builtin-apply: stronger indent-with-on-tab fixing
[git/jbf-maint.git] / dir.c
blob6b3273d1d18298dc6cb8c08b42948af05c775221
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 ALLOC_GROW(which->excludes, which->nr + 1, which->alloc);
148 which->excludes[which->nr++] = x;
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)
157 struct stat st;
158 int fd, i;
159 size_t size;
160 char *buf, *entry;
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;
170 buf = xmalloc(size+1);
171 if (read_in_full(fd, buf, size) != size)
172 goto err;
173 close(fd);
175 if (buf_p)
176 *buf_p = buf;
177 buf[size++] = '\n';
178 entry = buf;
179 for (i = 0; i < size; i++) {
180 if (buf[i] == '\n') {
181 if (entry != buf + i && entry[0] != '#') {
182 buf[i - (i && buf[i-1] == '\r')] = 0;
183 add_exclude(entry, base, baselen, which);
185 entry = buf + i + 1;
188 return 0;
190 err:
191 if (0 <= fd)
192 close(fd);
193 return -1;
196 void add_excludes_from_file(struct dir_struct *dir, const char *fname)
198 if (add_excludes_from_file_1(fname, "", 0, NULL,
199 &dir->exclude_list[EXC_FILE]) < 0)
200 die("cannot use %s as an exclude file", fname);
203 static void prep_exclude(struct dir_struct *dir, const char *base, int baselen)
205 struct exclude_list *el;
206 struct exclude_stack *stk = NULL;
207 int current;
209 if ((!dir->exclude_per_dir) ||
210 (baselen + strlen(dir->exclude_per_dir) >= PATH_MAX))
211 return; /* too long a path -- ignore */
213 /* Pop the ones that are not the prefix of the path being checked. */
214 el = &dir->exclude_list[EXC_DIRS];
215 while ((stk = dir->exclude_stack) != NULL) {
216 if (stk->baselen <= baselen &&
217 !strncmp(dir->basebuf, base, stk->baselen))
218 break;
219 dir->exclude_stack = stk->prev;
220 while (stk->exclude_ix < el->nr)
221 free(el->excludes[--el->nr]);
222 free(stk->filebuf);
223 free(stk);
226 /* Read from the parent directories and push them down. */
227 current = stk ? stk->baselen : -1;
228 while (current < baselen) {
229 struct exclude_stack *stk = xcalloc(1, sizeof(*stk));
230 const char *cp;
232 if (current < 0) {
233 cp = base;
234 current = 0;
236 else {
237 cp = strchr(base + current + 1, '/');
238 if (!cp)
239 die("oops in prep_exclude");
240 cp++;
242 stk->prev = dir->exclude_stack;
243 stk->baselen = cp - base;
244 stk->exclude_ix = el->nr;
245 memcpy(dir->basebuf + current, base + current,
246 stk->baselen - current);
247 strcpy(dir->basebuf + stk->baselen, dir->exclude_per_dir);
248 add_excludes_from_file_1(dir->basebuf,
249 dir->basebuf, stk->baselen,
250 &stk->filebuf, el);
251 dir->exclude_stack = stk;
252 current = stk->baselen;
254 dir->basebuf[baselen] = '\0';
257 /* Scan the list and let the last match determines the fate.
258 * Return 1 for exclude, 0 for include and -1 for undecided.
260 static int excluded_1(const char *pathname,
261 int pathlen, const char *basename,
262 struct exclude_list *el)
264 int i;
266 if (el->nr) {
267 for (i = el->nr - 1; 0 <= i; i--) {
268 struct exclude *x = el->excludes[i];
269 const char *exclude = x->pattern;
270 int to_exclude = x->to_exclude;
272 if (x->flags & EXC_FLAG_NODIR) {
273 /* match basename */
274 if (x->flags & EXC_FLAG_NOWILDCARD) {
275 if (!strcmp(exclude, basename))
276 return to_exclude;
277 } else if (x->flags & EXC_FLAG_ENDSWITH) {
278 if (x->patternlen - 1 <= pathlen &&
279 !strcmp(exclude + 1, pathname + pathlen - x->patternlen + 1))
280 return to_exclude;
281 } else {
282 if (fnmatch(exclude, basename, 0) == 0)
283 return to_exclude;
286 else {
287 /* match with FNM_PATHNAME:
288 * exclude has base (baselen long) implicitly
289 * in front of it.
291 int baselen = x->baselen;
292 if (*exclude == '/')
293 exclude++;
295 if (pathlen < baselen ||
296 (baselen && pathname[baselen-1] != '/') ||
297 strncmp(pathname, x->base, baselen))
298 continue;
300 if (x->flags & EXC_FLAG_NOWILDCARD) {
301 if (!strcmp(exclude, pathname + baselen))
302 return to_exclude;
303 } else {
304 if (fnmatch(exclude, pathname+baselen,
305 FNM_PATHNAME) == 0)
306 return to_exclude;
311 return -1; /* undecided */
314 int excluded(struct dir_struct *dir, const char *pathname)
316 int pathlen = strlen(pathname);
317 int st;
318 const char *basename = strrchr(pathname, '/');
319 basename = (basename) ? basename+1 : pathname;
321 prep_exclude(dir, pathname, basename-pathname);
322 for (st = EXC_CMDL; st <= EXC_FILE; st++) {
323 switch (excluded_1(pathname, pathlen, basename, &dir->exclude_list[st])) {
324 case 0:
325 return 0;
326 case 1:
327 return 1;
330 return 0;
333 static struct dir_entry *dir_entry_new(const char *pathname, int len)
335 struct dir_entry *ent;
337 ent = xmalloc(sizeof(*ent) + len + 1);
338 ent->len = len;
339 memcpy(ent->name, pathname, len);
340 ent->name[len] = 0;
341 return ent;
344 struct dir_entry *dir_add_name(struct dir_struct *dir, const char *pathname, int len)
346 if (cache_name_pos(pathname, len) >= 0)
347 return NULL;
349 ALLOC_GROW(dir->entries, dir->nr+1, dir->alloc);
350 return dir->entries[dir->nr++] = dir_entry_new(pathname, len);
353 struct dir_entry *dir_add_ignored(struct dir_struct *dir, const char *pathname, int len)
355 if (cache_name_pos(pathname, len) >= 0)
356 return NULL;
358 ALLOC_GROW(dir->ignored, dir->ignored_nr+1, dir->ignored_alloc);
359 return dir->ignored[dir->ignored_nr++] = dir_entry_new(pathname, len);
362 enum exist_status {
363 index_nonexistent = 0,
364 index_directory,
365 index_gitdir,
369 * The index sorts alphabetically by entry name, which
370 * means that a gitlink sorts as '\0' at the end, while
371 * a directory (which is defined not as an entry, but as
372 * the files it contains) will sort with the '/' at the
373 * end.
375 static enum exist_status directory_exists_in_index(const char *dirname, int len)
377 int pos = cache_name_pos(dirname, len);
378 if (pos < 0)
379 pos = -pos-1;
380 while (pos < active_nr) {
381 struct cache_entry *ce = active_cache[pos++];
382 unsigned char endchar;
384 if (strncmp(ce->name, dirname, len))
385 break;
386 endchar = ce->name[len];
387 if (endchar > '/')
388 break;
389 if (endchar == '/')
390 return index_directory;
391 if (!endchar && S_ISGITLINK(ntohl(ce->ce_mode)))
392 return index_gitdir;
394 return index_nonexistent;
398 * When we find a directory when traversing the filesystem, we
399 * have three distinct cases:
401 * - ignore it
402 * - see it as a directory
403 * - recurse into it
405 * and which one we choose depends on a combination of existing
406 * git index contents and the flags passed into the directory
407 * traversal routine.
409 * Case 1: If we *already* have entries in the index under that
410 * directory name, we always recurse into the directory to see
411 * all the files.
413 * Case 2: If we *already* have that directory name as a gitlink,
414 * we always continue to see it as a gitlink, regardless of whether
415 * there is an actual git directory there or not (it might not
416 * be checked out as a subproject!)
418 * Case 3: if we didn't have it in the index previously, we
419 * have a few sub-cases:
421 * (a) if "show_other_directories" is true, we show it as
422 * just a directory, unless "hide_empty_directories" is
423 * also true and the directory is empty, in which case
424 * we just ignore it entirely.
425 * (b) if it looks like a git directory, and we don't have
426 * 'no_gitlinks' set we treat it as a gitlink, and show it
427 * as a directory.
428 * (c) otherwise, we recurse into it.
430 enum directory_treatment {
431 show_directory,
432 ignore_directory,
433 recurse_into_directory,
436 static enum directory_treatment treat_directory(struct dir_struct *dir,
437 const char *dirname, int len,
438 const struct path_simplify *simplify)
440 /* The "len-1" is to strip the final '/' */
441 switch (directory_exists_in_index(dirname, len-1)) {
442 case index_directory:
443 return recurse_into_directory;
445 case index_gitdir:
446 if (dir->show_other_directories)
447 return ignore_directory;
448 return show_directory;
450 case index_nonexistent:
451 if (dir->show_other_directories)
452 break;
453 if (!dir->no_gitlinks) {
454 unsigned char sha1[20];
455 if (resolve_gitlink_ref(dirname, "HEAD", sha1) == 0)
456 return show_directory;
458 return recurse_into_directory;
461 /* This is the "show_other_directories" case */
462 if (!dir->hide_empty_directories)
463 return show_directory;
464 if (!read_directory_recursive(dir, dirname, dirname, len, 1, simplify))
465 return ignore_directory;
466 return show_directory;
470 * This is an inexact early pruning of any recursive directory
471 * reading - if the path cannot possibly be in the pathspec,
472 * return true, and we'll skip it early.
474 static int simplify_away(const char *path, int pathlen, const struct path_simplify *simplify)
476 if (simplify) {
477 for (;;) {
478 const char *match = simplify->path;
479 int len = simplify->len;
481 if (!match)
482 break;
483 if (len > pathlen)
484 len = pathlen;
485 if (!memcmp(path, match, len))
486 return 0;
487 simplify++;
489 return 1;
491 return 0;
494 static int in_pathspec(const char *path, int len, const struct path_simplify *simplify)
496 if (simplify) {
497 for (; simplify->path; simplify++) {
498 if (len == simplify->len
499 && !memcmp(path, simplify->path, len))
500 return 1;
503 return 0;
506 static int get_dtype(struct dirent *de, const char *path)
508 int dtype = DTYPE(de);
509 struct stat st;
511 if (dtype != DT_UNKNOWN)
512 return dtype;
513 if (lstat(path, &st))
514 return dtype;
515 if (S_ISREG(st.st_mode))
516 return DT_REG;
517 if (S_ISDIR(st.st_mode))
518 return DT_DIR;
519 if (S_ISLNK(st.st_mode))
520 return DT_LNK;
521 return dtype;
525 * Read a directory tree. We currently ignore anything but
526 * directories, regular files and symlinks. That's because git
527 * doesn't handle them at all yet. Maybe that will change some
528 * day.
530 * Also, we ignore the name ".git" (even if it is not a directory).
531 * That likely will not change.
533 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)
535 DIR *fdir = opendir(path);
536 int contents = 0;
538 if (fdir) {
539 struct dirent *de;
540 char fullname[PATH_MAX + 1];
541 memcpy(fullname, base, baselen);
543 while ((de = readdir(fdir)) != NULL) {
544 int len, dtype;
545 int exclude;
547 if ((de->d_name[0] == '.') &&
548 (de->d_name[1] == 0 ||
549 !strcmp(de->d_name + 1, ".") ||
550 !strcmp(de->d_name + 1, "git")))
551 continue;
552 len = strlen(de->d_name);
553 /* Ignore overly long pathnames! */
554 if (len + baselen + 8 > sizeof(fullname))
555 continue;
556 memcpy(fullname + baselen, de->d_name, len+1);
557 if (simplify_away(fullname, baselen + len, simplify))
558 continue;
560 exclude = excluded(dir, fullname);
561 if (exclude && dir->collect_ignored
562 && in_pathspec(fullname, baselen + len, simplify))
563 dir_add_ignored(dir, fullname, baselen + len);
566 * Excluded? If we don't explicitly want to show
567 * ignored files, ignore it
569 if (exclude && !dir->show_ignored)
570 continue;
572 dtype = get_dtype(de, fullname);
575 * Do we want to see just the ignored files?
576 * We still need to recurse into directories,
577 * even if we don't ignore them, since the
578 * directory may contain files that we do..
580 if (!exclude && dir->show_ignored) {
581 if (dtype != DT_DIR)
582 continue;
585 switch (dtype) {
586 default:
587 continue;
588 case DT_DIR:
589 memcpy(fullname + baselen + len, "/", 2);
590 len++;
591 switch (treat_directory(dir, fullname, baselen + len, simplify)) {
592 case show_directory:
593 if (exclude != dir->show_ignored)
594 continue;
595 break;
596 case recurse_into_directory:
597 contents += read_directory_recursive(dir,
598 fullname, fullname, baselen + len, 0, simplify);
599 continue;
600 case ignore_directory:
601 continue;
603 break;
604 case DT_REG:
605 case DT_LNK:
606 break;
608 contents++;
609 if (check_only)
610 goto exit_early;
611 else
612 dir_add_name(dir, fullname, baselen + len);
614 exit_early:
615 closedir(fdir);
618 return contents;
621 static int cmp_name(const void *p1, const void *p2)
623 const struct dir_entry *e1 = *(const struct dir_entry **)p1;
624 const struct dir_entry *e2 = *(const struct dir_entry **)p2;
626 return cache_name_compare(e1->name, e1->len,
627 e2->name, e2->len);
631 * Return the length of the "simple" part of a path match limiter.
633 static int simple_length(const char *match)
635 const char special[256] = {
636 [0] = 1, ['?'] = 1,
637 ['\\'] = 1, ['*'] = 1,
638 ['['] = 1
640 int len = -1;
642 for (;;) {
643 unsigned char c = *match++;
644 len++;
645 if (special[c])
646 return len;
650 static struct path_simplify *create_simplify(const char **pathspec)
652 int nr, alloc = 0;
653 struct path_simplify *simplify = NULL;
655 if (!pathspec)
656 return NULL;
658 for (nr = 0 ; ; nr++) {
659 const char *match;
660 if (nr >= alloc) {
661 alloc = alloc_nr(alloc);
662 simplify = xrealloc(simplify, alloc * sizeof(*simplify));
664 match = *pathspec++;
665 if (!match)
666 break;
667 simplify[nr].path = match;
668 simplify[nr].len = simple_length(match);
670 simplify[nr].path = NULL;
671 simplify[nr].len = 0;
672 return simplify;
675 static void free_simplify(struct path_simplify *simplify)
677 if (simplify)
678 free(simplify);
681 int read_directory(struct dir_struct *dir, const char *path, const char *base, int baselen, const char **pathspec)
683 struct path_simplify *simplify = create_simplify(pathspec);
685 read_directory_recursive(dir, path, base, baselen, 0, simplify);
686 free_simplify(simplify);
687 qsort(dir->entries, dir->nr, sizeof(struct dir_entry *), cmp_name);
688 qsort(dir->ignored, dir->ignored_nr, sizeof(struct dir_entry *), cmp_name);
689 return dir->nr;
692 int file_exists(const char *f)
694 struct stat sb;
695 return lstat(f, &sb) == 0;
699 * get_relative_cwd() gets the prefix of the current working directory
700 * relative to 'dir'. If we are not inside 'dir', it returns NULL.
702 * As a convenience, it also returns NULL if 'dir' is already NULL. The
703 * reason for this behaviour is that it is natural for functions returning
704 * directory names to return NULL to say "this directory does not exist"
705 * or "this directory is invalid". These cases are usually handled the
706 * same as if the cwd is not inside 'dir' at all, so get_relative_cwd()
707 * returns NULL for both of them.
709 * Most notably, get_relative_cwd(buffer, size, get_git_work_tree())
710 * unifies the handling of "outside work tree" with "no work tree at all".
712 char *get_relative_cwd(char *buffer, int size, const char *dir)
714 char *cwd = buffer;
716 if (!dir)
717 return NULL;
718 if (!getcwd(buffer, size))
719 die("can't find the current directory: %s", strerror(errno));
721 if (!is_absolute_path(dir))
722 dir = make_absolute_path(dir);
724 while (*dir && *dir == *cwd) {
725 dir++;
726 cwd++;
728 if (*dir)
729 return NULL;
730 if (*cwd == '/')
731 return cwd + 1;
732 return cwd;
735 int is_inside_dir(const char *dir)
737 char buffer[PATH_MAX];
738 return get_relative_cwd(buffer, sizeof(buffer), dir) != NULL;
741 int remove_dir_recursively(struct strbuf *path, int only_empty)
743 DIR *dir = opendir(path->buf);
744 struct dirent *e;
745 int ret = 0, original_len = path->len, len;
747 if (!dir)
748 return -1;
749 if (path->buf[original_len - 1] != '/')
750 strbuf_addch(path, '/');
752 len = path->len;
753 while ((e = readdir(dir)) != NULL) {
754 struct stat st;
755 if ((e->d_name[0] == '.') &&
756 ((e->d_name[1] == 0) ||
757 ((e->d_name[1] == '.') && e->d_name[2] == 0)))
758 continue; /* "." and ".." */
760 strbuf_setlen(path, len);
761 strbuf_addstr(path, e->d_name);
762 if (lstat(path->buf, &st))
763 ; /* fall thru */
764 else if (S_ISDIR(st.st_mode)) {
765 if (!remove_dir_recursively(path, only_empty))
766 continue; /* happy */
767 } else if (!only_empty && !unlink(path->buf))
768 continue; /* happy, too */
770 /* path too long, stat fails, or non-directory still exists */
771 ret = -1;
772 break;
774 closedir(dir);
776 strbuf_setlen(path, original_len);
777 if (!ret)
778 ret = rmdir(path->buf);
779 return ret;
782 void setup_standard_excludes(struct dir_struct *dir)
784 const char *path;
786 dir->exclude_per_dir = ".gitignore";
787 path = git_path("info/exclude");
788 if (!access(path, R_OK))
789 add_excludes_from_file(dir, path);
790 if (excludes_file && !access(excludes_file, R_OK))
791 add_excludes_from_file(dir, excludes_file);