archive: handle commits with an empty tree
[git/jnareb-git.git] / symlinks.c
blobc2b41a85013eec87a412badc4097f364d547726d
1 #include "cache.h"
3 static int threaded_check_leading_path(struct cache_def *cache, const char *name, int len);
4 static int threaded_has_dirs_only_path(struct cache_def *cache, const char *name, int len, int prefix_len);
6 /*
7 * Returns the length (on a path component basis) of the longest
8 * common prefix match of 'name_a' and 'name_b'.
9 */
10 static int longest_path_match(const char *name_a, int len_a,
11 const char *name_b, int len_b,
12 int *previous_slash)
14 int max_len, match_len = 0, match_len_prev = 0, i = 0;
16 max_len = len_a < len_b ? len_a : len_b;
17 while (i < max_len && name_a[i] == name_b[i]) {
18 if (name_a[i] == '/') {
19 match_len_prev = match_len;
20 match_len = i;
22 i++;
25 * Is 'name_b' a substring of 'name_a', the other way around,
26 * or is 'name_a' and 'name_b' the exact same string?
28 if (i >= max_len && ((len_a > len_b && name_a[len_b] == '/') ||
29 (len_a < len_b && name_b[len_a] == '/') ||
30 (len_a == len_b))) {
31 match_len_prev = match_len;
32 match_len = i;
34 *previous_slash = match_len_prev;
35 return match_len;
38 static struct cache_def default_cache;
40 static inline void reset_lstat_cache(struct cache_def *cache)
42 cache->path[0] = '\0';
43 cache->len = 0;
44 cache->flags = 0;
46 * The track_flags and prefix_len_stat_func members is only
47 * set by the safeguard rule inside lstat_cache()
51 #define FL_DIR (1 << 0)
52 #define FL_NOENT (1 << 1)
53 #define FL_SYMLINK (1 << 2)
54 #define FL_LSTATERR (1 << 3)
55 #define FL_ERR (1 << 4)
56 #define FL_FULLPATH (1 << 5)
59 * Check if name 'name' of length 'len' has a symlink leading
60 * component, or if the directory exists and is real, or not.
62 * To speed up the check, some information is allowed to be cached.
63 * This can be indicated by the 'track_flags' argument, which also can
64 * be used to indicate that we should check the full path.
66 * The 'prefix_len_stat_func' parameter can be used to set the length
67 * of the prefix, where the cache should use the stat() function
68 * instead of the lstat() function to test each path component.
70 static int lstat_cache_matchlen(struct cache_def *cache,
71 const char *name, int len,
72 int *ret_flags, int track_flags,
73 int prefix_len_stat_func)
75 int match_len, last_slash, last_slash_dir, previous_slash;
76 int save_flags, max_len, ret;
77 struct stat st;
79 if (cache->track_flags != track_flags ||
80 cache->prefix_len_stat_func != prefix_len_stat_func) {
82 * As a safeguard rule we clear the cache if the
83 * values of track_flags and/or prefix_len_stat_func
84 * does not match with the last supplied values.
86 reset_lstat_cache(cache);
87 cache->track_flags = track_flags;
88 cache->prefix_len_stat_func = prefix_len_stat_func;
89 match_len = last_slash = 0;
90 } else {
92 * Check to see if we have a match from the cache for
93 * the 2 "excluding" path types.
95 match_len = last_slash =
96 longest_path_match(name, len, cache->path, cache->len,
97 &previous_slash);
98 *ret_flags = cache->flags & track_flags & (FL_NOENT|FL_SYMLINK);
100 if (!(track_flags & FL_FULLPATH) && match_len == len)
101 match_len = last_slash = previous_slash;
103 if (*ret_flags && match_len == cache->len)
104 return match_len;
106 * If we now have match_len > 0, we would know that
107 * the matched part will always be a directory.
109 * Also, if we are tracking directories and 'name' is
110 * a substring of the cache on a path component basis,
111 * we can return immediately.
113 *ret_flags = track_flags & FL_DIR;
114 if (*ret_flags && len == match_len)
115 return match_len;
119 * Okay, no match from the cache so far, so now we have to
120 * check the rest of the path components.
122 *ret_flags = FL_DIR;
123 last_slash_dir = last_slash;
124 max_len = len < PATH_MAX ? len : PATH_MAX;
125 while (match_len < max_len) {
126 do {
127 cache->path[match_len] = name[match_len];
128 match_len++;
129 } while (match_len < max_len && name[match_len] != '/');
130 if (match_len >= max_len && !(track_flags & FL_FULLPATH))
131 break;
132 last_slash = match_len;
133 cache->path[last_slash] = '\0';
135 if (last_slash <= prefix_len_stat_func)
136 ret = stat(cache->path, &st);
137 else
138 ret = lstat(cache->path, &st);
140 if (ret) {
141 *ret_flags = FL_LSTATERR;
142 if (errno == ENOENT)
143 *ret_flags |= FL_NOENT;
144 } else if (S_ISDIR(st.st_mode)) {
145 last_slash_dir = last_slash;
146 continue;
147 } else if (S_ISLNK(st.st_mode)) {
148 *ret_flags = FL_SYMLINK;
149 } else {
150 *ret_flags = FL_ERR;
152 break;
156 * At the end update the cache. Note that max 3 different
157 * path types, FL_NOENT, FL_SYMLINK and FL_DIR, can be cached
158 * for the moment!
160 save_flags = *ret_flags & track_flags & (FL_NOENT|FL_SYMLINK);
161 if (save_flags && last_slash > 0 && last_slash <= PATH_MAX) {
162 cache->path[last_slash] = '\0';
163 cache->len = last_slash;
164 cache->flags = save_flags;
165 } else if ((track_flags & FL_DIR) &&
166 last_slash_dir > 0 && last_slash_dir <= PATH_MAX) {
168 * We have a separate test for the directory case,
169 * since it could be that we have found a symlink or a
170 * non-existing directory and the track_flags says
171 * that we cannot cache this fact, so the cache would
172 * then have been left empty in this case.
174 * But if we are allowed to track real directories, we
175 * can still cache the path components before the last
176 * one (the found symlink or non-existing component).
178 cache->path[last_slash_dir] = '\0';
179 cache->len = last_slash_dir;
180 cache->flags = FL_DIR;
181 } else {
182 reset_lstat_cache(cache);
184 return match_len;
187 static int lstat_cache(struct cache_def *cache, const char *name, int len,
188 int track_flags, int prefix_len_stat_func)
190 int flags;
191 (void)lstat_cache_matchlen(cache, name, len, &flags, track_flags,
192 prefix_len_stat_func);
193 return flags;
196 #define USE_ONLY_LSTAT 0
199 * Return non-zero if path 'name' has a leading symlink component
201 int threaded_has_symlink_leading_path(struct cache_def *cache, const char *name, int len)
203 return lstat_cache(cache, name, len, FL_SYMLINK|FL_DIR, USE_ONLY_LSTAT) & FL_SYMLINK;
207 * Return non-zero if path 'name' has a leading symlink component
209 int has_symlink_leading_path(const char *name, int len)
211 return threaded_has_symlink_leading_path(&default_cache, name, len);
215 * Return zero if path 'name' has a leading symlink component or
216 * if some leading path component does not exists.
218 * Return -1 if leading path exists and is a directory.
220 * Return path length if leading path exists and is neither a
221 * directory nor a symlink.
223 int check_leading_path(const char *name, int len)
225 return threaded_check_leading_path(&default_cache, name, len);
229 * Return zero if path 'name' has a leading symlink component or
230 * if some leading path component does not exists.
232 * Return -1 if leading path exists and is a directory.
234 * Return path length if leading path exists and is neither a
235 * directory nor a symlink.
237 static int threaded_check_leading_path(struct cache_def *cache, const char *name, int len)
239 int flags;
240 int match_len = lstat_cache_matchlen(cache, name, len, &flags,
241 FL_SYMLINK|FL_NOENT|FL_DIR, USE_ONLY_LSTAT);
242 if (flags & FL_NOENT)
243 return 0;
244 else if (flags & FL_DIR)
245 return -1;
246 else
247 return match_len;
251 * Return non-zero if all path components of 'name' exists as a
252 * directory. If prefix_len > 0, we will test with the stat()
253 * function instead of the lstat() function for a prefix length of
254 * 'prefix_len', thus we then allow for symlinks in the prefix part as
255 * long as those points to real existing directories.
257 int has_dirs_only_path(const char *name, int len, int prefix_len)
259 return threaded_has_dirs_only_path(&default_cache, name, len, prefix_len);
263 * Return non-zero if all path components of 'name' exists as a
264 * directory. If prefix_len > 0, we will test with the stat()
265 * function instead of the lstat() function for a prefix length of
266 * 'prefix_len', thus we then allow for symlinks in the prefix part as
267 * long as those points to real existing directories.
269 static int threaded_has_dirs_only_path(struct cache_def *cache, const char *name, int len, int prefix_len)
271 return lstat_cache(cache, name, len,
272 FL_DIR|FL_FULLPATH, prefix_len) &
273 FL_DIR;
276 static struct removal_def {
277 char path[PATH_MAX];
278 int len;
279 } removal;
281 static void do_remove_scheduled_dirs(int new_len)
283 while (removal.len > new_len) {
284 removal.path[removal.len] = '\0';
285 if (rmdir(removal.path))
286 break;
287 do {
288 removal.len--;
289 } while (removal.len > new_len &&
290 removal.path[removal.len] != '/');
292 removal.len = new_len;
295 void schedule_dir_for_removal(const char *name, int len)
297 int match_len, last_slash, i, previous_slash;
299 match_len = last_slash = i =
300 longest_path_match(name, len, removal.path, removal.len,
301 &previous_slash);
302 /* Find last slash inside 'name' */
303 while (i < len) {
304 if (name[i] == '/')
305 last_slash = i;
306 i++;
310 * If we are about to go down the directory tree, we check if
311 * we must first go upwards the tree, such that we then can
312 * remove possible empty directories as we go upwards.
314 if (match_len < last_slash && match_len < removal.len)
315 do_remove_scheduled_dirs(match_len);
317 * If we go deeper down the directory tree, we only need to
318 * save the new path components as we go down.
320 if (match_len < last_slash) {
321 memcpy(&removal.path[match_len], &name[match_len],
322 last_slash - match_len);
323 removal.len = last_slash;
327 void remove_scheduled_dirs(void)
329 do_remove_scheduled_dirs(0);