Make on-disk index representation separate from in-core one
[git/git.git] / dir.c
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
4 *
5 * Copyright (C) Linus Torvalds, 2005-2006
6 * Junio Hamano, 2005-2006
7 */
8 #include "cache.h"
9 #include "dir.h"
10 #include "refs.h"
11
12 struct path_simplify {
13 int len;
14 const char *path;
15 };
16
17 static int read_directory_recursive(struct dir_struct *dir,
18 const char *path, const char *base, int baselen,
19 int check_only, const struct path_simplify *simplify);
20
21 int common_prefix(const char **pathspec)
22 {
23 const char *path, *slash, *next;
24 int prefix;
25
26 if (!pathspec)
27 return 0;
28
29 path = *pathspec;
30 slash = strrchr(path, '/');
31 if (!slash)
32 return 0;
33
34 prefix = slash - path + 1;
35 while ((next = *++pathspec) != NULL) {
36 int len = strlen(next);
37 if (len >= prefix && !memcmp(path, next, prefix))
38 continue;
39 len = prefix - 1;
40 for (;;) {
41 if (!len)
42 return 0;
43 if (next[--len] != '/')
44 continue;
45 if (memcmp(path, next, len+1))
46 continue;
47 prefix = len + 1;
48 break;
49 }
50 }
51 return prefix;
52 }
53
54 /*
55 * Does 'match' matches the given name?
56 * A match is found if
57 *
58 * (1) the 'match' string is leading directory of 'name', or
59 * (2) the 'match' string is a wildcard and matches 'name', or
60 * (3) the 'match' string is exactly the same as 'name'.
61 *
62 * and the return value tells which case it was.
63 *
64 * It returns 0 when there is no match.
65 */
66 static int match_one(const char *match, const char *name, int namelen)
67 {
68 int matchlen;
69
70 /* If the match was just the prefix, we matched */
71 matchlen = strlen(match);
72 if (!matchlen)
73 return MATCHED_RECURSIVELY;
74
75 /*
76 * If we don't match the matchstring exactly,
77 * we need to match by fnmatch
78 */
79 if (strncmp(match, name, matchlen))
80 return !fnmatch(match, name, 0) ? MATCHED_FNMATCH : 0;
81
82 if (!name[matchlen])
83 return MATCHED_EXACTLY;
84 if (match[matchlen-1] == '/' || name[matchlen] == '/')
85 return MATCHED_RECURSIVELY;
86 return 0;
87 }
88
89 /*
90 * Given a name and a list of pathspecs, see if the name matches
91 * any of the pathspecs. The caller is also interested in seeing
92 * all pathspec matches some names it calls this function with
93 * (otherwise the user could have mistyped the unmatched pathspec),
94 * and a mark is left in seen[] array for pathspec element that
95 * actually matched anything.
96 */
97 int match_pathspec(const char **pathspec, const char *name, int namelen, int prefix, char *seen)
98 {
99 int retval;
100 const char *match;
101
102 name += prefix;
103 namelen -= prefix;
104
105 for (retval = 0; (match = *pathspec++) != NULL; seen++) {
106 int how;
107 if (retval && *seen == MATCHED_EXACTLY)
108 continue;
109 match += prefix;
110 how = match_one(match, name, namelen);
111 if (how) {
112 if (retval < how)
113 retval = how;
114 if (*seen < how)
115 *seen = how;
116 }
117 }
118 return retval;
119 }
120
121 static int no_wildcard(const char *string)
122 {
123 return string[strcspn(string, "*?[{")] == '\0';
124 }
125
126 void add_exclude(const char *string, const char *base,
127 int baselen, struct exclude_list *which)
128 {
129 struct exclude *x = xmalloc(sizeof (*x));
130
131 x->to_exclude = 1;
132 if (*string == '!') {
133 x->to_exclude = 0;
134 string++;
135 }
136 x->pattern = string;
137 x->patternlen = strlen(string);
138 x->base = base;
139 x->baselen = baselen;
140 x->flags = 0;
141 if (!strchr(string, '/'))
142 x->flags |= EXC_FLAG_NODIR;
143 if (no_wildcard(string))
144 x->flags |= EXC_FLAG_NOWILDCARD;
145 if (*string == '*' && no_wildcard(string+1))
146 x->flags |= EXC_FLAG_ENDSWITH;
147 ALLOC_GROW(which->excludes, which->nr + 1, which->alloc);
148 which->excludes[which->nr++] = x;
149 }
150
151 static int add_excludes_from_file_1(const char *fname,
152 const char *base,
153 int baselen,
154 char **buf_p,
155 struct exclude_list *which)
156 {
157 struct stat st;
158 int fd, i;
159 size_t size;
160 char *buf, *entry;
161
162 fd = open(fname, O_RDONLY);
163 if (fd < 0 || fstat(fd, &st) < 0)
164 goto err;
165 size = xsize_t(st.st_size);
166 if (size == 0) {
167 close(fd);
168 return 0;
169 }
170 buf = xmalloc(size+1);
171 if (read_in_full(fd, buf, size) != size)
172 {
173 free(buf);
174 goto err;
175 }
176 close(fd);
177
178 if (buf_p)
179 *buf_p = buf;
180 buf[size++] = '\n';
181 entry = buf;
182 for (i = 0; i < size; i++) {
183 if (buf[i] == '\n') {
184 if (entry != buf + i && entry[0] != '#') {
185 buf[i - (i && buf[i-1] == '\r')] = 0;
186 add_exclude(entry, base, baselen, which);
187 }
188 entry = buf + i + 1;
189 }
190 }
191 return 0;
192
193 err:
194 if (0 <= fd)
195 close(fd);
196 return -1;
197 }
198
199 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
200 {
201 if (add_excludes_from_file_1(fname, "", 0, NULL,
202 &dir->exclude_list[EXC_FILE]) < 0)
203 die("cannot use %s as an exclude file", fname);
204 }
205
206 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
207 {
208 struct exclude_list *el;
209 struct exclude_stack *stk = NULL;
210 int current;
211
212 if ((!dir->exclude_per_dir) ||
213 (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
214 return; /* too long a path -- ignore */
215
216 /* Pop the ones that are not the prefix of the path being checked. */
217 el = &dir->exclude_list[EXC_DIRS];
218 while ((stk = dir->exclude_stack) != NULL) {
219 if (stk->baselen <= baselen &&
220 !strncmp(dir->basebuf, base, stk->baselen))
221 break;
222 dir->exclude_stack = stk->prev;
223 while (stk->exclude_ix < el->nr)
224 free(el->excludes[--el->nr]);
225 free(stk->filebuf);
226 free(stk);
227 }
228
229 /* Read from the parent directories and push them down. */
230 current = stk ? stk->baselen : -1;
231 while (current < baselen) {
232 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
233 const char *cp;
234
235 if (current < 0) {
236 cp = base;
237 current = 0;
238 }
239 else {
240 cp = strchr(base + current + 1, '/');
241 if (!cp)
242 die("oops in prep_exclude");
243 cp++;
244 }
245 stk->prev = dir->exclude_stack;
246 stk->baselen = cp - base;
247 stk->exclude_ix = el->nr;
248 memcpy(dir->basebuf + current, base + current,
249 stk->baselen - current);
250 strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
251 add_excludes_from_file_1(dir->basebuf,
252 dir->basebuf, stk->baselen,
253 &stk->filebuf, el);
254 dir->exclude_stack = stk;
255 current = stk->baselen;
256 }
257 dir->basebuf[baselen] = '\0';
258 }
259
260 /* Scan the list and let the last match determines the fate.
261 * Return 1 for exclude, 0 for include and -1 for undecided.
262 */
263 static int excluded_1(const char *pathname,
264 int pathlen, const char *basename,
265 struct exclude_list *el)
266 {
267 int i;
268
269 if (el->nr) {
270 for (i = el->nr - 1; 0 <= i; i--) {
271 struct exclude *x = el->excludes[i];
272 const char *exclude = x->pattern;
273 int to_exclude = x->to_exclude;
274
275 if (x->flags & EXC_FLAG_NODIR) {
276 /* match basename */
277 if (x->flags & EXC_FLAG_NOWILDCARD) {
278 if (!strcmp(exclude, basename))
279 return to_exclude;
280 } else if (x->flags & EXC_FLAG_ENDSWITH) {
281 if (x->patternlen - 1 <= pathlen &&
282 !strcmp(exclude + 1, pathname + pathlen - x->patternlen + 1))
283 return to_exclude;
284 } else {
285 if (fnmatch(exclude, basename, 0) == 0)
286 return to_exclude;
287 }
288 }
289 else {
290 /* match with FNM_PATHNAME:
291 * exclude has base (baselen long) implicitly
292 * in front of it.
293 */
294 int baselen = x->baselen;
295 if (*exclude == '/')
296 exclude++;
297
298 if (pathlen < baselen ||
299 (baselen && pathname[baselen-1] != '/') ||
300 strncmp(pathname, x->base, baselen))
301 continue;
302
303 if (x->flags & EXC_FLAG_NOWILDCARD) {
304 if (!strcmp(exclude, pathname + baselen))
305 return to_exclude;
306 } else {
307 if (fnmatch(exclude, pathname+baselen,
308 FNM_PATHNAME) == 0)
309 return to_exclude;
310 }
311 }
312 }
313 }
314 return -1; /* undecided */
315 }
316
317 int excluded(struct dir_struct *dir, const char *pathname)
318 {
319 int pathlen = strlen(pathname);
320 int st;
321 const char *basename = strrchr(pathname, '/');
322 basename = (basename) ? basename+1 : pathname;
323
324 prep_exclude(dir, pathname, basename-pathname);
325 for (st = EXC_CMDL; st <= EXC_FILE; st++) {
326 switch (excluded_1(pathname, pathlen, basename, &dir->exclude_list[st])) {
327 case 0:
328 return 0;
329 case 1:
330 return 1;
331 }
332 }
333 return 0;
334 }
335
336 static struct dir_entry *dir_entry_new(const char *pathname, int len)
337 {
338 struct dir_entry *ent;
339
340 ent = xmalloc(sizeof(*ent) + len + 1);
341 ent->len = len;
342 memcpy(ent->name, pathname, len);
343 ent->name[len] = 0;
344 return ent;
345 }
346
347 struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
348 {
349 if (cache_name_pos(pathname, len) >= 0)
350 return NULL;
351
352 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
353 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
354 }
355
356 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
357 {
358 if (cache_name_pos(pathname, len) >= 0)
359 return NULL;
360
361 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
362 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
363 }
364
365 enum exist_status {
366 index_nonexistent = 0,
367 index_directory,
368 index_gitdir,
369 };
370
371 /*
372 * The index sorts alphabetically by entry name, which
373 * means that a gitlink sorts as '\0' at the end, while
374 * a directory (which is defined not as an entry, but as
375 * the files it contains) will sort with the '/' at the
376 * end.
377 */
378 static enum exist_status directory_exists_in_index(const char *dirname, int len)
379 {
380 int pos = cache_name_pos(dirname, len);
381 if (pos < 0)
382 pos = -pos-1;
383 while (pos < active_nr) {
384 struct cache_entry *ce = active_cache[pos++];
385 unsigned char endchar;
386
387 if (strncmp(ce->name, dirname, len))
388 break;
389 endchar = ce->name[len];
390 if (endchar > '/')
391 break;
392 if (endchar == '/')
393 return index_directory;
394 if (!endchar && S_ISGITLINK(ce->ce_mode))
395 return index_gitdir;
396 }
397 return index_nonexistent;
398 }
399
400 /*
401 * When we find a directory when traversing the filesystem, we
402 * have three distinct cases:
403 *
404 * - ignore it
405 * - see it as a directory
406 * - recurse into it
407 *
408 * and which one we choose depends on a combination of existing
409 * git index contents and the flags passed into the directory
410 * traversal routine.
411 *
412 * Case 1: If we *already* have entries in the index under that
413 * directory name, we always recurse into the directory to see
414 * all the files.
415 *
416 * Case 2: If we *already* have that directory name as a gitlink,
417 * we always continue to see it as a gitlink, regardless of whether
418 * there is an actual git directory there or not (it might not
419 * be checked out as a subproject!)
420 *
421 * Case 3: if we didn't have it in the index previously, we
422 * have a few sub-cases:
423 *
424 * (a) if "show_other_directories" is true, we show it as
425 * just a directory, unless "hide_empty_directories" is
426 * also true and the directory is empty, in which case
427 * we just ignore it entirely.
428 * (b) if it looks like a git directory, and we don't have
429 * 'no_gitlinks' set we treat it as a gitlink, and show it
430 * as a directory.
431 * (c) otherwise, we recurse into it.
432 */
433 enum directory_treatment {
434 show_directory,
435 ignore_directory,
436 recurse_into_directory,
437 };
438
439 static enum directory_treatment treat_directory(struct dir_struct *dir,
440 const char *dirname, int len,
441 const struct path_simplify *simplify)
442 {
443 /* The "len-1" is to strip the final '/' */
444 switch (directory_exists_in_index(dirname, len-1)) {
445 case index_directory:
446 return recurse_into_directory;
447
448 case index_gitdir:
449 if (dir->show_other_directories)
450 return ignore_directory;
451 return show_directory;
452
453 case index_nonexistent:
454 if (dir->show_other_directories)
455 break;
456 if (!dir->no_gitlinks) {
457 unsigned char sha1[20];
458 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
459 return show_directory;
460 }
461 return recurse_into_directory;
462 }
463
464 /* This is the "show_other_directories" case */
465 if (!dir->hide_empty_directories)
466 return show_directory;
467 if (!read_directory_recursive(dir, dirname, dirname, len, 1, simplify))
468 return ignore_directory;
469 return show_directory;
470 }
471
472 /*
473 * This is an inexact early pruning of any recursive directory
474 * reading - if the path cannot possibly be in the pathspec,
475 * return true, and we'll skip it early.
476 */
477 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
478 {
479 if (simplify) {
480 for (;;) {
481 const char *match = simplify->path;
482 int len = simplify->len;
483
484 if (!match)
485 break;
486 if (len > pathlen)
487 len = pathlen;
488 if (!memcmp(path, match, len))
489 return 0;
490 simplify++;
491 }
492 return 1;
493 }
494 return 0;
495 }
496
497 static int in_pathspec(const char *path, int len, const struct path_simplify *simplify)
498 {
499 if (simplify) {
500 for (; simplify->path; simplify++) {
501 if (len == simplify->len
502 && !memcmp(path, simplify->path, len))
503 return 1;
504 }
505 }
506 return 0;
507 }
508
509 static int get_dtype(struct dirent *de, const char *path)
510 {
511 int dtype = DTYPE(de);
512 struct stat st;
513
514 if (dtype != DT_UNKNOWN)
515 return dtype;
516 if (lstat(path, &st))
517 return dtype;
518 if (S_ISREG(st.st_mode))
519 return DT_REG;
520 if (S_ISDIR(st.st_mode))
521 return DT_DIR;
522 if (S_ISLNK(st.st_mode))
523 return DT_LNK;
524 return dtype;
525 }
526
527 /*
528 * Read a directory tree. We currently ignore anything but
529 * directories, regular files and symlinks. That's because git
530 * doesn't handle them at all yet. Maybe that will change some
531 * day.
532 *
533 * Also, we ignore the name ".git" (even if it is not a directory).
534 * That likely will not change.
535 */
536 static int read_directory_recursive(struct dir_struct *dir, const char *path, const char *base, int baselen, int check_only, const struct path_simplify *simplify)
537 {
538 DIR *fdir = opendir(path);
539 int contents = 0;
540
541 if (fdir) {
542 struct dirent *de;
543 char fullname[PATH_MAX + 1];
544 memcpy(fullname, base, baselen);
545
546 while ((de = readdir(fdir)) != NULL) {
547 int len, dtype;
548 int exclude;
549
550 if ((de->d_name[0] == '.') &&
551 (de->d_name[1] == 0 ||
552 !strcmp(de->d_name + 1, ".") ||
553 !strcmp(de->d_name + 1, "git")))
554 continue;
555 len = strlen(de->d_name);
556 /* Ignore overly long pathnames! */
557 if (len + baselen + 8 > sizeof(fullname))
558 continue;
559 memcpy(fullname + baselen, de->d_name, len+1);
560 if (simplify_away(fullname, baselen + len, simplify))
561 continue;
562
563 exclude = excluded(dir, fullname);
564 if (exclude && dir->collect_ignored
565 && in_pathspec(fullname, baselen + len, simplify))
566 dir_add_ignored(dir, fullname, baselen + len);
567
568 /*
569 * Excluded? If we don't explicitly want to show
570 * ignored files, ignore it
571 */
572 if (exclude && !dir->show_ignored)
573 continue;
574
575 dtype = get_dtype(de, fullname);
576
577 /*
578 * Do we want to see just the ignored files?
579 * We still need to recurse into directories,
580 * even if we don't ignore them, since the
581 * directory may contain files that we do..
582 */
583 if (!exclude && dir->show_ignored) {
584 if (dtype != DT_DIR)
585 continue;
586 }
587
588 switch (dtype) {
589 default:
590 continue;
591 case DT_DIR:
592 memcpy(fullname + baselen + len, "/", 2);
593 len++;
594 switch (treat_directory(dir, fullname, baselen + len, simplify)) {
595 case show_directory:
596 if (exclude != dir->show_ignored)
597 continue;
598 break;
599 case recurse_into_directory:
600 contents += read_directory_recursive(dir,
601 fullname, fullname, baselen + len, 0, simplify);
602 continue;
603 case ignore_directory:
604 continue;
605 }
606 break;
607 case DT_REG:
608 case DT_LNK:
609 break;
610 }
611 contents++;
612 if (check_only)
613 goto exit_early;
614 else
615 dir_add_name(dir, fullname, baselen + len);
616 }
617 exit_early:
618 closedir(fdir);
619 }
620
621 return contents;
622 }
623
624 static int cmp_name(const void *p1, const void *p2)
625 {
626 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
627 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
628
629 return cache_name_compare(e1->name, e1->len,
630 e2->name, e2->len);
631 }
632
633 /*
634 * Return the length of the "simple" part of a path match limiter.
635 */
636 static int simple_length(const char *match)
637 {
638 const char special[256] = {
639 [0] = 1, ['?'] = 1,
640 ['\\'] = 1, ['*'] = 1,
641 ['['] = 1
642 };
643 int len = -1;
644
645 for (;;) {
646 unsigned char c = *match++;
647 len++;
648 if (special[c])
649 return len;
650 }
651 }
652
653 static struct path_simplify *create_simplify(const char **pathspec)
654 {
655 int nr, alloc = 0;
656 struct path_simplify *simplify = NULL;
657
658 if (!pathspec)
659 return NULL;
660
661 for (nr = 0 ; ; nr++) {
662 const char *match;
663 if (nr >= alloc) {
664 alloc = alloc_nr(alloc);
665 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
666 }
667 match = *pathspec++;
668 if (!match)
669 break;
670 simplify[nr].path = match;
671 simplify[nr].len = simple_length(match);
672 }
673 simplify[nr].path = NULL;
674 simplify[nr].len = 0;
675 return simplify;
676 }
677
678 static void free_simplify(struct path_simplify *simplify)
679 {
680 if (simplify)
681 free(simplify);
682 }
683
684 int read_directory(struct dir_struct *dir, const char *path, const char *base, int baselen, const char **pathspec)
685 {
686 struct path_simplify *simplify = create_simplify(pathspec);
687
688 read_directory_recursive(dir, path, base, baselen, 0, simplify);
689 free_simplify(simplify);
690 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
691 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
692 return dir->nr;
693 }
694
695 int file_exists(const char *f)
696 {
697 struct stat sb;
698 return lstat(f, &sb) == 0;
699 }
700
701 /*
702 * get_relative_cwd() gets the prefix of the current working directory
703 * relative to 'dir'. If we are not inside 'dir', it returns NULL.
704 *
705 * As a convenience, it also returns NULL if 'dir' is already NULL. The
706 * reason for this behaviour is that it is natural for functions returning
707 * directory names to return NULL to say "this directory does not exist"
708 * or "this directory is invalid". These cases are usually handled the
709 * same as if the cwd is not inside 'dir' at all, so get_relative_cwd()
710 * returns NULL for both of them.
711 *
712 * Most notably, get_relative_cwd(buffer, size, get_git_work_tree())
713 * unifies the handling of "outside work tree" with "no work tree at all".
714 */
715 char *get_relative_cwd(char *buffer, int size, const char *dir)
716 {
717 char *cwd = buffer;
718
719 if (!dir)
720 return NULL;
721 if (!getcwd(buffer, size))
722 die("can't find the current directory: %s", strerror(errno));
723
724 if (!is_absolute_path(dir))
725 dir = make_absolute_path(dir);
726
727 while (*dir && *dir == *cwd) {
728 dir++;
729 cwd++;
730 }
731 if (*dir)
732 return NULL;
733 if (*cwd == '/')
734 return cwd + 1;
735 return cwd;
736 }
737
738 int is_inside_dir(const char *dir)
739 {
740 char buffer[PATH_MAX];
741 return get_relative_cwd(buffer, sizeof(buffer), dir) != NULL;
742 }
743
744 int remove_dir_recursively(struct strbuf *path, int only_empty)
745 {
746 DIR *dir = opendir(path->buf);
747 struct dirent *e;
748 int ret = 0, original_len = path->len, len;
749
750 if (!dir)
751 return -1;
752 if (path->buf[original_len - 1] != '/')
753 strbuf_addch(path, '/');
754
755 len = path->len;
756 while ((e = readdir(dir)) != NULL) {
757 struct stat st;
758 if ((e->d_name[0] == '.') &&
759 ((e->d_name[1] == 0) ||
760 ((e->d_name[1] == '.') && e->d_name[2] == 0)))
761 continue; /* "." and ".." */
762
763 strbuf_setlen(path, len);
764 strbuf_addstr(path, e->d_name);
765 if (lstat(path->buf, &st))
766 ; /* fall thru */
767 else if (S_ISDIR(st.st_mode)) {
768 if (!remove_dir_recursively(path, only_empty))
769 continue; /* happy */
770 } else if (!only_empty && !unlink(path->buf))
771 continue; /* happy, too */
772
773 /* path too long, stat fails, or non-directory still exists */
774 ret = -1;
775 break;
776 }
777 closedir(dir);
778
779 strbuf_setlen(path, original_len);
780 if (!ret)
781 ret = rmdir(path->buf);
782 return ret;
783 }
784
785 void setup_standard_excludes(struct dir_struct *dir)
786 {
787 const char *path;
788
789 dir->exclude_per_dir = ".gitignore";
790 path = git_path("info/exclude");
791 if (!access(path, R_OK))
792 add_excludes_from_file(dir, path);
793 if (excludes_file && !access(excludes_file, R_OK))
794 add_excludes_from_file(dir, excludes_file);
795 }