dir: allow a BOM at the beginning of exclude files
[git/mingw/j6t.git] / path.c
blobe6089938018b74fb8a6ea547ffc64d428a13daa0
1 /*
2 * Utilities for paths and pathnames
3 */
4 #include "cache.h"
5 #include "strbuf.h"
6 #include "string-list.h"
8 static int get_st_mode_bits(const char *path, int *mode)
10 struct stat st;
11 if (lstat(path, &st) < 0)
12 return -1;
13 *mode = st.st_mode;
14 return 0;
17 static char bad_path[] = "/bad-path/";
19 static char *get_pathname(void)
21 static char pathname_array[4][PATH_MAX];
22 static int index;
23 return pathname_array[3 & ++index];
26 static char *cleanup_path(char *path)
28 /* Clean it up */
29 if (!memcmp(path, "./", 2)) {
30 path += 2;
31 while (*path == '/')
32 path++;
34 return path;
37 char *mksnpath(char *buf, size_t n, const char *fmt, ...)
39 va_list args;
40 unsigned len;
42 va_start(args, fmt);
43 len = vsnprintf(buf, n, fmt, args);
44 va_end(args);
45 if (len >= n) {
46 strlcpy(buf, bad_path, n);
47 return buf;
49 return cleanup_path(buf);
52 static char *vsnpath(char *buf, size_t n, const char *fmt, va_list args)
54 const char *git_dir = get_git_dir();
55 size_t len;
57 len = strlen(git_dir);
58 if (n < len + 1)
59 goto bad;
60 memcpy(buf, git_dir, len);
61 if (len && !is_dir_sep(git_dir[len-1]))
62 buf[len++] = '/';
63 len += vsnprintf(buf + len, n - len, fmt, args);
64 if (len >= n)
65 goto bad;
66 return cleanup_path(buf);
67 bad:
68 strlcpy(buf, bad_path, n);
69 return buf;
72 char *git_snpath(char *buf, size_t n, const char *fmt, ...)
74 char *ret;
75 va_list args;
76 va_start(args, fmt);
77 ret = vsnpath(buf, n, fmt, args);
78 va_end(args);
79 return ret;
82 char *git_pathdup(const char *fmt, ...)
84 char path[PATH_MAX], *ret;
85 va_list args;
86 va_start(args, fmt);
87 ret = vsnpath(path, sizeof(path), fmt, args);
88 va_end(args);
89 return xstrdup(ret);
92 char *mkpathdup(const char *fmt, ...)
94 char *path;
95 struct strbuf sb = STRBUF_INIT;
96 va_list args;
98 va_start(args, fmt);
99 strbuf_vaddf(&sb, fmt, args);
100 va_end(args);
101 path = xstrdup(cleanup_path(sb.buf));
103 strbuf_release(&sb);
104 return path;
107 char *mkpath(const char *fmt, ...)
109 va_list args;
110 unsigned len;
111 char *pathname = get_pathname();
113 va_start(args, fmt);
114 len = vsnprintf(pathname, PATH_MAX, fmt, args);
115 va_end(args);
116 if (len >= PATH_MAX)
117 return bad_path;
118 return cleanup_path(pathname);
121 char *git_path(const char *fmt, ...)
123 char *pathname = get_pathname();
124 va_list args;
125 char *ret;
127 va_start(args, fmt);
128 ret = vsnpath(pathname, PATH_MAX, fmt, args);
129 va_end(args);
130 return ret;
133 void home_config_paths(char **global, char **xdg, char *file)
135 char *xdg_home = getenv("XDG_CONFIG_HOME");
136 char *home = getenv("HOME");
137 char *to_free = NULL;
139 if (!home) {
140 if (global)
141 *global = NULL;
142 } else {
143 if (!xdg_home) {
144 to_free = mkpathdup("%s/.config", home);
145 xdg_home = to_free;
147 if (global)
148 *global = mkpathdup("%s/.gitconfig", home);
151 if (xdg) {
152 if (!xdg_home)
153 *xdg = NULL;
154 else
155 *xdg = mkpathdup("%s/git/%s", xdg_home, file);
158 free(to_free);
161 char *git_path_submodule(const char *path, const char *fmt, ...)
163 char *pathname = get_pathname();
164 struct strbuf buf = STRBUF_INIT;
165 const char *git_dir;
166 va_list args;
167 unsigned len;
169 len = strlen(path);
170 if (len > PATH_MAX-100)
171 return bad_path;
173 strbuf_addstr(&buf, path);
174 if (len && path[len-1] != '/')
175 strbuf_addch(&buf, '/');
176 strbuf_addstr(&buf, ".git");
178 git_dir = read_gitfile(buf.buf);
179 if (git_dir) {
180 strbuf_reset(&buf);
181 strbuf_addstr(&buf, git_dir);
183 strbuf_addch(&buf, '/');
185 if (buf.len >= PATH_MAX)
186 return bad_path;
187 memcpy(pathname, buf.buf, buf.len + 1);
189 strbuf_release(&buf);
190 len = strlen(pathname);
192 va_start(args, fmt);
193 len += vsnprintf(pathname + len, PATH_MAX - len, fmt, args);
194 va_end(args);
195 if (len >= PATH_MAX)
196 return bad_path;
197 return cleanup_path(pathname);
200 int validate_headref(const char *path)
202 struct stat st;
203 char *buf, buffer[256];
204 unsigned char sha1[20];
205 int fd;
206 ssize_t len;
208 if (lstat(path, &st) < 0)
209 return -1;
211 /* Make sure it is a "refs/.." symlink */
212 if (S_ISLNK(st.st_mode)) {
213 len = readlink(path, buffer, sizeof(buffer)-1);
214 if (len >= 5 && !memcmp("refs/", buffer, 5))
215 return 0;
216 return -1;
220 * Anything else, just open it and try to see if it is a symbolic ref.
222 fd = open(path, O_RDONLY);
223 if (fd < 0)
224 return -1;
225 len = read_in_full(fd, buffer, sizeof(buffer)-1);
226 close(fd);
229 * Is it a symbolic ref?
231 if (len < 4)
232 return -1;
233 if (!memcmp("ref:", buffer, 4)) {
234 buf = buffer + 4;
235 len -= 4;
236 while (len && isspace(*buf))
237 buf++, len--;
238 if (len >= 5 && !memcmp("refs/", buf, 5))
239 return 0;
243 * Is this a detached HEAD?
245 if (!get_sha1_hex(buffer, sha1))
246 return 0;
248 return -1;
251 static struct passwd *getpw_str(const char *username, size_t len)
253 struct passwd *pw;
254 char *username_z = xmemdupz(username, len);
255 pw = getpwnam(username_z);
256 free(username_z);
257 return pw;
261 * Return a string with ~ and ~user expanded via getpw*. If buf != NULL,
262 * then it is a newly allocated string. Returns NULL on getpw failure or
263 * if path is NULL.
265 char *expand_user_path(const char *path)
267 struct strbuf user_path = STRBUF_INIT;
268 const char *to_copy = path;
270 if (path == NULL)
271 goto return_null;
272 if (path[0] == '~') {
273 const char *first_slash = strchrnul(path, '/');
274 const char *username = path + 1;
275 size_t username_len = first_slash - username;
276 if (username_len == 0) {
277 const char *home = getenv("HOME");
278 if (!home)
279 goto return_null;
280 strbuf_addstr(&user_path, home);
281 } else {
282 struct passwd *pw = getpw_str(username, username_len);
283 if (!pw)
284 goto return_null;
285 strbuf_addstr(&user_path, pw->pw_dir);
287 to_copy = first_slash;
289 strbuf_addstr(&user_path, to_copy);
290 return strbuf_detach(&user_path, NULL);
291 return_null:
292 strbuf_release(&user_path);
293 return NULL;
297 * First, one directory to try is determined by the following algorithm.
299 * (0) If "strict" is given, the path is used as given and no DWIM is
300 * done. Otherwise:
301 * (1) "~/path" to mean path under the running user's home directory;
302 * (2) "~user/path" to mean path under named user's home directory;
303 * (3) "relative/path" to mean cwd relative directory; or
304 * (4) "/absolute/path" to mean absolute directory.
306 * Unless "strict" is given, we try access() for existence of "%s.git/.git",
307 * "%s/.git", "%s.git", "%s" in this order. The first one that exists is
308 * what we try.
310 * Second, we try chdir() to that. Upon failure, we return NULL.
312 * Then, we try if the current directory is a valid git repository.
313 * Upon failure, we return NULL.
315 * If all goes well, we return the directory we used to chdir() (but
316 * before ~user is expanded), avoiding getcwd() resolving symbolic
317 * links. User relative paths are also returned as they are given,
318 * except DWIM suffixing.
320 const char *enter_repo(const char *path, int strict)
322 static char used_path[PATH_MAX];
323 static char validated_path[PATH_MAX];
325 if (!path)
326 return NULL;
328 if (!strict) {
329 static const char *suffix[] = {
330 "/.git", "", ".git/.git", ".git", NULL,
332 const char *gitfile;
333 int len = strlen(path);
334 int i;
335 while ((1 < len) && (path[len-1] == '/'))
336 len--;
338 if (PATH_MAX <= len)
339 return NULL;
340 strncpy(used_path, path, len); used_path[len] = 0 ;
341 strcpy(validated_path, used_path);
343 if (used_path[0] == '~') {
344 char *newpath = expand_user_path(used_path);
345 if (!newpath || (PATH_MAX - 10 < strlen(newpath))) {
346 free(newpath);
347 return NULL;
350 * Copy back into the static buffer. A pity
351 * since newpath was not bounded, but other
352 * branches of the if are limited by PATH_MAX
353 * anyway.
355 strcpy(used_path, newpath); free(newpath);
357 else if (PATH_MAX - 10 < len)
358 return NULL;
359 len = strlen(used_path);
360 for (i = 0; suffix[i]; i++) {
361 struct stat st;
362 strcpy(used_path + len, suffix[i]);
363 if (!stat(used_path, &st) &&
364 (S_ISREG(st.st_mode) ||
365 (S_ISDIR(st.st_mode) && is_git_directory(used_path)))) {
366 strcat(validated_path, suffix[i]);
367 break;
370 if (!suffix[i])
371 return NULL;
372 gitfile = read_gitfile(used_path) ;
373 if (gitfile)
374 strcpy(used_path, gitfile);
375 if (chdir(used_path))
376 return NULL;
377 path = validated_path;
379 else if (chdir(path))
380 return NULL;
382 if (access("objects", X_OK) == 0 && access("refs", X_OK) == 0 &&
383 validate_headref("HEAD") == 0) {
384 set_git_dir(".");
385 check_repository_format();
386 return path;
389 return NULL;
392 static int calc_shared_perm(int mode)
394 int tweak;
396 if (shared_repository < 0)
397 tweak = -shared_repository;
398 else
399 tweak = shared_repository;
401 if (!(mode & S_IWUSR))
402 tweak &= ~0222;
403 if (mode & S_IXUSR)
404 /* Copy read bits to execute bits */
405 tweak |= (tweak & 0444) >> 2;
406 if (shared_repository < 0)
407 mode = (mode & ~0777) | tweak;
408 else
409 mode |= tweak;
411 return mode;
415 int adjust_shared_perm(const char *path)
417 int old_mode, new_mode;
419 if (!shared_repository)
420 return 0;
421 if (get_st_mode_bits(path, &old_mode) < 0)
422 return -1;
424 new_mode = calc_shared_perm(old_mode);
425 if (S_ISDIR(old_mode)) {
426 /* Copy read bits to execute bits */
427 new_mode |= (new_mode & 0444) >> 2;
428 new_mode |= FORCE_DIR_SET_GID;
431 if (((old_mode ^ new_mode) & ~S_IFMT) &&
432 chmod(path, (new_mode & ~S_IFMT)) < 0)
433 return -2;
434 return 0;
437 static int have_same_root(const char *path1, const char *path2)
439 int is_abs1, is_abs2;
441 is_abs1 = is_absolute_path(path1);
442 is_abs2 = is_absolute_path(path2);
443 return (is_abs1 && is_abs2 && tolower(path1[0]) == tolower(path2[0])) ||
444 (!is_abs1 && !is_abs2);
448 * Give path as relative to prefix.
450 * The strbuf may or may not be used, so do not assume it contains the
451 * returned path.
453 const char *relative_path(const char *in, const char *prefix,
454 struct strbuf *sb)
456 int in_len = in ? strlen(in) : 0;
457 int prefix_len = prefix ? strlen(prefix) : 0;
458 int in_off = 0;
459 int prefix_off = 0;
460 int i = 0, j = 0;
462 if (!in_len)
463 return "./";
464 else if (!prefix_len)
465 return in;
467 if (have_same_root(in, prefix)) {
468 /* bypass dos_drive, for "c:" is identical to "C:" */
469 if (has_dos_drive_prefix(in)) {
470 i = 2;
471 j = 2;
473 } else {
474 return in;
477 while (i < prefix_len && j < in_len && prefix[i] == in[j]) {
478 if (is_dir_sep(prefix[i])) {
479 while (is_dir_sep(prefix[i]))
480 i++;
481 while (is_dir_sep(in[j]))
482 j++;
483 prefix_off = i;
484 in_off = j;
485 } else {
486 i++;
487 j++;
491 if (
492 /* "prefix" seems like prefix of "in" */
493 i >= prefix_len &&
495 * but "/foo" is not a prefix of "/foobar"
496 * (i.e. prefix not end with '/')
498 prefix_off < prefix_len) {
499 if (j >= in_len) {
500 /* in="/a/b", prefix="/a/b" */
501 in_off = in_len;
502 } else if (is_dir_sep(in[j])) {
503 /* in="/a/b/c", prefix="/a/b" */
504 while (is_dir_sep(in[j]))
505 j++;
506 in_off = j;
507 } else {
508 /* in="/a/bbb/c", prefix="/a/b" */
509 i = prefix_off;
511 } else if (
512 /* "in" is short than "prefix" */
513 j >= in_len &&
514 /* "in" not end with '/' */
515 in_off < in_len) {
516 if (is_dir_sep(prefix[i])) {
517 /* in="/a/b", prefix="/a/b/c/" */
518 while (is_dir_sep(prefix[i]))
519 i++;
520 in_off = in_len;
523 in += in_off;
524 in_len -= in_off;
526 if (i >= prefix_len) {
527 if (!in_len)
528 return "./";
529 else
530 return in;
533 strbuf_reset(sb);
534 strbuf_grow(sb, in_len);
536 while (i < prefix_len) {
537 if (is_dir_sep(prefix[i])) {
538 strbuf_addstr(sb, "../");
539 while (is_dir_sep(prefix[i]))
540 i++;
541 continue;
543 i++;
545 if (!is_dir_sep(prefix[prefix_len - 1]))
546 strbuf_addstr(sb, "../");
548 strbuf_addstr(sb, in);
550 return sb->buf;
554 * A simpler implementation of relative_path
556 * Get relative path by removing "prefix" from "in". This function
557 * first appears in v1.5.6-1-g044bbbc, and makes git_dir shorter
558 * to increase performance when traversing the path to work_tree.
560 const char *remove_leading_path(const char *in, const char *prefix)
562 static char buf[PATH_MAX + 1];
563 int i = 0, j = 0;
565 if (!prefix || !prefix[0])
566 return in;
567 while (prefix[i]) {
568 if (is_dir_sep(prefix[i])) {
569 if (!is_dir_sep(in[j]))
570 return in;
571 while (is_dir_sep(prefix[i]))
572 i++;
573 while (is_dir_sep(in[j]))
574 j++;
575 continue;
576 } else if (in[j] != prefix[i]) {
577 return in;
579 i++;
580 j++;
582 if (
583 /* "/foo" is a prefix of "/foo" */
584 in[j] &&
585 /* "/foo" is not a prefix of "/foobar" */
586 !is_dir_sep(prefix[i-1]) && !is_dir_sep(in[j])
588 return in;
589 while (is_dir_sep(in[j]))
590 j++;
591 if (!in[j])
592 strcpy(buf, ".");
593 else
594 strcpy(buf, in + j);
595 return buf;
599 * It is okay if dst == src, but they should not overlap otherwise.
601 * Performs the following normalizations on src, storing the result in dst:
602 * - Ensures that components are separated by '/' (Windows only)
603 * - Squashes sequences of '/'.
604 * - Removes "." components.
605 * - Removes ".." components, and the components the precede them.
606 * Returns failure (non-zero) if a ".." component appears as first path
607 * component anytime during the normalization. Otherwise, returns success (0).
609 * Note that this function is purely textual. It does not follow symlinks,
610 * verify the existence of the path, or make any system calls.
612 * prefix_len != NULL is for a specific case of prefix_pathspec():
613 * assume that src == dst and src[0..prefix_len-1] is already
614 * normalized, any time "../" eats up to the prefix_len part,
615 * prefix_len is reduced. In the end prefix_len is the remaining
616 * prefix that has not been overridden by user pathspec.
618 int normalize_path_copy_len(char *dst, const char *src, int *prefix_len)
620 char *dst0;
622 if (has_dos_drive_prefix(src)) {
623 *dst++ = *src++;
624 *dst++ = *src++;
626 dst0 = dst;
628 if (is_dir_sep(*src)) {
629 *dst++ = '/';
630 while (is_dir_sep(*src))
631 src++;
634 for (;;) {
635 char c = *src;
638 * A path component that begins with . could be
639 * special:
640 * (1) "." and ends -- ignore and terminate.
641 * (2) "./" -- ignore them, eat slash and continue.
642 * (3) ".." and ends -- strip one and terminate.
643 * (4) "../" -- strip one, eat slash and continue.
645 if (c == '.') {
646 if (!src[1]) {
647 /* (1) */
648 src++;
649 } else if (is_dir_sep(src[1])) {
650 /* (2) */
651 src += 2;
652 while (is_dir_sep(*src))
653 src++;
654 continue;
655 } else if (src[1] == '.') {
656 if (!src[2]) {
657 /* (3) */
658 src += 2;
659 goto up_one;
660 } else if (is_dir_sep(src[2])) {
661 /* (4) */
662 src += 3;
663 while (is_dir_sep(*src))
664 src++;
665 goto up_one;
670 /* copy up to the next '/', and eat all '/' */
671 while ((c = *src++) != '\0' && !is_dir_sep(c))
672 *dst++ = c;
673 if (is_dir_sep(c)) {
674 *dst++ = '/';
675 while (is_dir_sep(c))
676 c = *src++;
677 src--;
678 } else if (!c)
679 break;
680 continue;
682 up_one:
684 * dst0..dst is prefix portion, and dst[-1] is '/';
685 * go up one level.
687 dst--; /* go to trailing '/' */
688 if (dst <= dst0)
689 return -1;
690 /* Windows: dst[-1] cannot be backslash anymore */
691 while (dst0 < dst && dst[-1] != '/')
692 dst--;
693 if (prefix_len && *prefix_len > dst - dst0)
694 *prefix_len = dst - dst0;
696 *dst = '\0';
697 return 0;
700 int normalize_path_copy(char *dst, const char *src)
702 return normalize_path_copy_len(dst, src, NULL);
706 * path = Canonical absolute path
707 * prefixes = string_list containing normalized, absolute paths without
708 * trailing slashes (except for the root directory, which is denoted by "/").
710 * Determines, for each path in prefixes, whether the "prefix"
711 * is an ancestor directory of path. Returns the length of the longest
712 * ancestor directory, excluding any trailing slashes, or -1 if no prefix
713 * is an ancestor. (Note that this means 0 is returned if prefixes is
714 * ["/"].) "/foo" is not considered an ancestor of "/foobar". Directories
715 * are not considered to be their own ancestors. path must be in a
716 * canonical form: empty components, or "." or ".." components are not
717 * allowed.
719 int longest_ancestor_length(const char *path, struct string_list *prefixes)
721 int i, max_len = -1;
723 if (!strcmp(path, "/"))
724 return -1;
726 for (i = 0; i < prefixes->nr; i++) {
727 const char *ceil = prefixes->items[i].string;
728 int len = strlen(ceil);
730 if (len == 1 && ceil[0] == '/')
731 len = 0; /* root matches anything, with length 0 */
732 else if (!strncmp(path, ceil, len) && path[len] == '/')
733 ; /* match of length len */
734 else
735 continue; /* no match */
737 if (len > max_len)
738 max_len = len;
741 return max_len;
744 /* strip arbitrary amount of directory separators at end of path */
745 static inline int chomp_trailing_dir_sep(const char *path, int len)
747 while (len && is_dir_sep(path[len - 1]))
748 len--;
749 return len;
753 * If path ends with suffix (complete path components), returns the
754 * part before suffix (sans trailing directory separators).
755 * Otherwise returns NULL.
757 char *strip_path_suffix(const char *path, const char *suffix)
759 int path_len = strlen(path), suffix_len = strlen(suffix);
761 while (suffix_len) {
762 if (!path_len)
763 return NULL;
765 if (is_dir_sep(path[path_len - 1])) {
766 if (!is_dir_sep(suffix[suffix_len - 1]))
767 return NULL;
768 path_len = chomp_trailing_dir_sep(path, path_len);
769 suffix_len = chomp_trailing_dir_sep(suffix, suffix_len);
771 else if (path[--path_len] != suffix[--suffix_len])
772 return NULL;
775 if (path_len && !is_dir_sep(path[path_len - 1]))
776 return NULL;
777 return xstrndup(path, chomp_trailing_dir_sep(path, path_len));
780 int daemon_avoid_alias(const char *p)
782 int sl, ndot;
785 * This resurrects the belts and suspenders paranoia check by HPA
786 * done in <435560F7.4080006@zytor.com> thread, now enter_repo()
787 * does not do getcwd() based path canonicalization.
789 * sl becomes true immediately after seeing '/' and continues to
790 * be true as long as dots continue after that without intervening
791 * non-dot character.
793 if (!p || (*p != '/' && *p != '~'))
794 return -1;
795 sl = 1; ndot = 0;
796 p++;
798 while (1) {
799 char ch = *p++;
800 if (sl) {
801 if (ch == '.')
802 ndot++;
803 else if (ch == '/') {
804 if (ndot < 3)
805 /* reject //, /./ and /../ */
806 return -1;
807 ndot = 0;
809 else if (ch == 0) {
810 if (0 < ndot && ndot < 3)
811 /* reject /.$ and /..$ */
812 return -1;
813 return 0;
815 else
816 sl = ndot = 0;
818 else if (ch == 0)
819 return 0;
820 else if (ch == '/') {
821 sl = 1;
822 ndot = 0;
827 static int only_spaces_and_periods(const char *path, size_t len, size_t skip)
829 if (len < skip)
830 return 0;
831 len -= skip;
832 path += skip;
833 while (len-- > 0) {
834 char c = *(path++);
835 if (c != ' ' && c != '.')
836 return 0;
838 return 1;
841 int is_ntfs_dotgit(const char *name)
843 int len;
845 for (len = 0; ; len++)
846 if (!name[len] || name[len] == '\\' || is_dir_sep(name[len])) {
847 if (only_spaces_and_periods(name, len, 4) &&
848 !strncasecmp(name, ".git", 4))
849 return 1;
850 if (only_spaces_and_periods(name, len, 5) &&
851 !strncasecmp(name, "git~1", 5))
852 return 1;
853 if (name[len] != '\\')
854 return 0;
855 name += len + 1;
856 len = -1;