Improve sha1 object file writing.
[git/git.git] / update-cache.c
1 /*
2 * GIT - The information manager from hell
3 *
4 * Copyright (C) Linus Torvalds, 2005
5 */
6 #include <signal.h>
7 #include "cache.h"
8
9 /*
10 * Default to not allowing changes to the list of files. The
11 * tool doesn't actually care, but this makes it harder to add
12 * files to the revision control by mistake by doing something
13 * like "update-cache *" and suddenly having all the object
14 * files be revision controlled.
15 */
16 static int allow_add = 0, allow_remove = 0, not_new = 0;
17
18 /* Three functions to allow overloaded pointer return; see linux/err.h */
19 static inline void *ERR_PTR(long error)
20 {
21 return (void *) error;
22 }
23
24 static inline long PTR_ERR(const void *ptr)
25 {
26 return (long) ptr;
27 }
28
29 static inline long IS_ERR(const void *ptr)
30 {
31 return (unsigned long)ptr > (unsigned long)-1000L;
32 }
33
34 /*
35 * This only updates the "non-critical" parts of the directory
36 * cache, ie the parts that aren't tracked by GIT, and only used
37 * to validate the cache.
38 */
39 static void fill_stat_cache_info(struct cache_entry *ce, struct stat *st)
40 {
41 ce->ce_ctime.sec = htonl(st->st_ctime);
42 ce->ce_mtime.sec = htonl(st->st_mtime);
43 #ifdef NSEC
44 ce->ce_ctime.nsec = htonl(st->st_ctim.tv_nsec);
45 ce->ce_mtime.nsec = htonl(st->st_mtim.tv_nsec);
46 #endif
47 ce->ce_dev = htonl(st->st_dev);
48 ce->ce_ino = htonl(st->st_ino);
49 ce->ce_uid = htonl(st->st_uid);
50 ce->ce_gid = htonl(st->st_gid);
51 ce->ce_size = htonl(st->st_size);
52 }
53
54 static int add_file_to_cache(char *path)
55 {
56 int size, namelen;
57 struct cache_entry *ce;
58 struct stat st;
59 int fd;
60
61 fd = open(path, O_RDONLY);
62 if (fd < 0) {
63 if (errno == ENOENT || errno == ENOTDIR) {
64 if (allow_remove)
65 return remove_file_from_cache(path);
66 }
67 return -1;
68 }
69 if (fstat(fd, &st) < 0) {
70 close(fd);
71 return -1;
72 }
73 namelen = strlen(path);
74 size = cache_entry_size(namelen);
75 ce = xmalloc(size);
76 memset(ce, 0, size);
77 memcpy(ce->name, path, namelen);
78 fill_stat_cache_info(ce, &st);
79 ce->ce_mode = create_ce_mode(st.st_mode);
80 ce->ce_flags = htons(namelen);
81
82 if (index_fd(ce->sha1, fd, &st) < 0)
83 return -1;
84
85 return add_cache_entry(ce, allow_add);
86 }
87
88 static int match_data(int fd, void *buffer, unsigned long size)
89 {
90 while (size) {
91 char compare[1024];
92 int ret = read(fd, compare, sizeof(compare));
93
94 if (ret <= 0 || ret > size || memcmp(buffer, compare, ret))
95 return -1;
96 size -= ret;
97 buffer += ret;
98 }
99 return 0;
100 }
101
102 static int compare_data(struct cache_entry *ce, unsigned long expected_size)
103 {
104 int match = -1;
105 int fd = open(ce->name, O_RDONLY);
106
107 if (fd >= 0) {
108 void *buffer;
109 unsigned long size;
110 char type[10];
111
112 buffer = read_sha1_file(ce->sha1, type, &size);
113 if (buffer) {
114 if (size == expected_size && !strcmp(type, "blob"))
115 match = match_data(fd, buffer, size);
116 free(buffer);
117 }
118 close(fd);
119 }
120 return match;
121 }
122
123 /*
124 * "refresh" does not calculate a new sha1 file or bring the
125 * cache up-to-date for mode/content changes. But what it
126 * _does_ do is to "re-match" the stat information of a file
127 * with the cache, so that you can refresh the cache for a
128 * file that hasn't been changed but where the stat entry is
129 * out of date.
130 *
131 * For example, you'd want to do this after doing a "read-tree",
132 * to link up the stat cache details with the proper files.
133 */
134 static struct cache_entry *refresh_entry(struct cache_entry *ce)
135 {
136 struct stat st;
137 struct cache_entry *updated;
138 int changed, size;
139
140 if (stat(ce->name, &st) < 0)
141 return ERR_PTR(-errno);
142
143 changed = cache_match_stat(ce, &st);
144 if (!changed)
145 return ce;
146
147 /*
148 * If the mode has changed, there's no point in trying
149 * to refresh the entry - it's not going to match
150 */
151 if (changed & MODE_CHANGED)
152 return ERR_PTR(-EINVAL);
153
154 if (compare_data(ce, st.st_size))
155 return ERR_PTR(-EINVAL);
156
157 size = ce_size(ce);
158 updated = xmalloc(size);
159 memcpy(updated, ce, size);
160 fill_stat_cache_info(updated, &st);
161 return updated;
162 }
163
164 static int refresh_cache(void)
165 {
166 int i;
167 int has_errors = 0;
168
169 for (i = 0; i < active_nr; i++) {
170 struct cache_entry *ce, *new;
171 ce = active_cache[i];
172 if (ce_stage(ce)) {
173 printf("%s: needs merge\n", ce->name);
174 has_errors = 1;
175 while ((i < active_nr) &&
176 ! strcmp(active_cache[i]->name, ce->name))
177 i++;
178 i--;
179 continue;
180 }
181
182 new = refresh_entry(ce);
183 if (IS_ERR(new)) {
184 if (!(not_new && PTR_ERR(new) == -ENOENT)) {
185 printf("%s: needs update\n", ce->name);
186 has_errors = 1;
187 }
188 continue;
189 }
190 active_cache[i] = new;
191 }
192 return has_errors;
193 }
194
195 /*
196 * We fundamentally don't like some paths: we don't want
197 * dot or dot-dot anywhere, and in fact, we don't even want
198 * any other dot-files (.git or anything else). They
199 * are hidden, for chist sake.
200 *
201 * Also, we don't want double slashes or slashes at the
202 * end that can make pathnames ambiguous.
203 */
204 static int verify_path(char *path)
205 {
206 char c;
207
208 goto inside;
209 for (;;) {
210 if (!c)
211 return 1;
212 if (c == '/') {
213 inside:
214 c = *path++;
215 if (c != '/' && c != '.' && c != '\0')
216 continue;
217 return 0;
218 }
219 c = *path++;
220 }
221 }
222
223 static int add_cacheinfo(char *arg1, char *arg2, char *arg3)
224 {
225 int size, len;
226 unsigned int mode;
227 unsigned char sha1[20];
228 struct cache_entry *ce;
229
230 if (sscanf(arg1, "%o", &mode) != 1)
231 return -1;
232 if (get_sha1_hex(arg2, sha1))
233 return -1;
234 if (!verify_path(arg3))
235 return -1;
236
237 len = strlen(arg3);
238 size = cache_entry_size(len);
239 ce = xmalloc(size);
240 memset(ce, 0, size);
241
242 memcpy(ce->sha1, sha1, 20);
243 memcpy(ce->name, arg3, len);
244 ce->ce_flags = htons(len);
245 ce->ce_mode = create_ce_mode(mode);
246 return add_cache_entry(ce, allow_add);
247 }
248
249 static const char *lockfile_name = NULL;
250
251 static void remove_lock_file(void)
252 {
253 if (lockfile_name)
254 unlink(lockfile_name);
255 }
256
257 static void remove_lock_file_on_signal(int signo)
258 {
259 remove_lock_file();
260 }
261
262 int main(int argc, char **argv)
263 {
264 int i, newfd, entries, has_errors = 0;
265 int allow_options = 1;
266 static char lockfile[MAXPATHLEN+1];
267 const char *indexfile = get_index_file();
268
269 snprintf(lockfile, sizeof(lockfile), "%s.lock", indexfile);
270
271 newfd = open(lockfile, O_RDWR | O_CREAT | O_EXCL, 0600);
272 if (newfd < 0)
273 die("unable to create new cachefile");
274
275 signal(SIGINT, remove_lock_file_on_signal);
276 atexit(remove_lock_file);
277 lockfile_name = lockfile;
278
279 entries = read_cache();
280 if (entries < 0)
281 die("cache corrupted");
282
283 for (i = 1 ; i < argc; i++) {
284 char *path = argv[i];
285
286 if (allow_options && *path == '-') {
287 if (!strcmp(path, "--")) {
288 allow_options = 0;
289 continue;
290 }
291 if (!strcmp(path, "--add")) {
292 allow_add = 1;
293 continue;
294 }
295 if (!strcmp(path, "--remove")) {
296 allow_remove = 1;
297 continue;
298 }
299 if (!strcmp(path, "--refresh")) {
300 has_errors |= refresh_cache();
301 continue;
302 }
303 if (!strcmp(path, "--cacheinfo")) {
304 if (i+3 >= argc || add_cacheinfo(argv[i+1], argv[i+2], argv[i+3]))
305 die("update-cache: --cacheinfo <mode> <sha1> <path>");
306 i += 3;
307 continue;
308 }
309 if (!strcmp(path, "--force-remove")) {
310 if (argc <= i + 1)
311 die("update-cache: --force-remove <path>");
312 if (remove_file_from_cache(argv[i+1]))
313 die("update-cache: --force-remove cannot remove %s", argv[i+1]);
314 i++;
315 continue;
316 }
317
318 if (!strcmp(path, "--ignore-missing")) {
319 not_new = 1;
320 continue;
321 }
322 die("unknown option %s", path);
323 }
324 if (!verify_path(path)) {
325 fprintf(stderr, "Ignoring path %s\n", argv[i]);
326 continue;
327 }
328 if (add_file_to_cache(path))
329 die("Unable to add %s to database", path);
330 }
331 if (write_cache(newfd, active_cache, active_nr) || rename(lockfile, indexfile))
332 die("Unable to write new cachefile");
333
334 lockfile_name = NULL;
335 return has_errors;
336 }