Build-in send-pack, with an API for other programs to call.
[git/jnareb-git.git] / dir.c
blob5bcc764c97dca5e584d0a4227bfb299efb5699ce
1 /*
2 * This handles recursive filename detection with exclude
3 * files, index knowledge etc..
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"
12 struct path_simplify {
13 int len;
14 const char *path;
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);
21 int common_prefix(const char **pathspec)
23 const char *path, *slash, *next;
24 int prefix;
26 if (!pathspec)
27 return 0;
29 path = *pathspec;
30 slash = strrchr(path, '/');
31 if (!slash)
32 return 0;
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;
51 return prefix;
55 * Does 'match' matches the given name?
56 * A match is found if
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'.
62 * and the return value tells which case it was.
64 * It returns 0 when there is no match.
66 static int match_one(const char *match, const char *name, int namelen)
68 int matchlen;
70 /* If the match was just the prefix, we matched */
71 matchlen = strlen(match);
72 if (!matchlen)
73 return MATCHED_RECURSIVELY;
76 * If we don't match the matchstring exactly,
77 * we need to match by fnmatch
79 if (strncmp(match, name, matchlen))
80 return !fnmatch(match, name, 0) ? MATCHED_FNMATCH : 0;
82 if (!name[matchlen])
83 return MATCHED_EXACTLY;
84 if (match[matchlen-1] == '/' || name[matchlen] == '/')
85 return MATCHED_RECURSIVELY;
86 return 0;
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.
97 int match_pathspec(const char **pathspec, const char *name, int namelen, int prefix, char *seen)
99 int retval;
100 const char *match;
102 name += prefix;
103 namelen -= prefix;
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;
118 return retval;
121 static int no_wildcard(const char *string)
123 return string[strcspn(string, "*?[{")] == '\0';
126 void add_exclude(const char *string, const char *base,
127 int baselen, struct exclude_list *which)
129 struct exclude *x = xmalloc(sizeof (*x));
131 x->to_exclude = 1;
132 if (*string == '!') {
133 x->to_exclude = 0;
134 string++;
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 if (which->nr == which->alloc) {
148 which->alloc = alloc_nr(which->alloc);
149 which->excludes = xrealloc(which->excludes,
150 which->alloc * sizeof(x));
152 which->excludes[which->nr++] = x;
155 static int add_excludes_from_file_1(const char *fname,
156 const char *base,
157 int baselen,
158 struct exclude_list *which)
160 struct stat st;
161 int fd, i;
162 size_t size;
163 char *buf, *entry;
165 fd = open(fname, O_RDONLY);
166 if (fd < 0 || fstat(fd, &st) < 0)
167 goto err;
168 size = xsize_t(st.st_size);
169 if (size == 0) {
170 close(fd);
171 return 0;
173 buf = xmalloc(size+1);
174 if (read_in_full(fd, buf, size) != size)
175 goto err;
176 close(fd);
178 buf[size++] = '\n';
179 entry = buf;
180 for (i = 0; i < size; i++) {
181 if (buf[i] == '\n') {
182 if (entry != buf + i && entry[0] != '#') {
183 buf[i - (i && buf[i-1] == '\r')] = 0;
184 add_exclude(entry, base, baselen, which);
186 entry = buf + i + 1;
189 return 0;
191 err:
192 if (0 <= fd)
193 close(fd);
194 return -1;
197 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
199 if (add_excludes_from_file_1(fname, "", 0,
200 &dir->exclude_list[EXC_FILE]) < 0)
201 die("cannot use %s as an exclude file", fname);
204 int push_exclude_per_directory(struct dir_struct *dir, const char *base, int baselen)
206 char exclude_file[PATH_MAX];
207 struct exclude_list *el = &dir->exclude_list[EXC_DIRS];
208 int current_nr = el->nr;
210 if (dir->exclude_per_dir) {
211 memcpy(exclude_file, base, baselen);
212 strcpy(exclude_file + baselen, dir->exclude_per_dir);
213 add_excludes_from_file_1(exclude_file, base, baselen, el);
215 return current_nr;
218 void pop_exclude_per_directory(struct dir_struct *dir, int stk)
220 struct exclude_list *el = &dir->exclude_list[EXC_DIRS];
222 while (stk < el->nr)
223 free(el->excludes[--el->nr]);
226 /* Scan the list and let the last match determines the fate.
227 * Return 1 for exclude, 0 for include and -1 for undecided.
229 static int excluded_1(const char *pathname,
230 int pathlen, const char *basename,
231 struct exclude_list *el)
233 int i;
235 if (el->nr) {
236 for (i = el->nr - 1; 0 <= i; i--) {
237 struct exclude *x = el->excludes[i];
238 const char *exclude = x->pattern;
239 int to_exclude = x->to_exclude;
241 if (x->flags & EXC_FLAG_NODIR) {
242 /* match basename */
243 if (x->flags & EXC_FLAG_NOWILDCARD) {
244 if (!strcmp(exclude, basename))
245 return to_exclude;
246 } else if (x->flags & EXC_FLAG_ENDSWITH) {
247 if (x->patternlen - 1 <= pathlen &&
248 !strcmp(exclude + 1, pathname + pathlen - x->patternlen + 1))
249 return to_exclude;
250 } else {
251 if (fnmatch(exclude, basename, 0) == 0)
252 return to_exclude;
255 else {
256 /* match with FNM_PATHNAME:
257 * exclude has base (baselen long) implicitly
258 * in front of it.
260 int baselen = x->baselen;
261 if (*exclude == '/')
262 exclude++;
264 if (pathlen < baselen ||
265 (baselen && pathname[baselen-1] != '/') ||
266 strncmp(pathname, x->base, baselen))
267 continue;
269 if (x->flags & EXC_FLAG_NOWILDCARD) {
270 if (!strcmp(exclude, pathname + baselen))
271 return to_exclude;
272 } else {
273 if (fnmatch(exclude, pathname+baselen,
274 FNM_PATHNAME) == 0)
275 return to_exclude;
280 return -1; /* undecided */
283 int excluded(struct dir_struct *dir, const char *pathname)
285 int pathlen = strlen(pathname);
286 int st;
287 const char *basename = strrchr(pathname, '/');
288 basename = (basename) ? basename+1 : pathname;
290 for (st = EXC_CMDL; st <= EXC_FILE; st++) {
291 switch (excluded_1(pathname, pathlen, basename, &dir->exclude_list[st])) {
292 case 0:
293 return 0;
294 case 1:
295 return 1;
298 return 0;
301 static struct dir_entry *dir_entry_new(const char *pathname, int len) {
302 struct dir_entry *ent;
304 ent = xmalloc(sizeof(*ent) + len + 1);
305 ent->len = len;
306 memcpy(ent->name, pathname, len);
307 ent->name[len] = 0;
308 return ent;
311 struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
313 if (cache_name_pos(pathname, len) >= 0)
314 return NULL;
316 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
317 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
320 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
322 if (cache_name_pos(pathname, len) >= 0)
323 return NULL;
325 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
326 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
329 enum exist_status {
330 index_nonexistent = 0,
331 index_directory,
332 index_gitdir,
336 * The index sorts alphabetically by entry name, which
337 * means that a gitlink sorts as '\0' at the end, while
338 * a directory (which is defined not as an entry, but as
339 * the files it contains) will sort with the '/' at the
340 * end.
342 static enum exist_status directory_exists_in_index(const char *dirname, int len)
344 int pos = cache_name_pos(dirname, len);
345 if (pos < 0)
346 pos = -pos-1;
347 while (pos < active_nr) {
348 struct cache_entry *ce = active_cache[pos++];
349 unsigned char endchar;
351 if (strncmp(ce->name, dirname, len))
352 break;
353 endchar = ce->name[len];
354 if (endchar > '/')
355 break;
356 if (endchar == '/')
357 return index_directory;
358 if (!endchar && S_ISGITLINK(ntohl(ce->ce_mode)))
359 return index_gitdir;
361 return index_nonexistent;
365 * When we find a directory when traversing the filesystem, we
366 * have three distinct cases:
368 * - ignore it
369 * - see it as a directory
370 * - recurse into it
372 * and which one we choose depends on a combination of existing
373 * git index contents and the flags passed into the directory
374 * traversal routine.
376 * Case 1: If we *already* have entries in the index under that
377 * directory name, we always recurse into the directory to see
378 * all the files.
380 * Case 2: If we *already* have that directory name as a gitlink,
381 * we always continue to see it as a gitlink, regardless of whether
382 * there is an actual git directory there or not (it might not
383 * be checked out as a subproject!)
385 * Case 3: if we didn't have it in the index previously, we
386 * have a few sub-cases:
388 * (a) if "show_other_directories" is true, we show it as
389 * just a directory, unless "hide_empty_directories" is
390 * also true and the directory is empty, in which case
391 * we just ignore it entirely.
392 * (b) if it looks like a git directory, and we don't have
393 * 'no_gitlinks' set we treat it as a gitlink, and show it
394 * as a directory.
395 * (c) otherwise, we recurse into it.
397 enum directory_treatment {
398 show_directory,
399 ignore_directory,
400 recurse_into_directory,
403 static enum directory_treatment treat_directory(struct dir_struct *dir,
404 const char *dirname, int len,
405 const struct path_simplify *simplify)
407 /* The "len-1" is to strip the final '/' */
408 switch (directory_exists_in_index(dirname, len-1)) {
409 case index_directory:
410 return recurse_into_directory;
412 case index_gitdir:
413 if (dir->show_other_directories)
414 return ignore_directory;
415 return show_directory;
417 case index_nonexistent:
418 if (dir->show_other_directories)
419 break;
420 if (!dir->no_gitlinks) {
421 unsigned char sha1[20];
422 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
423 return show_directory;
425 return recurse_into_directory;
428 /* This is the "show_other_directories" case */
429 if (!dir->hide_empty_directories)
430 return show_directory;
431 if (!read_directory_recursive(dir, dirname, dirname, len, 1, simplify))
432 return ignore_directory;
433 return show_directory;
437 * This is an inexact early pruning of any recursive directory
438 * reading - if the path cannot possibly be in the pathspec,
439 * return true, and we'll skip it early.
441 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
443 if (simplify) {
444 for (;;) {
445 const char *match = simplify->path;
446 int len = simplify->len;
448 if (!match)
449 break;
450 if (len > pathlen)
451 len = pathlen;
452 if (!memcmp(path, match, len))
453 return 0;
454 simplify++;
456 return 1;
458 return 0;
461 static int in_pathspec(const char *path, int len, const struct path_simplify *simplify)
463 if (simplify) {
464 for (; simplify->path; simplify++) {
465 if (len == simplify->len
466 && !memcmp(path, simplify->path, len))
467 return 1;
470 return 0;
473 static int get_dtype(struct dirent *de, const char *path)
475 int dtype = DTYPE(de);
476 struct stat st;
478 if (dtype != DT_UNKNOWN)
479 return dtype;
480 if (lstat(path, &st))
481 return dtype;
482 if (S_ISREG(st.st_mode))
483 return DT_REG;
484 if (S_ISDIR(st.st_mode))
485 return DT_DIR;
486 if (S_ISLNK(st.st_mode))
487 return DT_LNK;
488 return dtype;
492 * Read a directory tree. We currently ignore anything but
493 * directories, regular files and symlinks. That's because git
494 * doesn't handle them at all yet. Maybe that will change some
495 * day.
497 * Also, we ignore the name ".git" (even if it is not a directory).
498 * That likely will not change.
500 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)
502 DIR *fdir = opendir(path);
503 int contents = 0;
505 if (fdir) {
506 int exclude_stk;
507 struct dirent *de;
508 char fullname[PATH_MAX + 1];
509 memcpy(fullname, base, baselen);
511 exclude_stk = push_exclude_per_directory(dir, base, baselen);
513 while ((de = readdir(fdir)) != NULL) {
514 int len, dtype;
515 int exclude;
517 if ((de->d_name[0] == '.') &&
518 (de->d_name[1] == 0 ||
519 !strcmp(de->d_name + 1, ".") ||
520 !strcmp(de->d_name + 1, "git")))
521 continue;
522 len = strlen(de->d_name);
523 /* Ignore overly long pathnames! */
524 if (len + baselen + 8 > sizeof(fullname))
525 continue;
526 memcpy(fullname + baselen, de->d_name, len+1);
527 if (simplify_away(fullname, baselen + len, simplify))
528 continue;
530 exclude = excluded(dir, fullname);
531 if (exclude && dir->collect_ignored
532 && in_pathspec(fullname, baselen + len, simplify))
533 dir_add_ignored(dir, fullname, baselen + len);
536 * Excluded? If we don't explicitly want to show
537 * ignored files, ignore it
539 if (exclude && !dir->show_ignored)
540 continue;
542 dtype = get_dtype(de, fullname);
545 * Do we want to see just the ignored files?
546 * We still need to recurse into directories,
547 * even if we don't ignore them, since the
548 * directory may contain files that we do..
550 if (!exclude && dir->show_ignored) {
551 if (dtype != DT_DIR)
552 continue;
555 switch (dtype) {
556 default:
557 continue;
558 case DT_DIR:
559 memcpy(fullname + baselen + len, "/", 2);
560 len++;
561 switch (treat_directory(dir, fullname, baselen + len, simplify)) {
562 case show_directory:
563 if (exclude != dir->show_ignored)
564 continue;
565 break;
566 case recurse_into_directory:
567 contents += read_directory_recursive(dir,
568 fullname, fullname, baselen + len, 0, simplify);
569 continue;
570 case ignore_directory:
571 continue;
573 break;
574 case DT_REG:
575 case DT_LNK:
576 break;
578 contents++;
579 if (check_only)
580 goto exit_early;
581 else
582 dir_add_name(dir, fullname, baselen + len);
584 exit_early:
585 closedir(fdir);
587 pop_exclude_per_directory(dir, exclude_stk);
590 return contents;
593 static int cmp_name(const void *p1, const void *p2)
595 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
596 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
598 return cache_name_compare(e1->name, e1->len,
599 e2->name, e2->len);
603 * Return the length of the "simple" part of a path match limiter.
605 static int simple_length(const char *match)
607 const char special[256] = {
608 [0] = 1, ['?'] = 1,
609 ['\\'] = 1, ['*'] = 1,
610 ['['] = 1
612 int len = -1;
614 for (;;) {
615 unsigned char c = *match++;
616 len++;
617 if (special[c])
618 return len;
622 static struct path_simplify *create_simplify(const char **pathspec)
624 int nr, alloc = 0;
625 struct path_simplify *simplify = NULL;
627 if (!pathspec)
628 return NULL;
630 for (nr = 0 ; ; nr++) {
631 const char *match;
632 if (nr >= alloc) {
633 alloc = alloc_nr(alloc);
634 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
636 match = *pathspec++;
637 if (!match)
638 break;
639 simplify[nr].path = match;
640 simplify[nr].len = simple_length(match);
642 simplify[nr].path = NULL;
643 simplify[nr].len = 0;
644 return simplify;
647 static void free_simplify(struct path_simplify *simplify)
649 if (simplify)
650 free(simplify);
653 int read_directory(struct dir_struct *dir, const char *path, const char *base, int baselen, const char **pathspec)
655 struct path_simplify *simplify = create_simplify(pathspec);
658 * Make sure to do the per-directory exclude for all the
659 * directories leading up to our base.
661 if (baselen) {
662 if (dir->exclude_per_dir) {
663 char *p, *pp = xmalloc(baselen+1);
664 memcpy(pp, base, baselen+1);
665 p = pp;
666 while (1) {
667 char save = *p;
668 *p = 0;
669 push_exclude_per_directory(dir, pp, p-pp);
670 *p++ = save;
671 if (!save)
672 break;
673 p = strchr(p, '/');
674 if (p)
675 p++;
676 else
677 p = pp + baselen;
679 free(pp);
683 read_directory_recursive(dir, path, base, baselen, 0, simplify);
684 free_simplify(simplify);
685 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
686 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
687 return dir->nr;
691 file_exists(const char *f)
693 struct stat sb;
694 return stat(f, &sb) == 0;
698 * get_relative_cwd() gets the prefix of the current working directory
699 * relative to 'dir'. If we are not inside 'dir', it returns NULL.
701 * As a convenience, it also returns NULL if 'dir' is already NULL. The
702 * reason for this behaviour is that it is natural for functions returning
703 * directory names to return NULL to say "this directory does not exist"
704 * or "this directory is invalid". These cases are usually handled the
705 * same as if the cwd is not inside 'dir' at all, so get_relative_cwd()
706 * returns NULL for both of them.
708 * Most notably, get_relative_cwd(buffer, size, get_git_work_tree())
709 * unifies the handling of "outside work tree" with "no work tree at all".
711 char *get_relative_cwd(char *buffer, int size, const char *dir)
713 char *cwd = buffer;
715 if (!dir)
716 return NULL;
717 if (!getcwd(buffer, size))
718 die("can't find the current directory: %s", strerror(errno));
720 if (!is_absolute_path(dir))
721 dir = make_absolute_path(dir);
723 while (*dir && *dir == *cwd) {
724 dir++;
725 cwd++;
727 if (*dir)
728 return NULL;
729 if (*cwd == '/')
730 return cwd + 1;
731 return cwd;
734 int is_inside_dir(const char *dir)
736 char buffer[PATH_MAX];
737 return get_relative_cwd(buffer, sizeof(buffer), dir) != NULL;
740 int remove_dir_recursively(struct strbuf *path, int only_empty)
742 DIR *dir = opendir(path->buf);
743 struct dirent *e;
744 int ret = 0, original_len = path->len, len;
746 if (!dir)
747 return -1;
748 if (path->buf[original_len - 1] != '/')
749 strbuf_addch(path, '/');
751 len = path->len;
752 while ((e = readdir(dir)) != NULL) {
753 struct stat st;
754 if ((e->d_name[0] == '.') &&
755 ((e->d_name[1] == 0) ||
756 ((e->d_name[1] == '.') && e->d_name[2] == 0)))
757 continue; /* "." and ".." */
759 strbuf_setlen(path, len);
760 strbuf_addstr(path, e->d_name);
761 if (lstat(path->buf, &st))
762 ; /* fall thru */
763 else if (S_ISDIR(st.st_mode)) {
764 if (!remove_dir_recursively(path, only_empty))
765 continue; /* happy */
766 } else if (!only_empty && !unlink(path->buf))
767 continue; /* happy, too */
769 /* path too long, stat fails, or non-directory still exists */
770 ret = -1;
771 break;
773 closedir(dir);
775 strbuf_setlen(path, original_len);
776 if (!ret)
777 ret = rmdir(path->buf);
778 return ret;