(path
, simplify
->path
, len
))
1069 static int get_index_dtype(const char *path
, int len
)
1072 const struct cache_entry
*ce
;
1074 ce
= cache_name_exists(path
, len
, 0);
1076 if (!ce_uptodate(ce
))
1078 if (S_ISGITLINK(ce
->ce_mode
))
1081 * Nobody actually cares about the
1082 * difference between DT_LNK and DT_REG
1087 /* Try to look it up as a directory */
1088 pos
= cache_name_pos(path
, len
);
1092 while (pos
< active_nr
) {
1093 ce
= active_cache
[pos
++];
1094 if (strncmp(ce
->name
, path
, len
))
1096 if (ce
->name
[len
] > '/')
1098 if (ce
->name
[len
] < '/')
1100 if (!ce_uptodate(ce
))
1101 break; /* continue? */
1107 static int get_dtype(struct dirent
*de
, const char *path
, int len
)
1109 int dtype
= de
? DTYPE(de
) : DT_UNKNOWN
;
1112 if (dtype
!= DT_UNKNOWN
)
1114 dtype
= get_index_dtype(path
, len
);
1115 if (dtype
!= DT_UNKNOWN
)
1117 if (lstat(path
, &st
))
1119 if (S_ISREG(st
.st_mode
))
1121 if (S_ISDIR(st
.st_mode
))
1123 if (S_ISLNK(st
.st_mode
))
1128 static enum path_treatment
treat_one_path(struct dir_struct
*dir
,
1129 struct strbuf
*path
,
1130 const struct path_simplify
*simplify
,
1131 int dtype
, struct dirent
*de
)
1134 int has_path_in_index
= !!cache_name_exists(path
->buf
, path
->len
, ignore_case
);
1136 if (dtype
== DT_UNKNOWN
)
1137 dtype
= get_dtype(de
, path
->buf
, path
->len
);
1139 /* Always exclude indexed files */
1140 if (dtype
!= DT_DIR
&& has_path_in_index
)
1144 * When we are looking at a directory P in the working tree,
1145 * there are three cases:
1147 * (1) P exists in the index. Everything inside the directory P in
1148 * the working tree needs to go when P is checked out from the
1151 * (2) P does not exist in the index, but there is P/Q in the index.
1152 * We know P will stay a directory when we check out the contents
1153 * of the index, but we do not know yet if there is a directory
1154 * P/Q in the working tree to be killed, so we need to recurse.
1156 * (3) P does not exist in the index, and there is no P/Q in the index
1157 * to require P to be a directory, either. Only in this case, we
1158 * know that everything inside P will not be killed without
1161 if ((dir
->flags
& DIR_COLLECT_KILLED_ONLY
) &&
1162 (dtype
== DT_DIR
) &&
1163 !has_path_in_index
) {
1165 * NEEDSWORK: directory_exists_in_index_icase()
1166 * assumes that one byte past the given path is
1167 * readable and has '/', which needs to be fixed, but
1168 * until then, work it around in the caller.
1170 strbuf_addch(path
, '/');
1171 if (directory_exists_in_index(path
->buf
, path
->len
- 1) ==
1172 index_nonexistent
) {
1173 strbuf_setlen(path
, path
->len
- 1);
1176 strbuf_setlen(path
, path
->len
- 1);
1179 exclude
= is_excluded(dir
, path
->buf
, &dtype
);
1182 * Excluded? If we don't explicitly want to show
1183 * ignored files, ignore it
1185 if (exclude
&& !(dir
->flags
& (DIR_SHOW_IGNORED
|DIR_SHOW_IGNORED_TOO
)))
1186 return path_excluded
;
1192 strbuf_addch(path
, '/');
1193 return treat_directory(dir
, path
->buf
, path
->len
, exclude
,
1197 return exclude
? path_excluded
: path_untracked
;
1201 static enum path_treatment
treat_path(struct dir_struct
*dir
,
1203 struct strbuf
*path
,
1205 const struct path_simplify
*simplify
)
1209 if (is_dot_or_dotdot(de
->d_name
) || !strcmp(de
->d_name
, ".git"))
1211 strbuf_setlen(path
, baselen
);
1212 strbuf_addstr(path
, de
->d_name
);
1213 if (simplify_away(path
->buf
, path
->len
, simplify
))
1217 return treat_one_path(dir
, path
, simplify
, dtype
, de
);
1221 * Read a directory tree. We currently ignore anything but
1222 * directories, regular files and symlinks. That's because git
1223 * doesn't handle them at all yet. Maybe that will change some
1226 * Also, we ignore the name ".git" (even if it is not a directory).
1227 * That likely will not change.
1229 * Returns the most significant path_treatment value encountered in the scan.
1231 static enum path_treatment
read_directory_recursive(struct dir_struct
*dir
,
1232 const char *base
, int baselen
,
1234 const struct path_simplify
*simplify
)
1237 enum path_treatment state
, subdir_state
, dir_state
= path_none
;
1239 struct strbuf path
= STRBUF_INIT
;
1241 strbuf_add(&path
, base
, baselen
);
1243 fdir
= opendir(path
.len
? path
.buf
: ".");
1247 while ((de
= readdir(fdir
)) != NULL
) {
1248 /* check how the file or directory should be treated */
1249 state
= treat_path(dir
, de
, &path
, baselen
, simplify
);
1250 if (state
> dir_state
)
1253 /* recurse into subdir if instructed by treat_path */
1254 if (state
== path_recurse
) {
1255 subdir_state
= read_directory_recursive(dir
, path
.buf
,
1256 path
.len
, check_only
, simplify
);
1257 if (subdir_state
> dir_state
)
1258 dir_state
= subdir_state
;
1262 /* abort early if maximum state has been reached */
1263 if (dir_state
== path_untracked
)
1265 /* skip the dir_add_* part */
1269 /* add the path to the appropriate result list */
1272 if (dir
->flags
& DIR_SHOW_IGNORED
)
1273 dir_add_name(dir
, path
.buf
, path
.len
);
1274 else if ((dir
->flags
& DIR_SHOW_IGNORED_TOO
) ||
1275 ((dir
->flags
& DIR_COLLECT_IGNORED
) &&
1276 exclude_matches_pathspec(path
.buf
, path
.len
,
1278 dir_add_ignored(dir
, path
.buf
, path
.len
);
1281 case path_untracked
:
1282 if (!(dir
->flags
& DIR_SHOW_IGNORED
))
1283 dir_add_name(dir
, path
.buf
, path
.len
);
1292 strbuf_release(&path
);
1297 static int cmp_name(const void *p1
, const void *p2
)
1299 const struct dir_entry
*e1
= *(const struct dir_entry
**)p1
;
1300 const struct dir_entry
*e2
= *(const struct dir_entry
**)p2
;
1302 return cache_name_compare(e1
->name
, e1
->len
,
1306 static struct path_simplify
*create_simplify(const char **pathspec
)
1309 struct path_simplify
*simplify
= NULL
;
1314 for (nr
= 0 ; ; nr
++) {
1317 alloc
= alloc_nr(alloc
);
1318 simplify
= xrealloc(simplify
, alloc
* sizeof(*simplify
));
1320 match
= *pathspec
++;
1323 simplify
[nr
].path
= match
;
1324 simplify
[nr
].len
= simple_length(match
);
1326 simplify
[nr
].path
= NULL
;
1327 simplify
[nr
].len
= 0;
1331 static void free_simplify(struct path_simplify
*simplify
)
1336 static int treat_leading_path(struct dir_struct
*dir
,
1337 const char *path
, int len
,
1338 const struct path_simplify
*simplify
)
1340 struct strbuf sb
= STRBUF_INIT
;
1341 int baselen
, rc
= 0;
1343 int old_flags
= dir
->flags
;
1345 while (len
&& path
[len
- 1] == '/')
1350 dir
->flags
&= ~DIR_SHOW_OTHER_DIRECTORIES
;
1352 cp
= path
+ baselen
+ !!baselen
;
1353 cp
= memchr(cp
, '/', path
+ len
- cp
);
1357 baselen
= cp
- path
;
1358 strbuf_setlen(&sb
, 0);
1359 strbuf_add(&sb
, path
, baselen
);
1360 if (!is_directory(sb
.buf
))
1362 if (simplify_away(sb
.buf
, sb
.len
, simplify
))
1364 if (treat_one_path(dir
, &sb
, simplify
,
1365 DT_DIR
, NULL
) == path_none
)
1366 break; /* do not recurse into it */
1367 if (len
<= baselen
) {
1369 break; /* finished checking */
1372 strbuf_release(&sb
);
1373 dir
->flags
= old_flags
;
1377 int read_directory(struct dir_struct
*dir
, const char *path
, int len
, const struct pathspec
*pathspec
)
1379 struct path_simplify
*simplify
;
1382 * Check out create_simplify()
1385 GUARD_PATHSPEC(pathspec
,
1392 if (has_symlink_leading_path(path
, len
))
1395 simplify
= create_simplify(pathspec
? pathspec
->_raw
: NULL
);
1396 if (!len
|| treat_leading_path(dir
, path
, len
, simplify
))
1397 read_directory_recursive(dir
, path
, len
, 0, simplify
);
1398 free_simplify(simplify
);
1399 qsort(dir
->entries
, dir
->nr
, sizeof(struct dir_entry
*), cmp_name
);
1400 qsort(dir
->ignored
, dir
->ignored_nr
, sizeof(struct dir_entry
*), cmp_name
);
1404 int file_exists(const char *f
)
1407 return lstat(f
, &sb
) == 0;
1411 * Given two normalized paths (a trailing slash is ok), if subdir is
1412 * outside dir, return -1. Otherwise return the offset in subdir that
1413 * can be used as relative path to dir.
1415 int dir_inside_of(const char *subdir
, const char *dir
)
1419 assert(dir
&& subdir
&& *dir
&& *subdir
);
1421 while (*dir
&& *subdir
&& *dir
== *subdir
) {
1427 /* hel[p]/me vs hel[l]/yeah */
1428 if (*dir
&& *subdir
)
1432 return !*dir
? offset
: -1; /* same dir */
1434 /* foo/[b]ar vs foo/[] */
1435 if (is_dir_sep(dir
[-1]))
1436 return is_dir_sep(subdir
[-1]) ? offset
: -1;
1438 /* foo[/]bar vs foo[] */
1439 return is_dir_sep(*subdir
) ? offset
+ 1 : -1;
1442 int is_inside_dir(const char *dir
)
1447 if (!getcwd(cwd
, sizeof(cwd
)))
1448 die_errno("can't find the current directory");
1449 return dir_inside_of(cwd
, dir
) >= 0;
1452 int is_empty_dir(const char *path
)
1454 DIR *dir
= opendir(path
);
1461 while ((e
= readdir(dir
)) != NULL
)
1462 if (!is_dot_or_dotdot(e
->d_name
)) {
1471 static int remove_dir_recurse(struct strbuf
*path
, int flag
, int *kept_up
)
1475 int ret
= 0, original_len
= path
->len
, len
, kept_down
= 0;
1476 int only_empty
= (flag
& REMOVE_DIR_EMPTY_ONLY
);
1477 int keep_toplevel
= (flag
& REMOVE_DIR_KEEP_TOPLEVEL
);
1478 unsigned char submodule_head
[20];
1480 if ((flag
& REMOVE_DIR_KEEP_NESTED_GIT
) &&
1481 !resolve_gitlink_ref(path
->buf
, "HEAD", submodule_head
)) {
1482 /* Do not descend and nuke a nested git work tree. */
1488 flag
&= ~REMOVE_DIR_KEEP_TOPLEVEL
;
1489 dir
= opendir(path
->buf
);
1491 /* an empty dir could be removed even if it is unreadble */
1493 return rmdir(path
->buf
);
1497 if (path
->buf
[original_len
- 1] != '/')
1498 strbuf_addch(path
, '/');
1501 while ((e
= readdir(dir
)) != NULL
) {
1503 if (is_dot_or_dotdot(e
->d_name
))
1506 strbuf_setlen(path
, len
);
1507 strbuf_addstr(path
, e
->d_name
);
1508 if (lstat(path
->buf
, &st
))
1510 else if (S_ISDIR(st
.st_mode
)) {
1511 if (!remove_dir_recurse(path
, flag
, &kept_down
))
1512 continue; /* happy */
1513 } else if (!only_empty
&& !unlink(path
->buf
))
1514 continue; /* happy, too */
1516 /* path too long, stat fails, or non-directory still exists */
1522 strbuf_setlen(path
, original_len
);
1523 if (!ret
&& !keep_toplevel
&& !kept_down
)
1524 ret
= rmdir(path
->buf
);
1527 * report the uplevel that it is not an error that we
1528 * did not rmdir() our directory.
1534 int remove_dir_recursively(struct strbuf
*path
, int flag
)
1536 return remove_dir_recurse(path
, flag
, NULL
);
1539 void setup_standard_excludes(struct dir_struct
*dir
)
1544 dir
->exclude_per_dir
= ".gitignore";
1545 path
= git_path("info/exclude");
1546 if (!excludes_file
) {
1547 home_config_paths(NULL
, &xdg_path
, "ignore");
1548 excludes_file
= xdg_path
;
1550 if (!access_or_warn(path
, R_OK
, 0))
1551 add_excludes_from_file(dir
, path
);
1552 if (excludes_file
&& !access_or_warn(excludes_file
, R_OK
, 0))
1553 add_excludes_from_file(dir
, excludes_file
);
1556 int remove_path(const char *name
)
1560 if (unlink(name
) && errno
!= ENOENT
&& errno
!= ENOTDIR
)
1563 slash
= strrchr(name
, '/');
1565 char *dirs
= xstrdup(name
);
1566 slash
= dirs
+ (slash
- name
);
1569 } while (rmdir(dirs
) == 0 && (slash
= strrchr(dirs
, '/')));
1576 * Frees memory within dir which was allocated for exclude lists and
1577 * the exclude_stack. Does not free dir itself.
1579 void clear_directory(struct dir_struct
*dir
)
1582 struct exclude_list_group
*group
;
1583 struct exclude_list
*el
;
1584 struct exclude_stack
*stk
;
1586 for (i
= EXC_CMDL
; i
<= EXC_FILE
; i
++) {
1587 group
= &dir
->exclude_list_group
[i
];
1588 for (j
= 0; j
< group
->nr
; j
++) {
1591 free((char *)el
->src
);
1592 clear_exclude_list(el
);
1597 stk
= dir
->exclude_stack
;
1599 struct exclude_stack
*prev
= stk
->prev
;