1/*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
5 * See Documentation/technical/api-directory-listing.txt
6 *
7 * Copyright (C) Linus Torvalds, 2005-2006
8 * Junio Hamano, 2005-2006
9 */
10#include "cache.h"
11#include "dir.h"
12#include "refs.h"
13#include "wildmatch.h"
14#include "pathspec.h"
15#include "utf8.h"
16
17struct path_simplify {
18 int len;
19 const char *path;
20};
21
22/*
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.
27 */
28enum path_treatment {
29 path_none = 0,
30 path_recurse,
31 path_excluded,
32 path_untracked
33};
34
35static 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);
38static int get_dtype(struct dirent *de, const char *path, int len);
39
40/* helper string functions with support for the ignore_case flag */
41int strcmp_icase(const char *a, const char *b)
42{
43 return ignore_case ? strcasecmp(a, b) : strcmp(a, b);
44}
45
46int strncmp_icase(const char *a, const char *b, size_t count)
47{
48 return ignore_case ? strncasecmp(a, b, count) : strncmp(a, b, count);
49}
50
51int fnmatch_icase(const char *pattern, const char *string, int flags)
52{
53 return wildmatch(pattern, string,
54 flags | (ignore_case ? WM_CASEFOLD : 0),
55 NULL);
56}
57
58int git_fnmatch(const struct pathspec_item *item,
59 const char *pattern, const char *string,
60 int prefix)
61{
62 if (prefix > 0) {
63 if (ps_strncmp(item, pattern, string, prefix))
64 return WM_NOMATCH;
65 pattern += prefix;
66 string += prefix;
67 }
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);
74 }
75 if (item->magic & PATHSPEC_GLOB)
76 return wildmatch(pattern, string,
77 WM_PATHNAME |
78 (item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0),
79 NULL);
80 else
81 /* wildmatch has not learned no FNM_PATHNAME mode yet */
82 return wildmatch(pattern, string,
83 item->magic & PATHSPEC_ICASE ? WM_CASEFOLD : 0,
84 NULL);
85}
86
87static int fnmatch_icase_mem(const char *pattern, int patternlen,
88 const char *string, int stringlen,
89 int flags)
90{
91 int match_status;
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;
96
97 if (pattern[patternlen]) {
98 strbuf_add(&pat_buf, pattern, patternlen);
99 use_pat = pat_buf.buf;
100 }
101 if (string[stringlen]) {
102 strbuf_add(&str_buf, string, stringlen);
103 use_str = str_buf.buf;
104 }
105
106 if (ignore_case)
107 flags |= WM_CASEFOLD;
108 match_status = wildmatch(use_pat, use_str, flags, NULL);
109
110 strbuf_release(&pat_buf);
111 strbuf_release(&str_buf);
112
113 return match_status;
114}
115
116static size_t common_prefix_len(const struct pathspec *pathspec)
117{
118 int n;
119 size_t max = 0;
120
121 /*
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
126 * in non-:(icase).
127 */
128 GUARD_PATHSPEC(pathspec,
129 PATHSPEC_FROMTOP |
130 PATHSPEC_MAXDEPTH |
131 PATHSPEC_LITERAL |
132 PATHSPEC_GLOB |
133 PATHSPEC_ICASE |
134 PATHSPEC_EXCLUDE);
135
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)
139 continue;
140 if (pathspec->items[n].magic & PATHSPEC_ICASE)
141 item_len = pathspec->items[n].prefix;
142 else
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])
147 break;
148 if (c == '/')
149 len = i + 1;
150 i++;
151 }
152 if (n == 0 || len < max) {
153 max = len;
154 if (!max)
155 break;
156 }
157 }
158 return max;
159}
160
161/*
162 * Returns a copy of the longest leading path common among all
163 * pathspecs.
164 */
165char *common_prefix(const struct pathspec *pathspec)
166{
167 unsigned long len = common_prefix_len(pathspec);
168
169 return len ? xmemdupz(pathspec->items[0].match, len) : NULL;
170}
171
172int fill_directory(struct dir_struct *dir, const struct pathspec *pathspec)
173{
174 size_t len;
175
176 /*
177 * Calculate common prefix for the pathspec, and
178 * use that to optimize the directory walk
179 */
180 len = common_prefix_len(pathspec);
181
182 /* Read the directory and prune it */
183 read_directory(dir, pathspec->nr ? pathspec->_raw[0] : "", len, pathspec);
184 return len;
185}
186
187int within_depth(const char *name, int namelen,
188 int depth, int max_depth)
189{
190 const char *cp = name, *cpe = name + namelen;
191
192 while (cp < cpe) {
193 if (*cp++ != '/')
194 continue;
195 depth++;
196 if (depth > max_depth)
197 return 0;
198 }
199 return 1;
200}
201
202#define DO_MATCH_EXCLUDE 1
203#define DO_MATCH_DIRECTORY 2
204
205/*
206 * Does 'match' match the given name?
207 * A match is found if
208 *
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'.
212 *
213 * and the return value tells which case it was.
214 *
215 * It returns 0 when there is no match.
216 */
217static int match_pathspec_item(const struct pathspec_item *item, int prefix,
218 const char *name, int namelen, unsigned flags)
219{
220 /* name/namelen has prefix cut off by caller */
221 const char *match = item->match + prefix;
222 int matchlen = item->len - prefix;
223
224 /*
225 * The normal call pattern is:
226 * 1. prefix = common_prefix_len(ps);
227 * 2. prune something, or fill_directory
228 * 3. match_pathspec()
229 *
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
232 * trimmed at #3.
233 *
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.
242 *
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.
248 *
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)
252 */
253 if (item->prefix && (item->magic & PATHSPEC_ICASE) &&
254 strncmp(item->match, name - prefix, item->prefix))
255 return 0;
256
257 /* If the match was just the prefix, we matched */
258 if (!*match)
259 return MATCHED_RECURSIVELY;
260
261 if (matchlen <= namelen && !ps_strncmp(item, match, name, matchlen)) {
262 if (matchlen == namelen)
263 return MATCHED_EXACTLY;
264
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;
272
273 if (item->nowildcard_len < item->len &&
274 !git_fnmatch(item, match, name,
275 item->nowildcard_len - prefix))
276 return MATCHED_FNMATCH;
277
278 return 0;
279}
280
281/*
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
284 * pathspecs.
285 *
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.
295 */
296static int do_match_pathspec(const struct pathspec *ps,
297 const char *name, int namelen,
298 int prefix, char *seen,
299 unsigned flags)
300{
301 int i, retval = 0, exclude = flags & DO_MATCH_EXCLUDE;
302
303 GUARD_PATHSPEC(ps,
304 PATHSPEC_FROMTOP |
305 PATHSPEC_MAXDEPTH |
306 PATHSPEC_LITERAL |
307 PATHSPEC_GLOB |
308 PATHSPEC_ICASE |
309 PATHSPEC_EXCLUDE);
310
311 if (!ps->nr) {
312 if (!ps->recursive ||
313 !(ps->magic & PATHSPEC_MAXDEPTH) ||
314 ps->max_depth == -1)
315 return MATCHED_RECURSIVELY;
316
317 if (within_depth(name, namelen, 0, ps->max_depth))
318 return MATCHED_EXACTLY;
319 else
320 return 0;
321 }
322
323 name += prefix;
324 namelen -= prefix;
325
326 for (i = ps->nr - 1; i >= 0; i--) {
327 int how;
328
329 if ((!exclude && ps->items[i].magic & PATHSPEC_EXCLUDE) ||
330 ( exclude && !(ps->items[i].magic & PATHSPEC_EXCLUDE)))
331 continue;
332
333 if (seen && seen[i] == MATCHED_EXACTLY)
334 continue;
335 /*
336 * Make exclude patterns optional and never report
337 * "pathspec ':(exclude)foo' matches no files"
338 */
339 if (seen && ps->items[i].magic & PATHSPEC_EXCLUDE)
340 seen[i] = MATCHED_FNMATCH;
341 how = match_pathspec_item(ps->items+i, prefix, name,
342 namelen, flags);
343 if (ps->recursive &&
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] == '/')
349 len++;
350 if (within_depth(name+len, namelen-len, 0, ps->max_depth))
351 how = MATCHED_EXACTLY;
352 else
353 how = 0;
354 }
355 if (how) {
356 if (retval < how)
357 retval = how;
358 if (seen && seen[i] < how)
359 seen[i] = how;
360 }
361 }
362 return retval;
363}
364
365int match_pathspec(const struct pathspec *ps,
366 const char *name, int namelen,
367 int prefix, char *seen, int is_dir)
368{
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)
374 return positive;
375 negative = do_match_pathspec(ps, name, namelen,
376 prefix, seen,
377 flags | DO_MATCH_EXCLUDE);
378 return negative ? 0 : positive;
379}
380
381/*
382 * Return the length of the "simple" part of a path match limiter.
383 */
384int simple_length(const char *match)
385{
386 int len = -1;
387
388 for (;;) {
389 unsigned char c = *match++;
390 len++;
391 if (c == '\0' || is_glob_special(c))
392 return len;
393 }
394}
395
396int no_wildcard(const char *string)
397{
398 return string[simple_length(string)] == '\0';
399}
400
401void parse_exclude_pattern(const char **pattern,
402 int *patternlen,
403 int *flags,
404 int *nowildcardlen)
405{
406 const char *p = *pattern;
407 size_t i, len;
408
409 *flags = 0;
410 if (*p == '!') {
411 *flags |= EXC_FLAG_NEGATIVE;
412 p++;
413 }
414 len = strlen(p);
415 if (len && p[len - 1] == '/') {
416 len--;
417 *flags |= EXC_FLAG_MUSTBEDIR;
418 }
419 for (i = 0; i < len; i++) {
420 if (p[i] == '/')
421 break;
422 }
423 if (i == len)
424 *flags |= EXC_FLAG_NODIR;
425 *nowildcardlen = simple_length(p);
426 /*
427 * we should have excluded the trailing slash from 'p' too,
428 * but that's one more allocation. Instead just make sure
429 * nowildcardlen does not exceed real patternlen
430 */
431 if (*nowildcardlen > len)
432 *nowildcardlen = len;
433 if (*p == '*' && no_wildcard(p + 1))
434 *flags |= EXC_FLAG_ENDSWITH;
435 *pattern = p;
436 *patternlen = len;
437}
438
439void add_exclude(const char *string, const char *base,
440 int baselen, struct exclude_list *el, int srcpos)
441{
442 struct exclude *x;
443 int patternlen;
444 int flags;
445 int nowildcardlen;
446
447 parse_exclude_pattern(&string, &patternlen, &flags, &nowildcardlen);
448 if (flags & EXC_FLAG_MUSTBEDIR) {
449 char *s;
450 x = xmalloc(sizeof(*x) + patternlen + 1);
451 s = (char *)(x+1);
452 memcpy(s, string, patternlen);
453 s[patternlen] = '\0';
454 x->pattern = s;
455 } else {
456 x = xmalloc(sizeof(*x));
457 x->pattern = string;
458 }
459 x->patternlen = patternlen;
460 x->nowildcardlen = nowildcardlen;
461 x->base = base;
462 x->baselen = baselen;
463 x->flags = flags;
464 x->srcpos = srcpos;
465 ALLOC_GROW(el->excludes, el->nr + 1, el->alloc);
466 el->excludes[el->nr++] = x;
467 x->el = el;
468}
469
470static void *read_skip_worktree_file_from_index(const char *path, size_t *size)
471{
472 int pos, len;
473 unsigned long sz;
474 enum object_type type;
475 void *data;
476
477 len = strlen(path);
478 pos = cache_name_pos(path, len);
479 if (pos < 0)
480 return NULL;
481 if (!ce_skip_worktree(active_cache[pos]))
482 return NULL;
483 data = read_sha1_file(active_cache[pos]->sha1, &type, &sz);
484 if (!data || type != OBJ_BLOB) {
485 free(data);
486 return NULL;
487 }
488 *size = xsize_t(sz);
489 return data;
490}
491
492/*
493 * Frees memory within el which was allocated for exclude patterns and
494 * the file buffer. Does not free el itself.
495 */
496void clear_exclude_list(struct exclude_list *el)
497{
498 int i;
499
500 for (i = 0; i < el->nr; i++)
501 free(el->excludes[i]);
502 free(el->excludes);
503 free(el->filebuf);
504
505 el->nr = 0;
506 el->excludes = NULL;
507 el->filebuf = NULL;
508}
509
510static void trim_trailing_spaces(char *buf)
511{
512 char *p, *last_space = NULL;
513
514 for (p = buf; *p; p++)
515 switch (*p) {
516 case ' ':
517 if (!last_space)
518 last_space = p;
519 break;
520 case '\\':
521 p++;
522 if (!*p)
523 return;
524 /* fallthrough */
525 default:
526 last_space = NULL;
527 }
528
529 if (last_space)
530 *last_space = '\0';
531}
532
533int add_excludes_from_file_to_list(const char *fname,
534 const char *base,
535 int baselen,
536 struct exclude_list *el,
537 int check_index)
538{
539 struct stat st;
540 int fd, i, lineno = 1;
541 size_t size = 0;
542 char *buf, *entry;
543
544 fd = open(fname, O_RDONLY);
545 if (fd < 0 || fstat(fd, &st) < 0) {
546 if (errno != ENOENT)
547 warn_on_inaccessible(fname);
548 if (0 <= fd)
549 close(fd);
550 if (!check_index ||
551 (buf = read_skip_worktree_file_from_index(fname, &size)) == NULL)
552 return -1;
553 if (size == 0) {
554 free(buf);
555 return 0;
556 }
557 if (buf[size-1] != '\n') {
558 buf = xrealloc(buf, size+1);
559 buf[size++] = '\n';
560 }
561 } else {
562 size = xsize_t(st.st_size);
563 if (size == 0) {
564 close(fd);
565 return 0;
566 }
567 buf = xmalloc(size+1);
568 if (read_in_full(fd, buf, size) != size) {
569 free(buf);
570 close(fd);
571 return -1;
572 }
573 buf[size++] = '\n';
574 close(fd);
575 }
576
577 el->filebuf = buf;
578
579 if (skip_utf8_bom(&buf, size))
580 size -= buf - el->filebuf;
581
582 entry = buf;
583
584 for (i = 0; i < size; i++) {
585 if (buf[i] == '\n') {
586 if (entry != buf + i && entry[0] != '#') {
587 buf[i - (i && buf[i-1] == '\r')] = 0;
588 trim_trailing_spaces(entry);
589 add_exclude(entry, base, baselen, el, lineno);
590 }
591 lineno++;
592 entry = buf + i + 1;
593 }
594 }
595 return 0;
596}
597
598struct exclude_list *add_exclude_list(struct dir_struct *dir,
599 int group_type, const char *src)
600{
601 struct exclude_list *el;
602 struct exclude_list_group *group;
603
604 group = &dir->exclude_list_group[group_type];
605 ALLOC_GROW(group->el, group->nr + 1, group->alloc);
606 el = &group->el[group->nr++];
607 memset(el, 0, sizeof(*el));
608 el->src = src;
609 return el;
610}
611
612/*
613 * Used to set up core.excludesfile and .git/info/exclude lists.
614 */
615void add_excludes_from_file(struct dir_struct *dir, const char *fname)
616{
617 struct exclude_list *el;
618 el = add_exclude_list(dir, EXC_FILE, fname);
619 if (add_excludes_from_file_to_list(fname, "", 0, el, 0) < 0)
620 die("cannot use %s as an exclude file", fname);
621}
622
623int match_basename(const char *basename, int basenamelen,
624 const char *pattern, int prefix, int patternlen,
625 int flags)
626{
627 if (prefix == patternlen) {
628 if (patternlen == basenamelen &&
629 !strncmp_icase(pattern, basename, basenamelen))
630 return 1;
631 } else if (flags & EXC_FLAG_ENDSWITH) {
632 /* "*literal" matching against "fooliteral" */
633 if (patternlen - 1 <= basenamelen &&
634 !strncmp_icase(pattern + 1,
635 basename + basenamelen - (patternlen - 1),
636 patternlen - 1))
637 return 1;
638 } else {
639 if (fnmatch_icase_mem(pattern, patternlen,
640 basename, basenamelen,
641 0) == 0)
642 return 1;
643 }
644 return 0;
645}
646
647int match_pathname(const char *pathname, int pathlen,
648 const char *base, int baselen,
649 const char *pattern, int prefix, int patternlen,
650 int flags)
651{
652 const char *name;
653 int namelen;
654
655 /*
656 * match with FNM_PATHNAME; the pattern has base implicitly
657 * in front of it.
658 */
659 if (*pattern == '/') {
660 pattern++;
661 patternlen--;
662 prefix--;
663 }
664
665 /*
666 * baselen does not count the trailing slash. base[] may or
667 * may not end with a trailing slash though.
668 */
669 if (pathlen < baselen + 1 ||
670 (baselen && pathname[baselen] != '/') ||
671 strncmp_icase(pathname, base, baselen))
672 return 0;
673
674 namelen = baselen ? pathlen - baselen - 1 : pathlen;
675 name = pathname + pathlen - namelen;
676
677 if (prefix) {
678 /*
679 * if the non-wildcard part is longer than the
680 * remaining pathname, surely it cannot match.
681 */
682 if (prefix > namelen)
683 return 0;
684
685 if (strncmp_icase(pattern, name, prefix))
686 return 0;
687 pattern += prefix;
688 patternlen -= prefix;
689 name += prefix;
690 namelen -= prefix;
691
692 /*
693 * If the whole pattern did not have a wildcard,
694 * then our prefix match is all we need; we
695 * do not need to call fnmatch at all.
696 */
697 if (!patternlen && !namelen)
698 return 1;
699 }
700
701 return fnmatch_icase_mem(pattern, patternlen,
702 name, namelen,
703 WM_PATHNAME) == 0;
704}
705
706/*
707 * Scan the given exclude list in reverse to see whether pathname
708 * should be ignored. The first match (i.e. the last on the list), if
709 * any, determines the fate. Returns the exclude_list element which
710 * matched, or NULL for undecided.
711 */
712static struct exclude *last_exclude_matching_from_list(const char *pathname,
713 int pathlen,
714 const char *basename,
715 int *dtype,
716 struct exclude_list *el)
717{
718 int i;
719
720 if (!el->nr)
721 return NULL; /* undefined */
722
723 for (i = el->nr - 1; 0 <= i; i--) {
724 struct exclude *x = el->excludes[i];
725 const char *exclude = x->pattern;
726 int prefix = x->nowildcardlen;
727
728 if (x->flags & EXC_FLAG_MUSTBEDIR) {
729 if (*dtype == DT_UNKNOWN)
730 *dtype = get_dtype(NULL, pathname, pathlen);
731 if (*dtype != DT_DIR)
732 continue;
733 }
734
735 if (x->flags & EXC_FLAG_NODIR) {
736 if (match_basename(basename,
737 pathlen - (basename - pathname),
738 exclude, prefix, x->patternlen,
739 x->flags))
740 return x;
741 continue;
742 }
743
744 assert(x->baselen == 0 || x->base[x->baselen - 1] == '/');
745 if (match_pathname(pathname, pathlen,
746 x->base, x->baselen ? x->baselen - 1 : 0,
747 exclude, prefix, x->patternlen, x->flags))
748 return x;
749 }
750 return NULL; /* undecided */
751}
752
753/*
754 * Scan the list and let the last match determine the fate.
755 * Return 1 for exclude, 0 for include and -1 for undecided.
756 */
757int is_excluded_from_list(const char *pathname,
758 int pathlen, const char *basename, int *dtype,
759 struct exclude_list *el)
760{
761 struct exclude *exclude;
762 exclude = last_exclude_matching_from_list(pathname, pathlen, basename, dtype, el);
763 if (exclude)
764 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
765 return -1; /* undecided */
766}
767
768static struct exclude *last_exclude_matching_from_lists(struct dir_struct *dir,
769 const char *pathname, int pathlen, const char *basename,
770 int *dtype_p)
771{
772 int i, j;
773 struct exclude_list_group *group;
774 struct exclude *exclude;
775 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
776 group = &dir->exclude_list_group[i];
777 for (j = group->nr - 1; j >= 0; j--) {
778 exclude = last_exclude_matching_from_list(
779 pathname, pathlen, basename, dtype_p,
780 &group->el[j]);
781 if (exclude)
782 return exclude;
783 }
784 }
785 return NULL;
786}
787
788/*
789 * Loads the per-directory exclude list for the substring of base
790 * which has a char length of baselen.
791 */
792static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
793{
794 struct exclude_list_group *group;
795 struct exclude_list *el;
796 struct exclude_stack *stk = NULL;
797 int current;
798
799 group = &dir->exclude_list_group[EXC_DIRS];
800
801 /*
802 * Pop the exclude lists from the EXCL_DIRS exclude_list_group
803 * which originate from directories not in the prefix of the
804 * path being checked.
805 */
806 while ((stk = dir->exclude_stack) != NULL) {
807 if (stk->baselen <= baselen &&
808 !strncmp(dir->basebuf.buf, base, stk->baselen))
809 break;
810 el = &group->el[dir->exclude_stack->exclude_ix];
811 dir->exclude_stack = stk->prev;
812 dir->exclude = NULL;
813 free((char *)el->src); /* see strbuf_detach() below */
814 clear_exclude_list(el);
815 free(stk);
816 group->nr--;
817 }
818
819 /* Skip traversing into sub directories if the parent is excluded */
820 if (dir->exclude)
821 return;
822
823 /*
824 * Lazy initialization. All call sites currently just
825 * memset(dir, 0, sizeof(*dir)) before use. Changing all of
826 * them seems lots of work for little benefit.
827 */
828 if (!dir->basebuf.buf)
829 strbuf_init(&dir->basebuf, PATH_MAX);
830
831 /* Read from the parent directories and push them down. */
832 current = stk ? stk->baselen : -1;
833 strbuf_setlen(&dir->basebuf, current < 0 ? 0 : current);
834 while (current < baselen) {
835 const char *cp;
836
837 stk = xcalloc(1, sizeof(*stk));
838 if (current < 0) {
839 cp = base;
840 current = 0;
841 } else {
842 cp = strchr(base + current + 1, '/');
843 if (!cp)
844 die("oops in prep_exclude");
845 cp++;
846 }
847 stk->prev = dir->exclude_stack;
848 stk->baselen = cp - base;
849 stk->exclude_ix = group->nr;
850 el = add_exclude_list(dir, EXC_DIRS, NULL);
851 strbuf_add(&dir->basebuf, base + current, stk->baselen - current);
852 assert(stk->baselen == dir->basebuf.len);
853
854 /* Abort if the directory is excluded */
855 if (stk->baselen) {
856 int dt = DT_DIR;
857 dir->basebuf.buf[stk->baselen - 1] = 0;
858 dir->exclude = last_exclude_matching_from_lists(dir,
859 dir->basebuf.buf, stk->baselen - 1,
860 dir->basebuf.buf + current, &dt);
861 dir->basebuf.buf[stk->baselen - 1] = '/';
862 if (dir->exclude &&
863 dir->exclude->flags & EXC_FLAG_NEGATIVE)
864 dir->exclude = NULL;
865 if (dir->exclude) {
866 dir->exclude_stack = stk;
867 return;
868 }
869 }
870
871 /* Try to read per-directory file */
872 if (dir->exclude_per_dir) {
873 /*
874 * dir->basebuf gets reused by the traversal, but we
875 * need fname to remain unchanged to ensure the src
876 * member of each struct exclude correctly
877 * back-references its source file. Other invocations
878 * of add_exclude_list provide stable strings, so we
879 * strbuf_detach() and free() here in the caller.
880 */
881 struct strbuf sb = STRBUF_INIT;
882 strbuf_addbuf(&sb, &dir->basebuf);
883 strbuf_addstr(&sb, dir->exclude_per_dir);
884 el->src = strbuf_detach(&sb, NULL);
885 add_excludes_from_file_to_list(el->src, el->src,
886 stk->baselen, el, 1);
887 }
888 dir->exclude_stack = stk;
889 current = stk->baselen;
890 }
891 strbuf_setlen(&dir->basebuf, baselen);
892}
893
894/*
895 * Loads the exclude lists for the directory containing pathname, then
896 * scans all exclude lists to determine whether pathname is excluded.
897 * Returns the exclude_list element which matched, or NULL for
898 * undecided.
899 */
900struct exclude *last_exclude_matching(struct dir_struct *dir,
901 const char *pathname,
902 int *dtype_p)
903{
904 int pathlen = strlen(pathname);
905 const char *basename = strrchr(pathname, '/');
906 basename = (basename) ? basename+1 : pathname;
907
908 prep_exclude(dir, pathname, basename-pathname);
909
910 if (dir->exclude)
911 return dir->exclude;
912
913 return last_exclude_matching_from_lists(dir, pathname, pathlen,
914 basename, dtype_p);
915}
916
917/*
918 * Loads the exclude lists for the directory containing pathname, then
919 * scans all exclude lists to determine whether pathname is excluded.
920 * Returns 1 if true, otherwise 0.
921 */
922int is_excluded(struct dir_struct *dir, const char *pathname, int *dtype_p)
923{
924 struct exclude *exclude =
925 last_exclude_matching(dir, pathname, dtype_p);
926 if (exclude)
927 return exclude->flags & EXC_FLAG_NEGATIVE ? 0 : 1;
928 return 0;
929}
930
931static struct dir_entry *dir_entry_new(const char *pathname, int len)
932{
933 struct dir_entry *ent;
934
935 ent = xmalloc(sizeof(*ent) + len + 1);
936 ent->len = len;
937 memcpy(ent->name, pathname, len);
938 ent->name[len] = 0;
939 return ent;
940}
941
942static struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
943{
944 if (cache_file_exists(pathname, len, ignore_case))
945 return NULL;
946
947 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
948 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
949}
950
951struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
952{
953 if (!cache_name_is_other(pathname, len))
954 return NULL;
955
956 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
957 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
958}
959
960enum exist_status {
961 index_nonexistent = 0,
962 index_directory,
963 index_gitdir
964};
965
966/*
967 * Do not use the alphabetically sorted index to look up
968 * the directory name; instead, use the case insensitive
969 * directory hash.
970 */
971static enum exist_status directory_exists_in_index_icase(const char *dirname, int len)
972{
973 const struct cache_entry *ce = cache_dir_exists(dirname, len);
974 unsigned char endchar;
975
976 if (!ce)
977 return index_nonexistent;
978 endchar = ce->name[len];
979
980 /*
981 * The cache_entry structure returned will contain this dirname
982 * and possibly additional path components.
983 */
984 if (endchar == '/')
985 return index_directory;
986
987 /*
988 * If there are no additional path components, then this cache_entry
989 * represents a submodule. Submodules, despite being directories,
990 * are stored in the cache without a closing slash.
991 */
992 if (!endchar && S_ISGITLINK(ce->ce_mode))
993 return index_gitdir;
994
995 /* This should never be hit, but it exists just in case. */
996 return index_nonexistent;
997}
998
999/*
1000 * The index sorts alphabetically by entry name, which
1001 * means that a gitlink sorts as '\0' at the end, while
1002 * a directory (which is defined not as an entry, but as
1003 * the files it contains) will sort with the '/' at the
1004 * end.
1005 */
1006static enum exist_status directory_exists_in_index(const char *dirname, int len)
1007{
1008 int pos;
1009
1010 if (ignore_case)
1011 return directory_exists_in_index_icase(dirname, len);
1012
1013 pos = cache_name_pos(dirname, len);
1014 if (pos < 0)
1015 pos = -pos-1;
1016 while (pos < active_nr) {
1017 const struct cache_entry *ce = active_cache[pos++];
1018 unsigned char endchar;
1019
1020 if (strncmp(ce->name, dirname, len))
1021 break;
1022 endchar = ce->name[len];
1023 if (endchar > '/')
1024 break;
1025 if (endchar == '/')
1026 return index_directory;
1027 if (!endchar && S_ISGITLINK(ce->ce_mode))
1028 return index_gitdir;
1029 }
1030 return index_nonexistent;
1031}
1032
1033/*
1034 * When we find a directory when traversing the filesystem, we
1035 * have three distinct cases:
1036 *
1037 * - ignore it
1038 * - see it as a directory
1039 * - recurse into it
1040 *
1041 * and which one we choose depends on a combination of existing
1042 * git index contents and the flags passed into the directory
1043 * traversal routine.
1044 *
1045 * Case 1: If we *already* have entries in the index under that
1046 * directory name, we always recurse into the directory to see
1047 * all the files.
1048 *
1049 * Case 2: If we *already* have that directory name as a gitlink,
1050 * we always continue to see it as a gitlink, regardless of whether
1051 * there is an actual git directory there or not (it might not
1052 * be checked out as a subproject!)
1053 *
1054 * Case 3: if we didn't have it in the index previously, we
1055 * have a few sub-cases:
1056 *
1057 * (a) if "show_other_directories" is true, we show it as
1058 * just a directory, unless "hide_empty_directories" is
1059 * also true, in which case we need to check if it contains any
1060 * untracked and / or ignored files.
1061 * (b) if it looks like a git directory, and we don't have
1062 * 'no_gitlinks' set we treat it as a gitlink, and show it
1063 * as a directory.
1064 * (c) otherwise, we recurse into it.
1065 */
1066static enum path_treatment treat_directory(struct dir_struct *dir,
1067 const char *dirname, int len, int exclude,
1068 const struct path_simplify *simplify)
1069{
1070 /* The "len-1" is to strip the final '/' */
1071 switch (directory_exists_in_index(dirname, len-1)) {
1072 case index_directory:
1073 return path_recurse;
1074
1075 case index_gitdir:
1076 return path_none;
1077
1078 case index_nonexistent:
1079 if (dir->flags & DIR_SHOW_OTHER_DIRECTORIES)
1080 break;
1081 if (!(dir->flags & DIR_NO_GITLINKS)) {
1082 unsigned char sha1[20];
1083 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
1084 return path_untracked;
1085 }
1086 return path_recurse;
1087 }
1088
1089 /* This is the "show_other_directories" case */
1090
1091 if (!(dir->flags & DIR_HIDE_EMPTY_DIRECTORIES))
1092 return exclude ? path_excluded : path_untracked;
1093
1094 return read_directory_recursive(dir, dirname, len, 1, simplify);
1095}
1096
1097/*
1098 * This is an inexact early pruning of any recursive directory
1099 * reading - if the path cannot possibly be in the pathspec,
1100 * return true, and we'll skip it early.
1101 */
1102static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
1103{
1104 if (simplify) {
1105 for (;;) {
1106 const char *match = simplify->path;
1107 int len = simplify->len;
1108
1109 if (!match)
1110 break;
1111 if (len > pathlen)
1112 len = pathlen;
1113 if (!memcmp(path, match, len))
1114 return 0;
1115 simplify++;
1116 }
1117 return 1;
1118 }
1119 return 0;
1120}
1121
1122/*
1123 * This function tells us whether an excluded path matches a
1124 * list of "interesting" pathspecs. That is, whether a path matched
1125 * by any of the pathspecs could possibly be ignored by excluding
1126 * the specified path. This can happen if:
1127 *
1128 * 1. the path is mentioned explicitly in the pathspec
1129 *
1130 * 2. the path is a directory prefix of some element in the
1131 * pathspec
1132 */
1133static int exclude_matches_pathspec(const char *path, int len,
1134 const struct path_simplify *simplify)
1135{
1136 if (simplify) {
1137 for (; simplify->path; simplify++) {
1138 if (len == simplify->len
1139 && !memcmp(path, simplify->path, len))
1140 return 1;
1141 if (len < simplify->len
1142 && simplify->path[len] == '/'
1143 && !memcmp(path, simplify->path, len))
1144 return 1;
1145 }
1146 }
1147 return 0;
1148}
1149
1150static int get_index_dtype(const char *path, int len)
1151{
1152 int pos;
1153 const struct cache_entry *ce;
1154
1155 ce = cache_file_exists(path, len, 0);
1156 if (ce) {
1157 if (!ce_uptodate(ce))
1158 return DT_UNKNOWN;
1159 if (S_ISGITLINK(ce->ce_mode))
1160 return DT_DIR;
1161 /*
1162 * Nobody actually cares about the
1163 * difference between DT_LNK and DT_REG
1164 */
1165 return DT_REG;
1166 }
1167
1168 /* Try to look it up as a directory */
1169 pos = cache_name_pos(path, len);
1170 if (pos >= 0)
1171 return DT_UNKNOWN;
1172 pos = -pos-1;
1173 while (pos < active_nr) {
1174 ce = active_cache[pos++];
1175 if (strncmp(ce->name, path, len))
1176 break;
1177 if (ce->name[len] > '/')
1178 break;
1179 if (ce->name[len] < '/')
1180 continue;
1181 if (!ce_uptodate(ce))
1182 break; /* continue? */
1183 return DT_DIR;
1184 }
1185 return DT_UNKNOWN;
1186}
1187
1188static int get_dtype(struct dirent *de, const char *path, int len)
1189{
1190 int dtype = de ? DTYPE(de) : DT_UNKNOWN;
1191 struct stat st;
1192
1193 if (dtype != DT_UNKNOWN)
1194 return dtype;
1195 dtype = get_index_dtype(path, len);
1196 if (dtype != DT_UNKNOWN)
1197 return dtype;
1198 if (lstat(path, &st))
1199 return dtype;
1200 if (S_ISREG(st.st_mode))
1201 return DT_REG;
1202 if (S_ISDIR(st.st_mode))
1203 return DT_DIR;
1204 if (S_ISLNK(st.st_mode))
1205 return DT_LNK;
1206 return dtype;
1207}
1208
1209static enum path_treatment treat_one_path(struct dir_struct *dir,
1210 struct strbuf *path,
1211 const struct path_simplify *simplify,
1212 int dtype, struct dirent *de)
1213{
1214 int exclude;
1215 int has_path_in_index = !!cache_file_exists(path->buf, path->len, ignore_case);
1216
1217 if (dtype == DT_UNKNOWN)
1218 dtype = get_dtype(de, path->buf, path->len);
1219
1220 /* Always exclude indexed files */
1221 if (dtype != DT_DIR && has_path_in_index)
1222 return path_none;
1223
1224 /*
1225 * When we are looking at a directory P in the working tree,
1226 * there are three cases:
1227 *
1228 * (1) P exists in the index. Everything inside the directory P in
1229 * the working tree needs to go when P is checked out from the
1230 * index.
1231 *
1232 * (2) P does not exist in the index, but there is P/Q in the index.
1233 * We know P will stay a directory when we check out the contents
1234 * of the index, but we do not know yet if there is a directory
1235 * P/Q in the working tree to be killed, so we need to recurse.
1236 *
1237 * (3) P does not exist in the index, and there is no P/Q in the index
1238 * to require P to be a directory, either. Only in this case, we
1239 * know that everything inside P will not be killed without
1240 * recursing.
1241 */
1242 if ((dir->flags & DIR_COLLECT_KILLED_ONLY) &&
1243 (dtype == DT_DIR) &&
1244 !has_path_in_index &&
1245 (directory_exists_in_index(path->buf, path->len) == index_nonexistent))
1246 return path_none;
1247
1248 exclude = is_excluded(dir, path->buf, &dtype);
1249
1250 /*
1251 * Excluded? If we don't explicitly want to show
1252 * ignored files, ignore it
1253 */
1254 if (exclude && !(dir->flags & (DIR_SHOW_IGNORED|DIR_SHOW_IGNORED_TOO)))
1255 return path_excluded;
1256
1257 switch (dtype) {
1258 default:
1259 return path_none;
1260 case DT_DIR:
1261 strbuf_addch(path, '/');
1262 return treat_directory(dir, path->buf, path->len, exclude,
1263 simplify);
1264 case DT_REG:
1265 case DT_LNK:
1266 return exclude ? path_excluded : path_untracked;
1267 }
1268}
1269
1270static enum path_treatment treat_path(struct dir_struct *dir,
1271 struct dirent *de,
1272 struct strbuf *path,
1273 int baselen,
1274 const struct path_simplify *simplify)
1275{
1276 int dtype;
1277
1278 if (is_dot_or_dotdot(de->d_name) || !strcmp(de->d_name, ".git"))
1279 return path_none;
1280 strbuf_setlen(path, baselen);
1281 strbuf_addstr(path, de->d_name);
1282 if (simplify_away(path->buf, path->len, simplify))
1283 return path_none;
1284
1285 dtype = DTYPE(de);
1286 return treat_one_path(dir, path, simplify, dtype, de);
1287}
1288
1289/*
1290 * Read a directory tree. We currently ignore anything but
1291 * directories, regular files and symlinks. That's because git
1292 * doesn't handle them at all yet. Maybe that will change some
1293 * day.
1294 *
1295 * Also, we ignore the name ".git" (even if it is not a directory).
1296 * That likely will not change.
1297 *
1298 * Returns the most significant path_treatment value encountered in the scan.
1299 */
1300static enum path_treatment read_directory_recursive(struct dir_struct *dir,
1301 const char *base, int baselen,
1302 int check_only,
1303 const struct path_simplify *simplify)
1304{
1305 DIR *fdir;
1306 enum path_treatment state, subdir_state, dir_state = path_none;
1307 struct dirent *de;
1308 struct strbuf path = STRBUF_INIT;
1309
1310 strbuf_add(&path, base, baselen);
1311
1312 fdir = opendir(path.len ? path.buf : ".");
1313 if (!fdir)
1314 goto out;
1315
1316 while ((de = readdir(fdir)) != NULL) {
1317 /* check how the file or directory should be treated */
1318 state = treat_path(dir, de, &path, baselen, simplify);
1319 if (state > dir_state)
1320 dir_state = state;
1321
1322 /* recurse into subdir if instructed by treat_path */
1323 if (state == path_recurse) {
1324 subdir_state = read_directory_recursive(dir, path.buf,
1325 path.len, check_only, simplify);
1326 if (subdir_state > dir_state)
1327 dir_state = subdir_state;
1328 }
1329
1330 if (check_only) {
1331 /* abort early if maximum state has been reached */
1332 if (dir_state == path_untracked)
1333 break;
1334 /* skip the dir_add_* part */
1335 continue;
1336 }
1337
1338 /* add the path to the appropriate result list */
1339 switch (state) {
1340 case path_excluded:
1341 if (dir->flags & DIR_SHOW_IGNORED)
1342 dir_add_name(dir, path.buf, path.len);
1343 else if ((dir->flags & DIR_SHOW_IGNORED_TOO) ||
1344 ((dir->flags & DIR_COLLECT_IGNORED) &&
1345 exclude_matches_pathspec(path.buf, path.len,
1346 simplify)))
1347 dir_add_ignored(dir, path.buf, path.len);
1348 break;
1349
1350 case path_untracked:
1351 if (!(dir->flags & DIR_SHOW_IGNORED))
1352 dir_add_name(dir, path.buf, path.len);
1353 break;
1354
1355 default:
1356 break;
1357 }
1358 }
1359 closedir(fdir);
1360 out:
1361 strbuf_release(&path);
1362
1363 return dir_state;
1364}
1365
1366static int cmp_name(const void *p1, const void *p2)
1367{
1368 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
1369 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
1370
1371 return name_compare(e1->name, e1->len, e2->name, e2->len);
1372}
1373
1374static struct path_simplify *create_simplify(const char **pathspec)
1375{
1376 int nr, alloc = 0;
1377 struct path_simplify *simplify = NULL;
1378
1379 if (!pathspec)
1380 return NULL;
1381
1382 for (nr = 0 ; ; nr++) {
1383 const char *match;
1384 ALLOC_GROW(simplify, nr + 1, alloc);
1385 match = *pathspec++;
1386 if (!match)
1387 break;
1388 simplify[nr].path = match;
1389 simplify[nr].len = simple_length(match);
1390 }
1391 simplify[nr].path = NULL;
1392 simplify[nr].len = 0;
1393 return simplify;
1394}
1395
1396static void free_simplify(struct path_simplify *simplify)
1397{
1398 free(simplify);
1399}
1400
1401static int treat_leading_path(struct dir_struct *dir,
1402 const char *path, int len,
1403 const struct path_simplify *simplify)
1404{
1405 struct strbuf sb = STRBUF_INIT;
1406 int baselen, rc = 0;
1407 const char *cp;
1408 int old_flags = dir->flags;
1409
1410 while (len && path[len - 1] == '/')
1411 len--;
1412 if (!len)
1413 return 1;
1414 baselen = 0;
1415 dir->flags &= ~DIR_SHOW_OTHER_DIRECTORIES;
1416 while (1) {
1417 cp = path + baselen + !!baselen;
1418 cp = memchr(cp, '/', path + len - cp);
1419 if (!cp)
1420 baselen = len;
1421 else
1422 baselen = cp - path;
1423 strbuf_setlen(&sb, 0);
1424 strbuf_add(&sb, path, baselen);
1425 if (!is_directory(sb.buf))
1426 break;
1427 if (simplify_away(sb.buf, sb.len, simplify))
1428 break;
1429 if (treat_one_path(dir, &sb, simplify,
1430 DT_DIR, NULL) == path_none)
1431 break; /* do not recurse into it */
1432 if (len <= baselen) {
1433 rc = 1;
1434 break; /* finished checking */
1435 }
1436 }
1437 strbuf_release(&sb);
1438 dir->flags = old_flags;
1439 return rc;
1440}
1441
1442int read_directory(struct dir_struct *dir, const char *path, int len, const struct pathspec *pathspec)
1443{
1444 struct path_simplify *simplify;
1445
1446 /*
1447 * Check out create_simplify()
1448 */
1449 if (pathspec)
1450 GUARD_PATHSPEC(pathspec,
1451 PATHSPEC_FROMTOP |
1452 PATHSPEC_MAXDEPTH |
1453 PATHSPEC_LITERAL |
1454 PATHSPEC_GLOB |
1455 PATHSPEC_ICASE |
1456 PATHSPEC_EXCLUDE);
1457
1458 if (has_symlink_leading_path(path, len))
1459 return dir->nr;
1460
1461 /*
1462 * exclude patterns are treated like positive ones in
1463 * create_simplify. Usually exclude patterns should be a
1464 * subset of positive ones, which has no impacts on
1465 * create_simplify().
1466 */
1467 simplify = create_simplify(pathspec ? pathspec->_raw : NULL);
1468 if (!len || treat_leading_path(dir, path, len, simplify))
1469 read_directory_recursive(dir, path, len, 0, simplify);
1470 free_simplify(simplify);
1471 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
1472 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
1473 return dir->nr;
1474}
1475
1476int file_exists(const char *f)
1477{
1478 struct stat sb;
1479 return lstat(f, &sb) == 0;
1480}
1481
1482/*
1483 * Given two normalized paths (a trailing slash is ok), if subdir is
1484 * outside dir, return -1. Otherwise return the offset in subdir that
1485 * can be used as relative path to dir.
1486 */
1487int dir_inside_of(const char *subdir, const char *dir)
1488{
1489 int offset = 0;
1490
1491 assert(dir && subdir && *dir && *subdir);
1492
1493 while (*dir && *subdir && *dir == *subdir) {
1494 dir++;
1495 subdir++;
1496 offset++;
1497 }
1498
1499 /* hel[p]/me vs hel[l]/yeah */
1500 if (*dir && *subdir)
1501 return -1;
1502
1503 if (!*subdir)
1504 return !*dir ? offset : -1; /* same dir */
1505
1506 /* foo/[b]ar vs foo/[] */
1507 if (is_dir_sep(dir[-1]))
1508 return is_dir_sep(subdir[-1]) ? offset : -1;
1509
1510 /* foo[/]bar vs foo[] */
1511 return is_dir_sep(*subdir) ? offset + 1 : -1;
1512}
1513
1514int is_inside_dir(const char *dir)
1515{
1516 char *cwd;
1517 int rc;
1518
1519 if (!dir)
1520 return 0;
1521
1522 cwd = xgetcwd();
1523 rc = (dir_inside_of(cwd, dir) >= 0);
1524 free(cwd);
1525 return rc;
1526}
1527
1528int is_empty_dir(const char *path)
1529{
1530 DIR *dir = opendir(path);
1531 struct dirent *e;
1532 int ret = 1;
1533
1534 if (!dir)
1535 return 0;
1536
1537 while ((e = readdir(dir)) != NULL)
1538 if (!is_dot_or_dotdot(e->d_name)) {
1539 ret = 0;
1540 break;
1541 }
1542
1543 closedir(dir);
1544 return ret;
1545}
1546
1547static int remove_dir_recurse(struct strbuf *path, int flag, int *kept_up)
1548{
1549 DIR *dir;
1550 struct dirent *e;
1551 int ret = 0, original_len = path->len, len, kept_down = 0;
1552 int only_empty = (flag & REMOVE_DIR_EMPTY_ONLY);
1553 int keep_toplevel = (flag & REMOVE_DIR_KEEP_TOPLEVEL);
1554 unsigned char submodule_head[20];
1555
1556 if ((flag & REMOVE_DIR_KEEP_NESTED_GIT) &&
1557 !resolve_gitlink_ref(path->buf, "HEAD", submodule_head)) {
1558 /* Do not descend and nuke a nested git work tree. */
1559 if (kept_up)
1560 *kept_up = 1;
1561 return 0;
1562 }
1563
1564 flag &= ~REMOVE_DIR_KEEP_TOPLEVEL;
1565 dir = opendir(path->buf);
1566 if (!dir) {
1567 if (errno == ENOENT)
1568 return keep_toplevel ? -1 : 0;
1569 else if (errno == EACCES && !keep_toplevel)
1570 /*
1571 * An empty dir could be removable even if it
1572 * is unreadable:
1573 */
1574 return rmdir(path->buf);
1575 else
1576 return -1;
1577 }
1578 if (path->buf[original_len - 1] != '/')
1579 strbuf_addch(path, '/');
1580
1581 len = path->len;
1582 while ((e = readdir(dir)) != NULL) {
1583 struct stat st;
1584 if (is_dot_or_dotdot(e->d_name))
1585 continue;
1586
1587 strbuf_setlen(path, len);
1588 strbuf_addstr(path, e->d_name);
1589 if (lstat(path->buf, &st)) {
1590 if (errno == ENOENT)
1591 /*
1592 * file disappeared, which is what we
1593 * wanted anyway
1594 */
1595 continue;
1596 /* fall thru */
1597 } else if (S_ISDIR(st.st_mode)) {
1598 if (!remove_dir_recurse(path, flag, &kept_down))
1599 continue; /* happy */
1600 } else if (!only_empty &&
1601 (!unlink(path->buf) || errno == ENOENT)) {
1602 continue; /* happy, too */
1603 }
1604
1605 /* path too long, stat fails, or non-directory still exists */
1606 ret = -1;
1607 break;
1608 }
1609 closedir(dir);
1610
1611 strbuf_setlen(path, original_len);
1612 if (!ret && !keep_toplevel && !kept_down)
1613 ret = (!rmdir(path->buf) || errno == ENOENT) ? 0 : -1;
1614 else if (kept_up)
1615 /*
1616 * report the uplevel that it is not an error that we
1617 * did not rmdir() our directory.
1618 */
1619 *kept_up = !ret;
1620 return ret;
1621}
1622
1623int remove_dir_recursively(struct strbuf *path, int flag)
1624{
1625 return remove_dir_recurse(path, flag, NULL);
1626}
1627
1628void setup_standard_excludes(struct dir_struct *dir)
1629{
1630 const char *path;
1631 char *xdg_path;
1632
1633 dir->exclude_per_dir = ".gitignore";
1634 path = git_path("info/exclude");
1635 if (!excludes_file) {
1636 home_config_paths(NULL, &xdg_path, "ignore");
1637 excludes_file = xdg_path;
1638 }
1639 if (!access_or_warn(path, R_OK, 0))
1640 add_excludes_from_file(dir, path);
1641 if (excludes_file && !access_or_warn(excludes_file, R_OK, 0))
1642 add_excludes_from_file(dir, excludes_file);
1643}
1644
1645int remove_path(const char *name)
1646{
1647 char *slash;
1648
1649 if (unlink(name) && errno != ENOENT && errno != ENOTDIR)
1650 return -1;
1651
1652 slash = strrchr(name, '/');
1653 if (slash) {
1654 char *dirs = xstrdup(name);
1655 slash = dirs + (slash - name);
1656 do {
1657 *slash = '\0';
1658 } while (rmdir(dirs) == 0 && (slash = strrchr(dirs, '/')));
1659 free(dirs);
1660 }
1661 return 0;
1662}
1663
1664/*
1665 * Frees memory within dir which was allocated for exclude lists and
1666 * the exclude_stack. Does not free dir itself.
1667 */
1668void clear_directory(struct dir_struct *dir)
1669{
1670 int i, j;
1671 struct exclude_list_group *group;
1672 struct exclude_list *el;
1673 struct exclude_stack *stk;
1674
1675 for (i = EXC_CMDL; i <= EXC_FILE; i++) {
1676 group = &dir->exclude_list_group[i];
1677 for (j = 0; j < group->nr; j++) {
1678 el = &group->el[j];
1679 if (i == EXC_DIRS)
1680 free((char *)el->src);
1681 clear_exclude_list(el);
1682 }
1683 free(group->el);
1684 }
1685
1686 stk = dir->exclude_stack;
1687 while (stk) {
1688 struct exclude_stack *prev = stk->prev;
1689 free(stk);
1690 stk = prev;
1691 }
1692 strbuf_release(&dir->basebuf);
1693}