]>
Commit | Line | Data |
---|---|---|
8bc9a0c7 LT |
1 | /* |
2 | * GIT - The information manager from hell | |
3 | * | |
4 | * Copyright (C) Linus Torvalds, 2005 | |
5 | */ | |
4aab5b46 | 6 | #define NO_THE_INDEX_COMPATIBILITY_MACROS |
e83c5163 | 7 | #include "cache.h" |
b2141fc1 | 8 | #include "config.h" |
f6ecc62d | 9 | #include "tempfile.h" |
697cc8ef | 10 | #include "lockfile.h" |
bad68ec9 | 11 | #include "cache-tree.h" |
f35a6d3b | 12 | #include "refs.h" |
d616813d | 13 | #include "dir.h" |
041aee31 JH |
14 | #include "tree.h" |
15 | #include "commit.h" | |
39425819 | 16 | #include "blob.h" |
cfc5789a | 17 | #include "resolve-undo.h" |
6c9cd161 JH |
18 | #include "strbuf.h" |
19 | #include "varint.h" | |
5fc2fc8f | 20 | #include "split-index.h" |
a42643aa | 21 | #include "utf8.h" |
883e248b | 22 | #include "fsmonitor.h" |
bad68ec9 | 23 | |
b60e188c TG |
24 | /* Mask for the name length in ce_flags in the on-disk index */ |
25 | ||
26 | #define CE_NAMEMASK (0x0fff) | |
27 | ||
bad68ec9 JH |
28 | /* Index extensions. |
29 | * | |
30 | * The first letter should be 'A'..'Z' for extensions that are not | |
31 | * necessary for a correct operation (i.e. optimization data). | |
32 | * When new extensions are added that _needs_ to be understood in | |
33 | * order to correctly interpret the index file, pick character that | |
34 | * is outside the range, to cause the reader to abort. | |
35 | */ | |
36 | ||
37 | #define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) ) | |
38 | #define CACHE_EXT_TREE 0x54524545 /* "TREE" */ | |
b659b49b | 39 | #define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */ |
5fc2fc8f | 40 | #define CACHE_EXT_LINK 0x6c696e6b /* "link" */ |
83c094ad | 41 | #define CACHE_EXT_UNTRACKED 0x554E5452 /* "UNTR" */ |
883e248b | 42 | #define CACHE_EXT_FSMONITOR 0x46534D4E /* "FSMN" */ |
5fc2fc8f NTND |
43 | |
44 | /* changes that can be kept in $GIT_DIR/index (basically all extensions) */ | |
e0cf0d7d | 45 | #define EXTMASK (RESOLVE_UNDO_CHANGED | CACHE_TREE_CHANGED | \ |
c18b80a0 | 46 | CE_ENTRY_ADDED | CE_ENTRY_REMOVED | CE_ENTRY_CHANGED | \ |
883e248b | 47 | SPLIT_INDEX_ORDERED | UNTRACKED_CHANGED | FSMONITOR_CHANGED) |
e83c5163 | 48 | |
8e72d675 JM |
49 | |
50 | /* | |
51 | * This is an estimate of the pathname length in the index. We use | |
52 | * this for V4 index files to guess the un-deltafied size of the index | |
53 | * in memory because of pathname deltafication. This is not required | |
54 | * for V2/V3 index formats because their pathnames are not compressed. | |
55 | * If the initial amount of memory set aside is not sufficient, the | |
56 | * mem pool will allocate extra memory. | |
57 | */ | |
58 | #define CACHE_ENTRY_PATH_LENGTH 80 | |
59 | ||
60 | static inline struct cache_entry *mem_pool__ce_alloc(struct mem_pool *mem_pool, size_t len) | |
61 | { | |
62 | struct cache_entry *ce; | |
63 | ce = mem_pool_alloc(mem_pool, cache_entry_size(len)); | |
64 | ce->mem_pool_allocated = 1; | |
65 | return ce; | |
66 | } | |
67 | ||
68 | static inline struct cache_entry *mem_pool__ce_calloc(struct mem_pool *mem_pool, size_t len) | |
69 | { | |
70 | struct cache_entry * ce; | |
71 | ce = mem_pool_calloc(mem_pool, 1, cache_entry_size(len)); | |
72 | ce->mem_pool_allocated = 1; | |
73 | return ce; | |
74 | } | |
75 | ||
76 | static struct mem_pool *find_mem_pool(struct index_state *istate) | |
77 | { | |
78 | struct mem_pool **pool_ptr; | |
79 | ||
80 | if (istate->split_index && istate->split_index->base) | |
81 | pool_ptr = &istate->split_index->base->ce_mem_pool; | |
82 | else | |
83 | pool_ptr = &istate->ce_mem_pool; | |
84 | ||
85 | if (!*pool_ptr) | |
86 | mem_pool_init(pool_ptr, 0); | |
87 | ||
88 | return *pool_ptr; | |
89 | } | |
90 | ||
228e94f9 | 91 | struct index_state the_index; |
626f35c8 | 92 | static const char *alternate_index_output; |
8fd2cb40 | 93 | |
9cb76b8c JH |
94 | static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
95 | { | |
96 | istate->cache[nr] = ce; | |
96872bc2 | 97 | add_name_hash(istate, ce); |
9cb76b8c JH |
98 | } |
99 | ||
cf558704 LT |
100 | static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
101 | { | |
102 | struct cache_entry *old = istate->cache[nr]; | |
103 | ||
078a58e8 | 104 | replace_index_entry_in_base(istate, old, ce); |
2092678c | 105 | remove_name_hash(istate, old); |
a849735b | 106 | discard_cache_entry(old); |
0e267b7a | 107 | ce->ce_flags &= ~CE_HASHED; |
a22c6371 | 108 | set_index_entry(istate, nr, ce); |
078a58e8 | 109 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 110 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 111 | istate->cache_changed |= CE_ENTRY_CHANGED; |
cf558704 LT |
112 | } |
113 | ||
81dc2307 PB |
114 | void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name) |
115 | { | |
285c2e25 | 116 | struct cache_entry *old_entry = istate->cache[nr], *new_entry; |
81dc2307 PB |
117 | int namelen = strlen(new_name); |
118 | ||
a849735b | 119 | new_entry = make_empty_cache_entry(istate, namelen); |
285c2e25 BW |
120 | copy_cache_entry(new_entry, old_entry); |
121 | new_entry->ce_flags &= ~CE_HASHED; | |
122 | new_entry->ce_namelen = namelen; | |
123 | new_entry->index = 0; | |
124 | memcpy(new_entry->name, new_name, namelen + 1); | |
81dc2307 | 125 | |
285c2e25 BW |
126 | cache_tree_invalidate_path(istate, old_entry->name); |
127 | untracked_cache_remove_from_index(istate, old_entry->name); | |
81dc2307 | 128 | remove_index_entry_at(istate, nr); |
285c2e25 | 129 | add_index_entry(istate, new_entry, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); |
81dc2307 PB |
130 | } |
131 | ||
c21d39d7 MH |
132 | void fill_stat_data(struct stat_data *sd, struct stat *st) |
133 | { | |
134 | sd->sd_ctime.sec = (unsigned int)st->st_ctime; | |
135 | sd->sd_mtime.sec = (unsigned int)st->st_mtime; | |
136 | sd->sd_ctime.nsec = ST_CTIME_NSEC(*st); | |
137 | sd->sd_mtime.nsec = ST_MTIME_NSEC(*st); | |
138 | sd->sd_dev = st->st_dev; | |
139 | sd->sd_ino = st->st_ino; | |
140 | sd->sd_uid = st->st_uid; | |
141 | sd->sd_gid = st->st_gid; | |
142 | sd->sd_size = st->st_size; | |
143 | } | |
144 | ||
145 | int match_stat_data(const struct stat_data *sd, struct stat *st) | |
146 | { | |
147 | int changed = 0; | |
148 | ||
149 | if (sd->sd_mtime.sec != (unsigned int)st->st_mtime) | |
150 | changed |= MTIME_CHANGED; | |
151 | if (trust_ctime && check_stat && | |
152 | sd->sd_ctime.sec != (unsigned int)st->st_ctime) | |
153 | changed |= CTIME_CHANGED; | |
154 | ||
155 | #ifdef USE_NSEC | |
156 | if (check_stat && sd->sd_mtime.nsec != ST_MTIME_NSEC(*st)) | |
157 | changed |= MTIME_CHANGED; | |
158 | if (trust_ctime && check_stat && | |
159 | sd->sd_ctime.nsec != ST_CTIME_NSEC(*st)) | |
160 | changed |= CTIME_CHANGED; | |
161 | #endif | |
162 | ||
163 | if (check_stat) { | |
164 | if (sd->sd_uid != (unsigned int) st->st_uid || | |
165 | sd->sd_gid != (unsigned int) st->st_gid) | |
166 | changed |= OWNER_CHANGED; | |
167 | if (sd->sd_ino != (unsigned int) st->st_ino) | |
168 | changed |= INODE_CHANGED; | |
169 | } | |
170 | ||
171 | #ifdef USE_STDEV | |
172 | /* | |
173 | * st_dev breaks on network filesystems where different | |
174 | * clients will have different views of what "device" | |
175 | * the filesystem is on | |
176 | */ | |
177 | if (check_stat && sd->sd_dev != (unsigned int) st->st_dev) | |
178 | changed |= INODE_CHANGED; | |
179 | #endif | |
180 | ||
181 | if (sd->sd_size != (unsigned int) st->st_size) | |
182 | changed |= DATA_CHANGED; | |
183 | ||
184 | return changed; | |
185 | } | |
186 | ||
415e96c8 JH |
187 | /* |
188 | * This only updates the "non-critical" parts of the directory | |
189 | * cache, ie the parts that aren't tracked by GIT, and only used | |
190 | * to validate the cache. | |
191 | */ | |
192 | void fill_stat_cache_info(struct cache_entry *ce, struct stat *st) | |
193 | { | |
c21d39d7 | 194 | fill_stat_data(&ce->ce_stat_data, st); |
5f73076c JH |
195 | |
196 | if (assume_unchanged) | |
7a51ed66 | 197 | ce->ce_flags |= CE_VALID; |
eadb5831 | 198 | |
883e248b | 199 | if (S_ISREG(st->st_mode)) { |
eadb5831 | 200 | ce_mark_uptodate(ce); |
883e248b BP |
201 | mark_fsmonitor_valid(ce); |
202 | } | |
415e96c8 JH |
203 | } |
204 | ||
21a6b9fa | 205 | static int ce_compare_data(const struct cache_entry *ce, struct stat *st) |
29e4d363 JH |
206 | { |
207 | int match = -1; | |
1b8ac5ea | 208 | int fd = git_open_cloexec(ce->name, O_RDONLY); |
29e4d363 JH |
209 | |
210 | if (fd >= 0) { | |
bebfecb9 | 211 | struct object_id oid; |
e3506559 | 212 | if (!index_fd(&oid, fd, st, OBJ_BLOB, ce->name, 0)) |
bebfecb9 | 213 | match = oidcmp(&oid, &ce->oid); |
7f8508e8 | 214 | /* index_fd() closed the file descriptor already */ |
29e4d363 JH |
215 | } |
216 | return match; | |
217 | } | |
218 | ||
21a6b9fa | 219 | static int ce_compare_link(const struct cache_entry *ce, size_t expected_size) |
29e4d363 JH |
220 | { |
221 | int match = -1; | |
29e4d363 JH |
222 | void *buffer; |
223 | unsigned long size; | |
21666f1a | 224 | enum object_type type; |
a60272b3 | 225 | struct strbuf sb = STRBUF_INIT; |
29e4d363 | 226 | |
a60272b3 | 227 | if (strbuf_readlink(&sb, ce->name, expected_size)) |
29e4d363 | 228 | return -1; |
a60272b3 | 229 | |
b4f5aca4 | 230 | buffer = read_object_file(&ce->oid, &type, &size); |
a60272b3 LT |
231 | if (buffer) { |
232 | if (size == sb.len) | |
233 | match = memcmp(buffer, sb.buf, size); | |
234 | free(buffer); | |
29e4d363 | 235 | } |
a60272b3 | 236 | strbuf_release(&sb); |
29e4d363 JH |
237 | return match; |
238 | } | |
239 | ||
21a6b9fa | 240 | static int ce_compare_gitlink(const struct cache_entry *ce) |
f35a6d3b | 241 | { |
1053fe82 | 242 | struct object_id oid; |
f35a6d3b LT |
243 | |
244 | /* | |
245 | * We don't actually require that the .git directory | |
302b9282 | 246 | * under GITLINK directory be a valid git directory. It |
f35a6d3b LT |
247 | * might even be missing (in case nobody populated that |
248 | * sub-project). | |
249 | * | |
250 | * If so, we consider it always to match. | |
251 | */ | |
a98e6101 | 252 | if (resolve_gitlink_ref(ce->name, "HEAD", &oid) < 0) |
f35a6d3b | 253 | return 0; |
1053fe82 | 254 | return oidcmp(&oid, &ce->oid); |
f35a6d3b LT |
255 | } |
256 | ||
21a6b9fa | 257 | static int ce_modified_check_fs(const struct cache_entry *ce, struct stat *st) |
29e4d363 JH |
258 | { |
259 | switch (st->st_mode & S_IFMT) { | |
260 | case S_IFREG: | |
261 | if (ce_compare_data(ce, st)) | |
262 | return DATA_CHANGED; | |
263 | break; | |
264 | case S_IFLNK: | |
dc49cd76 | 265 | if (ce_compare_link(ce, xsize_t(st->st_size))) |
29e4d363 JH |
266 | return DATA_CHANGED; |
267 | break; | |
a8ee75bc | 268 | case S_IFDIR: |
7a51ed66 | 269 | if (S_ISGITLINK(ce->ce_mode)) |
c70115b4 | 270 | return ce_compare_gitlink(ce) ? DATA_CHANGED : 0; |
1cf01a34 | 271 | /* else fallthrough */ |
29e4d363 JH |
272 | default: |
273 | return TYPE_CHANGED; | |
274 | } | |
275 | return 0; | |
276 | } | |
277 | ||
21a6b9fa | 278 | static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st) |
734aab75 LT |
279 | { |
280 | unsigned int changed = 0; | |
281 | ||
7a51ed66 LT |
282 | if (ce->ce_flags & CE_REMOVE) |
283 | return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED; | |
284 | ||
285 | switch (ce->ce_mode & S_IFMT) { | |
8ae0a8c5 KS |
286 | case S_IFREG: |
287 | changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0; | |
3e09cdfd JH |
288 | /* We consider only the owner x bit to be relevant for |
289 | * "mode changes" | |
290 | */ | |
291 | if (trust_executable_bit && | |
7a51ed66 | 292 | (0100 & (ce->ce_mode ^ st->st_mode))) |
ffbe1add | 293 | changed |= MODE_CHANGED; |
8ae0a8c5 KS |
294 | break; |
295 | case S_IFLNK: | |
78a8d641 JS |
296 | if (!S_ISLNK(st->st_mode) && |
297 | (has_symlinks || !S_ISREG(st->st_mode))) | |
298 | changed |= TYPE_CHANGED; | |
8ae0a8c5 | 299 | break; |
302b9282 | 300 | case S_IFGITLINK: |
c70115b4 | 301 | /* We ignore most of the st_xxx fields for gitlinks */ |
f35a6d3b LT |
302 | if (!S_ISDIR(st->st_mode)) |
303 | changed |= TYPE_CHANGED; | |
304 | else if (ce_compare_gitlink(ce)) | |
305 | changed |= DATA_CHANGED; | |
a8ee75bc | 306 | return changed; |
8ae0a8c5 | 307 | default: |
7a51ed66 | 308 | die("internal error: ce_mode is %o", ce->ce_mode); |
8ae0a8c5 | 309 | } |
2cb45e95 | 310 | |
c21d39d7 | 311 | changed |= match_stat_data(&ce->ce_stat_data, st); |
b0391890 | 312 | |
f49c2c22 | 313 | /* Racily smudged entry? */ |
c21d39d7 | 314 | if (!ce->ce_stat_data.sd_size) { |
99d1a986 | 315 | if (!is_empty_blob_sha1(ce->oid.hash)) |
f49c2c22 LT |
316 | changed |= DATA_CHANGED; |
317 | } | |
318 | ||
407c8eb0 JH |
319 | return changed; |
320 | } | |
321 | ||
2bb4cda1 NTND |
322 | static int is_racy_stat(const struct index_state *istate, |
323 | const struct stat_data *sd) | |
6d91da6d | 324 | { |
2bb4cda1 | 325 | return (istate->timestamp.sec && |
fba2f38a KB |
326 | #ifdef USE_NSEC |
327 | /* nanosecond timestamped files can also be racy! */ | |
2bb4cda1 NTND |
328 | (istate->timestamp.sec < sd->sd_mtime.sec || |
329 | (istate->timestamp.sec == sd->sd_mtime.sec && | |
330 | istate->timestamp.nsec <= sd->sd_mtime.nsec)) | |
fba2f38a | 331 | #else |
2bb4cda1 | 332 | istate->timestamp.sec <= sd->sd_mtime.sec |
fba2f38a | 333 | #endif |
2bb4cda1 NTND |
334 | ); |
335 | } | |
336 | ||
337 | static int is_racy_timestamp(const struct index_state *istate, | |
338 | const struct cache_entry *ce) | |
339 | { | |
340 | return (!S_ISGITLINK(ce->ce_mode) && | |
341 | is_racy_stat(istate, &ce->ce_stat_data)); | |
6d91da6d JH |
342 | } |
343 | ||
ed4efab1 NTND |
344 | int match_stat_data_racy(const struct index_state *istate, |
345 | const struct stat_data *sd, struct stat *st) | |
346 | { | |
347 | if (is_racy_stat(istate, sd)) | |
348 | return MTIME_CHANGED; | |
349 | return match_stat_data(sd, st); | |
6d91da6d JH |
350 | } |
351 | ||
883e248b | 352 | int ie_match_stat(struct index_state *istate, |
21a6b9fa | 353 | const struct cache_entry *ce, struct stat *st, |
4bd5b7da | 354 | unsigned int options) |
407c8eb0 | 355 | { |
5f73076c | 356 | unsigned int changed; |
4bd5b7da | 357 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 358 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
4bd5b7da | 359 | int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY; |
883e248b | 360 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
5f73076c | 361 | |
883e248b BP |
362 | if (!ignore_fsmonitor) |
363 | refresh_fsmonitor(istate); | |
5f73076c JH |
364 | /* |
365 | * If it's marked as always valid in the index, it's | |
366 | * valid whatever the checked-out copy says. | |
56cac48c NTND |
367 | * |
368 | * skip-worktree has the same effect with higher precedence | |
5f73076c | 369 | */ |
56cac48c NTND |
370 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) |
371 | return 0; | |
7a51ed66 | 372 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) |
5f73076c | 373 | return 0; |
883e248b BP |
374 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) |
375 | return 0; | |
5f73076c | 376 | |
331fcb59 JH |
377 | /* |
378 | * Intent-to-add entries have not been added, so the index entry | |
379 | * by definition never matches what is in the work tree until it | |
380 | * actually gets added. | |
381 | */ | |
895ff3b2 | 382 | if (ce_intent_to_add(ce)) |
331fcb59 JH |
383 | return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED; |
384 | ||
5f73076c | 385 | changed = ce_match_stat_basic(ce, st); |
407c8eb0 | 386 | |
29e4d363 JH |
387 | /* |
388 | * Within 1 second of this sequence: | |
389 | * echo xyzzy >file && git-update-index --add file | |
390 | * running this command: | |
391 | * echo frotz >file | |
392 | * would give a falsely clean cache entry. The mtime and | |
393 | * length match the cache, and other stat fields do not change. | |
394 | * | |
395 | * We could detect this at update-index time (the cache entry | |
396 | * being registered/updated records the same time as "now") | |
397 | * and delay the return from git-update-index, but that would | |
398 | * effectively mean we can make at most one commit per second, | |
399 | * which is not acceptable. Instead, we check cache entries | |
400 | * whose mtime are the same as the index file timestamp more | |
5f73076c | 401 | * carefully than others. |
29e4d363 | 402 | */ |
6d91da6d | 403 | if (!changed && is_racy_timestamp(istate, ce)) { |
42f77406 JH |
404 | if (assume_racy_is_modified) |
405 | changed |= DATA_CHANGED; | |
406 | else | |
407 | changed |= ce_modified_check_fs(ce, st); | |
408 | } | |
b0391890 | 409 | |
29e4d363 | 410 | return changed; |
b0391890 JH |
411 | } |
412 | ||
883e248b | 413 | int ie_modified(struct index_state *istate, |
21a6b9fa RS |
414 | const struct cache_entry *ce, |
415 | struct stat *st, unsigned int options) | |
b0391890 | 416 | { |
29e4d363 | 417 | int changed, changed_fs; |
4bd5b7da JH |
418 | |
419 | changed = ie_match_stat(istate, ce, st, options); | |
b0391890 JH |
420 | if (!changed) |
421 | return 0; | |
b0391890 JH |
422 | /* |
423 | * If the mode or type has changed, there's no point in trying | |
424 | * to refresh the entry - it's not going to match | |
425 | */ | |
426 | if (changed & (MODE_CHANGED | TYPE_CHANGED)) | |
427 | return changed; | |
428 | ||
c70115b4 JH |
429 | /* |
430 | * Immediately after read-tree or update-index --cacheinfo, | |
431 | * the length field is zero, as we have never even read the | |
432 | * lstat(2) information once, and we cannot trust DATA_CHANGED | |
433 | * returned by ie_match_stat() which in turn was returned by | |
434 | * ce_match_stat_basic() to signal that the filesize of the | |
435 | * blob changed. We have to actually go to the filesystem to | |
436 | * see if the contents match, and if so, should answer "unchanged". | |
437 | * | |
438 | * The logic does not apply to gitlinks, as ce_match_stat_basic() | |
439 | * already has checked the actual HEAD from the filesystem in the | |
440 | * subproject. If ie_match_stat() already said it is different, | |
441 | * then we know it is. | |
b0391890 | 442 | */ |
c70115b4 | 443 | if ((changed & DATA_CHANGED) && |
c21d39d7 | 444 | (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0)) |
b0391890 JH |
445 | return changed; |
446 | ||
29e4d363 JH |
447 | changed_fs = ce_modified_check_fs(ce, st); |
448 | if (changed_fs) | |
449 | return changed | changed_fs; | |
b0391890 JH |
450 | return 0; |
451 | } | |
452 | ||
958ba6c9 LT |
453 | int base_name_compare(const char *name1, int len1, int mode1, |
454 | const char *name2, int len2, int mode2) | |
455 | { | |
456 | unsigned char c1, c2; | |
457 | int len = len1 < len2 ? len1 : len2; | |
458 | int cmp; | |
459 | ||
460 | cmp = memcmp(name1, name2, len); | |
461 | if (cmp) | |
462 | return cmp; | |
463 | c1 = name1[len]; | |
464 | c2 = name2[len]; | |
1833a925 | 465 | if (!c1 && S_ISDIR(mode1)) |
958ba6c9 | 466 | c1 = '/'; |
1833a925 | 467 | if (!c2 && S_ISDIR(mode2)) |
958ba6c9 LT |
468 | c2 = '/'; |
469 | return (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0; | |
470 | } | |
471 | ||
0ab9e1e8 LT |
472 | /* |
473 | * df_name_compare() is identical to base_name_compare(), except it | |
474 | * compares conflicting directory/file entries as equal. Note that | |
475 | * while a directory name compares as equal to a regular file, they | |
476 | * then individually compare _differently_ to a filename that has | |
477 | * a dot after the basename (because '\0' < '.' < '/'). | |
478 | * | |
479 | * This is used by routines that want to traverse the git namespace | |
480 | * but then handle conflicting entries together when possible. | |
481 | */ | |
482 | int df_name_compare(const char *name1, int len1, int mode1, | |
483 | const char *name2, int len2, int mode2) | |
484 | { | |
485 | int len = len1 < len2 ? len1 : len2, cmp; | |
486 | unsigned char c1, c2; | |
487 | ||
488 | cmp = memcmp(name1, name2, len); | |
489 | if (cmp) | |
490 | return cmp; | |
491 | /* Directories and files compare equal (same length, same name) */ | |
492 | if (len1 == len2) | |
493 | return 0; | |
494 | c1 = name1[len]; | |
495 | if (!c1 && S_ISDIR(mode1)) | |
496 | c1 = '/'; | |
497 | c2 = name2[len]; | |
498 | if (!c2 && S_ISDIR(mode2)) | |
499 | c2 = '/'; | |
500 | if (c1 == '/' && !c2) | |
501 | return 0; | |
502 | if (c2 == '/' && !c1) | |
503 | return 0; | |
504 | return c1 - c2; | |
505 | } | |
506 | ||
ccdd4a0f | 507 | int name_compare(const char *name1, size_t len1, const char *name2, size_t len2) |
eb38c22f | 508 | { |
ccdd4a0f JM |
509 | size_t min_len = (len1 < len2) ? len1 : len2; |
510 | int cmp = memcmp(name1, name2, min_len); | |
eb38c22f LT |
511 | if (cmp) |
512 | return cmp; | |
513 | if (len1 < len2) | |
514 | return -1; | |
515 | if (len1 > len2) | |
516 | return 1; | |
ccdd4a0f JM |
517 | return 0; |
518 | } | |
519 | ||
520 | int cache_name_stage_compare(const char *name1, int len1, int stage1, const char *name2, int len2, int stage2) | |
521 | { | |
522 | int cmp; | |
523 | ||
524 | cmp = name_compare(name1, len1, name2, len2); | |
525 | if (cmp) | |
526 | return cmp; | |
5f73076c | 527 | |
b60e188c | 528 | if (stage1 < stage2) |
95fd5bf8 | 529 | return -1; |
b60e188c | 530 | if (stage1 > stage2) |
95fd5bf8 | 531 | return 1; |
eb38c22f LT |
532 | return 0; |
533 | } | |
534 | ||
357e9c69 | 535 | static int index_name_stage_pos(const struct index_state *istate, const char *name, int namelen, int stage) |
eb38c22f LT |
536 | { |
537 | int first, last; | |
538 | ||
539 | first = 0; | |
4aab5b46 | 540 | last = istate->cache_nr; |
eb38c22f LT |
541 | while (last > first) { |
542 | int next = (last + first) >> 1; | |
4aab5b46 | 543 | struct cache_entry *ce = istate->cache[next]; |
b60e188c | 544 | int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce)); |
eb38c22f | 545 | if (!cmp) |
76e7f4ec | 546 | return next; |
eb38c22f LT |
547 | if (cmp < 0) { |
548 | last = next; | |
549 | continue; | |
550 | } | |
551 | first = next+1; | |
552 | } | |
76e7f4ec | 553 | return -first-1; |
eb38c22f LT |
554 | } |
555 | ||
b60e188c TG |
556 | int index_name_pos(const struct index_state *istate, const char *name, int namelen) |
557 | { | |
558 | return index_name_stage_pos(istate, name, namelen, 0); | |
559 | } | |
560 | ||
4aab5b46 | 561 | int remove_index_entry_at(struct index_state *istate, int pos) |
7b937ca3 | 562 | { |
cf558704 LT |
563 | struct cache_entry *ce = istate->cache[pos]; |
564 | ||
cfc5789a | 565 | record_resolve_undo(istate, ce); |
2092678c | 566 | remove_name_hash(istate, ce); |
045113a5 | 567 | save_or_free_index_entry(istate, ce); |
e636a7b4 | 568 | istate->cache_changed |= CE_ENTRY_REMOVED; |
4aab5b46 JH |
569 | istate->cache_nr--; |
570 | if (pos >= istate->cache_nr) | |
7b937ca3 | 571 | return 0; |
f331ab9d RS |
572 | MOVE_ARRAY(istate->cache + pos, istate->cache + pos + 1, |
573 | istate->cache_nr - pos); | |
7b937ca3 LT |
574 | return 1; |
575 | } | |
576 | ||
36419c8e | 577 | /* |
98e023de | 578 | * Remove all cache entries marked for removal, that is where |
36419c8e KB |
579 | * CE_REMOVE is set in ce_flags. This is much more effective than |
580 | * calling remove_index_entry_at() for each entry to be removed. | |
581 | */ | |
582 | void remove_marked_cache_entries(struct index_state *istate) | |
583 | { | |
584 | struct cache_entry **ce_array = istate->cache; | |
585 | unsigned int i, j; | |
586 | ||
587 | for (i = j = 0; i < istate->cache_nr; i++) { | |
5699d17e | 588 | if (ce_array[i]->ce_flags & CE_REMOVE) { |
2092678c | 589 | remove_name_hash(istate, ce_array[i]); |
045113a5 | 590 | save_or_free_index_entry(istate, ce_array[i]); |
5699d17e | 591 | } |
36419c8e KB |
592 | else |
593 | ce_array[j++] = ce_array[i]; | |
594 | } | |
ad837d9e NTND |
595 | if (j == istate->cache_nr) |
596 | return; | |
e636a7b4 | 597 | istate->cache_changed |= CE_ENTRY_REMOVED; |
36419c8e KB |
598 | istate->cache_nr = j; |
599 | } | |
600 | ||
4aab5b46 | 601 | int remove_file_from_index(struct index_state *istate, const char *path) |
197ee8c9 | 602 | { |
4aab5b46 | 603 | int pos = index_name_pos(istate, path, strlen(path)); |
c4e3cca1 JH |
604 | if (pos < 0) |
605 | pos = -pos-1; | |
a5400efe | 606 | cache_tree_invalidate_path(istate, path); |
e931371a | 607 | untracked_cache_remove_from_index(istate, path); |
4aab5b46 JH |
608 | while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path)) |
609 | remove_index_entry_at(istate, pos); | |
197ee8c9 LT |
610 | return 0; |
611 | } | |
612 | ||
20314271 JS |
613 | static int compare_name(struct cache_entry *ce, const char *path, int namelen) |
614 | { | |
615 | return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen); | |
616 | } | |
617 | ||
618 | static int index_name_pos_also_unmerged(struct index_state *istate, | |
619 | const char *path, int namelen) | |
620 | { | |
621 | int pos = index_name_pos(istate, path, namelen); | |
622 | struct cache_entry *ce; | |
623 | ||
624 | if (pos >= 0) | |
625 | return pos; | |
626 | ||
627 | /* maybe unmerged? */ | |
628 | pos = -1 - pos; | |
629 | if (pos >= istate->cache_nr || | |
630 | compare_name((ce = istate->cache[pos]), path, namelen)) | |
631 | return -1; | |
632 | ||
633 | /* order of preference: stage 2, 1, 3 */ | |
634 | if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr && | |
635 | ce_stage((ce = istate->cache[pos + 1])) == 2 && | |
636 | !compare_name(ce, path, namelen)) | |
637 | pos++; | |
638 | return pos; | |
639 | } | |
640 | ||
1102952b LT |
641 | static int different_name(struct cache_entry *ce, struct cache_entry *alias) |
642 | { | |
643 | int len = ce_namelen(ce); | |
644 | return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len); | |
645 | } | |
646 | ||
647 | /* | |
648 | * If we add a filename that aliases in the cache, we will use the | |
649 | * name that we already have - but we don't want to update the same | |
650 | * alias twice, because that implies that there were actually two | |
651 | * different files with aliasing names! | |
652 | * | |
653 | * So we use the CE_ADDED flag to verify that the alias was an old | |
654 | * one before we accept it as | |
655 | */ | |
045113a5 NTND |
656 | static struct cache_entry *create_alias_ce(struct index_state *istate, |
657 | struct cache_entry *ce, | |
658 | struct cache_entry *alias) | |
1102952b LT |
659 | { |
660 | int len; | |
285c2e25 | 661 | struct cache_entry *new_entry; |
1102952b LT |
662 | |
663 | if (alias->ce_flags & CE_ADDED) | |
664 | die("Will not add file alias '%s' ('%s' already exists in index)", ce->name, alias->name); | |
665 | ||
666 | /* Ok, create the new entry using the name of the existing alias */ | |
667 | len = ce_namelen(alias); | |
a849735b | 668 | new_entry = make_empty_cache_entry(istate, len); |
285c2e25 BW |
669 | memcpy(new_entry->name, alias->name, len); |
670 | copy_cache_entry(new_entry, ce); | |
045113a5 | 671 | save_or_free_index_entry(istate, ce); |
285c2e25 | 672 | return new_entry; |
1102952b LT |
673 | } |
674 | ||
b4b313f9 | 675 | void set_object_name_for_intent_to_add_entry(struct cache_entry *ce) |
39425819 | 676 | { |
a09c985e PO |
677 | struct object_id oid; |
678 | if (write_object_file("", 0, blob_type, &oid)) | |
39425819 | 679 | die("cannot create an empty blob in the object database"); |
a09c985e | 680 | oidcpy(&ce->oid, &oid); |
39425819 JH |
681 | } |
682 | ||
610d55af | 683 | int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags) |
11be42a4 | 684 | { |
a849735b | 685 | int namelen, was_same; |
d177cab0 | 686 | mode_t st_mode = st->st_mode; |
9472935d | 687 | struct cache_entry *ce, *alias = NULL; |
56cac48c | 688 | unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY; |
38ed1d89 JH |
689 | int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND); |
690 | int pretend = flags & ADD_CACHE_PRETEND; | |
39425819 JH |
691 | int intent_only = flags & ADD_CACHE_INTENT; |
692 | int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE| | |
693 | (intent_only ? ADD_CACHE_NEW_ONLY : 0)); | |
9472935d TB |
694 | int newflags = HASH_WRITE_OBJECT; |
695 | ||
696 | if (flags & HASH_RENORMALIZE) | |
697 | newflags |= HASH_RENORMALIZE; | |
11be42a4 | 698 | |
d177cab0 | 699 | if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode)) |
960b8ad1 | 700 | return error("%s: can only add regular files, symbolic links or git-directories", path); |
11be42a4 JS |
701 | |
702 | namelen = strlen(path); | |
d177cab0 | 703 | if (S_ISDIR(st_mode)) { |
09595258 LT |
704 | while (namelen && path[namelen-1] == '/') |
705 | namelen--; | |
706 | } | |
a849735b | 707 | ce = make_empty_cache_entry(istate, namelen); |
11be42a4 | 708 | memcpy(ce->name, path, namelen); |
b60e188c | 709 | ce->ce_namelen = namelen; |
39425819 JH |
710 | if (!intent_only) |
711 | fill_stat_cache_info(ce, st); | |
388b2acd JH |
712 | else |
713 | ce->ce_flags |= CE_INTENT_TO_ADD; | |
11be42a4 | 714 | |
610d55af TG |
715 | |
716 | if (trust_executable_bit && has_symlinks) { | |
d177cab0 | 717 | ce->ce_mode = create_ce_mode(st_mode); |
610d55af | 718 | } else { |
78a8d641 JS |
719 | /* If there is an existing entry, pick the mode bits and type |
720 | * from it, otherwise assume unexecutable regular file. | |
11be42a4 | 721 | */ |
185c975f | 722 | struct cache_entry *ent; |
20314271 | 723 | int pos = index_name_pos_also_unmerged(istate, path, namelen); |
185c975f | 724 | |
4aab5b46 | 725 | ent = (0 <= pos) ? istate->cache[pos] : NULL; |
d177cab0 | 726 | ce->ce_mode = ce_mode_from_stat(ent, st_mode); |
11be42a4 JS |
727 | } |
728 | ||
dc1ae704 JJ |
729 | /* When core.ignorecase=true, determine if a directory of the same name but differing |
730 | * case already exists within the Git repository. If it does, ensure the directory | |
731 | * case of the file being added to the repository matches (is folded into) the existing | |
732 | * entry's directory case. | |
733 | */ | |
734 | if (ignore_case) { | |
41284eb0 | 735 | adjust_dirname_case(istate, ce->name); |
dc1ae704 | 736 | } |
9472935d TB |
737 | if (!(flags & HASH_RENORMALIZE)) { |
738 | alias = index_file_exists(istate, ce->name, | |
739 | ce_namelen(ce), ignore_case); | |
740 | if (alias && | |
741 | !ce_stage(alias) && | |
742 | !ie_match_stat(istate, alias, st, ce_option)) { | |
743 | /* Nothing changed, really */ | |
744 | if (!S_ISGITLINK(alias->ce_mode)) | |
745 | ce_mark_uptodate(alias); | |
746 | alias->ce_flags |= CE_ADDED; | |
dc1ae704 | 747 | |
a849735b | 748 | discard_cache_entry(ce); |
9472935d TB |
749 | return 0; |
750 | } | |
0781b8a9 | 751 | } |
39425819 | 752 | if (!intent_only) { |
9472935d | 753 | if (index_path(&ce->oid, path, st, newflags)) { |
a849735b | 754 | discard_cache_entry(ce); |
39425819 | 755 | return error("unable to index file %s", path); |
2d9426b0 | 756 | } |
39425819 | 757 | } else |
b4b313f9 | 758 | set_object_name_for_intent_to_add_entry(ce); |
39425819 | 759 | |
1102952b | 760 | if (ignore_case && alias && different_name(ce, alias)) |
045113a5 | 761 | ce = create_alias_ce(istate, ce, alias); |
1102952b | 762 | ce->ce_flags |= CE_ADDED; |
38ed1d89 | 763 | |
3bf0dd1f | 764 | /* It was suspected to be racily clean, but it turns out to be Ok */ |
38ed1d89 JH |
765 | was_same = (alias && |
766 | !ce_stage(alias) && | |
99d1a986 | 767 | !oidcmp(&alias->oid, &ce->oid) && |
38ed1d89 JH |
768 | ce->ce_mode == alias->ce_mode); |
769 | ||
770 | if (pretend) | |
a849735b | 771 | discard_cache_entry(ce); |
067178ed | 772 | else if (add_index_entry(istate, ce, add_option)) { |
a849735b | 773 | discard_cache_entry(ce); |
067178ed JH |
774 | return error("unable to add %s to index", path); |
775 | } | |
38ed1d89 | 776 | if (verbose && !was_same) |
11be42a4 | 777 | printf("add '%s'\n", path); |
11be42a4 JS |
778 | return 0; |
779 | } | |
780 | ||
610d55af | 781 | int add_file_to_index(struct index_state *istate, const char *path, int flags) |
d177cab0 LT |
782 | { |
783 | struct stat st; | |
784 | if (lstat(path, &st)) | |
d824cbba | 785 | die_errno("unable to stat '%s'", path); |
610d55af | 786 | return add_to_index(istate, path, &st, flags); |
d177cab0 LT |
787 | } |
788 | ||
a849735b JM |
789 | struct cache_entry *make_empty_cache_entry(struct index_state *istate, size_t len) |
790 | { | |
8e72d675 | 791 | return mem_pool__ce_calloc(find_mem_pool(istate), len); |
a849735b JM |
792 | } |
793 | ||
794 | struct cache_entry *make_empty_transient_cache_entry(size_t len) | |
795 | { | |
796 | return xcalloc(1, cache_entry_size(len)); | |
797 | } | |
798 | ||
799 | struct cache_entry *make_cache_entry(struct index_state *istate, | |
800 | unsigned int mode, | |
825ed4d9 JM |
801 | const struct object_id *oid, |
802 | const char *path, | |
803 | int stage, | |
804 | unsigned int refresh_options) | |
6640f881 | 805 | { |
bc1c2caa | 806 | struct cache_entry *ce, *ret; |
a849735b | 807 | int len; |
6640f881 | 808 | |
10ecfa76 | 809 | if (!verify_path(path, mode)) { |
7e7abea9 | 810 | error("Invalid path '%s'", path); |
6640f881 | 811 | return NULL; |
7e7abea9 | 812 | } |
6640f881 CR |
813 | |
814 | len = strlen(path); | |
a849735b | 815 | ce = make_empty_cache_entry(istate, len); |
6640f881 | 816 | |
825ed4d9 | 817 | oidcpy(&ce->oid, oid); |
6640f881 | 818 | memcpy(ce->name, path, len); |
b60e188c TG |
819 | ce->ce_flags = create_ce_flags(stage); |
820 | ce->ce_namelen = len; | |
6640f881 CR |
821 | ce->ce_mode = create_ce_mode(mode); |
822 | ||
768d7965 | 823 | ret = refresh_cache_entry(&the_index, ce, refresh_options); |
915e44c6 | 824 | if (ret != ce) |
a849735b | 825 | discard_cache_entry(ce); |
915e44c6 | 826 | return ret; |
6640f881 CR |
827 | } |
828 | ||
a849735b JM |
829 | struct cache_entry *make_transient_cache_entry(unsigned int mode, const struct object_id *oid, |
830 | const char *path, int stage) | |
831 | { | |
832 | struct cache_entry *ce; | |
833 | int len; | |
834 | ||
835 | if (!verify_path(path, mode)) { | |
836 | error("Invalid path '%s'", path); | |
837 | return NULL; | |
838 | } | |
839 | ||
840 | len = strlen(path); | |
841 | ce = make_empty_transient_cache_entry(len); | |
842 | ||
843 | oidcpy(&ce->oid, oid); | |
844 | memcpy(ce->name, path, len); | |
845 | ce->ce_flags = create_ce_flags(stage); | |
846 | ce->ce_namelen = len; | |
847 | ce->ce_mode = create_ce_mode(mode); | |
848 | ||
849 | return ce; | |
850 | } | |
851 | ||
d9d70966 TG |
852 | /* |
853 | * Chmod an index entry with either +x or -x. | |
854 | * | |
855 | * Returns -1 if the chmod for the particular cache entry failed (if it's | |
856 | * not a regular file), -2 if an invalid flip argument is passed in, 0 | |
857 | * otherwise. | |
858 | */ | |
859 | int chmod_index_entry(struct index_state *istate, struct cache_entry *ce, | |
860 | char flip) | |
861 | { | |
862 | if (!S_ISREG(ce->ce_mode)) | |
863 | return -1; | |
864 | switch (flip) { | |
865 | case '+': | |
866 | ce->ce_mode |= 0111; | |
867 | break; | |
868 | case '-': | |
869 | ce->ce_mode &= ~0111; | |
870 | break; | |
871 | default: | |
872 | return -2; | |
873 | } | |
874 | cache_tree_invalidate_path(istate, ce->name); | |
875 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
883e248b | 876 | mark_fsmonitor_invalid(istate, ce); |
d9d70966 TG |
877 | istate->cache_changed |= CE_ENTRY_CHANGED; |
878 | ||
879 | return 0; | |
880 | } | |
881 | ||
9c5e6c80 | 882 | int ce_same_name(const struct cache_entry *a, const struct cache_entry *b) |
7b937ca3 LT |
883 | { |
884 | int len = ce_namelen(a); | |
885 | return ce_namelen(b) == len && !memcmp(a->name, b->name, len); | |
886 | } | |
887 | ||
8dcf39c4 LT |
888 | /* |
889 | * We fundamentally don't like some paths: we don't want | |
890 | * dot or dot-dot anywhere, and for obvious reasons don't | |
891 | * want to recurse into ".git" either. | |
892 | * | |
893 | * Also, we don't want double slashes or slashes at the | |
894 | * end that can make pathnames ambiguous. | |
895 | */ | |
10ecfa76 | 896 | static int verify_dotfile(const char *rest, unsigned mode) |
8dcf39c4 LT |
897 | { |
898 | /* | |
899 | * The first character was '.', but that | |
900 | * has already been discarded, we now test | |
901 | * the rest. | |
902 | */ | |
e0f530ff | 903 | |
8dcf39c4 | 904 | /* "." is not allowed */ |
e0f530ff | 905 | if (*rest == '\0' || is_dir_sep(*rest)) |
8dcf39c4 LT |
906 | return 0; |
907 | ||
e0f530ff | 908 | switch (*rest) { |
8dcf39c4 | 909 | /* |
641084b6 JK |
910 | * ".git" followed by NUL or slash is bad. Note that we match |
911 | * case-insensitively here, even if ignore_case is not set. | |
912 | * This outlaws ".GIT" everywhere out of an abundance of caution, | |
913 | * since there's really no good reason to allow it. | |
10ecfa76 JK |
914 | * |
915 | * Once we've seen ".git", we can also find ".gitmodules", etc (also | |
916 | * case-insensitively). | |
8dcf39c4 LT |
917 | */ |
918 | case 'g': | |
cc2fc7c2 JK |
919 | case 'G': |
920 | if (rest[1] != 'i' && rest[1] != 'I') | |
8dcf39c4 | 921 | break; |
cc2fc7c2 | 922 | if (rest[2] != 't' && rest[2] != 'T') |
8dcf39c4 | 923 | break; |
e19e5e66 JK |
924 | if (rest[3] == '\0' || is_dir_sep(rest[3])) |
925 | return 0; | |
10ecfa76 JK |
926 | if (S_ISLNK(mode)) { |
927 | rest += 3; | |
928 | if (skip_iprefix(rest, "modules", &rest) && | |
929 | (*rest == '\0' || is_dir_sep(*rest))) | |
930 | return 0; | |
931 | } | |
e19e5e66 | 932 | break; |
8dcf39c4 | 933 | case '.': |
e0f530ff | 934 | if (rest[1] == '\0' || is_dir_sep(rest[1])) |
8dcf39c4 LT |
935 | return 0; |
936 | } | |
937 | return 1; | |
938 | } | |
939 | ||
10ecfa76 | 940 | int verify_path(const char *path, unsigned mode) |
8dcf39c4 LT |
941 | { |
942 | char c; | |
943 | ||
56948cb6 EFL |
944 | if (has_dos_drive_prefix(path)) |
945 | return 0; | |
946 | ||
8dcf39c4 LT |
947 | goto inside; |
948 | for (;;) { | |
949 | if (!c) | |
950 | return 1; | |
56948cb6 | 951 | if (is_dir_sep(c)) { |
8dcf39c4 | 952 | inside: |
10ecfa76 JK |
953 | if (protect_hfs) { |
954 | if (is_hfs_dotgit(path)) | |
955 | return 0; | |
956 | if (S_ISLNK(mode)) { | |
957 | if (is_hfs_dotgitmodules(path)) | |
958 | return 0; | |
959 | } | |
960 | } | |
961 | if (protect_ntfs) { | |
962 | if (is_ntfs_dotgit(path)) | |
963 | return 0; | |
964 | if (S_ISLNK(mode)) { | |
965 | if (is_ntfs_dotgitmodules(path)) | |
966 | return 0; | |
967 | } | |
968 | } | |
969 | ||
8dcf39c4 | 970 | c = *path++; |
10ecfa76 | 971 | if ((c == '.' && !verify_dotfile(path, mode)) || |
3bdf09c7 JH |
972 | is_dir_sep(c) || c == '\0') |
973 | return 0; | |
8dcf39c4 LT |
974 | } |
975 | c = *path++; | |
976 | } | |
977 | } | |
978 | ||
12676608 LT |
979 | /* |
980 | * Do we have another file that has the beginning components being a | |
981 | * proper superset of the name we're trying to add? | |
0f1e4f04 | 982 | */ |
4aab5b46 JH |
983 | static int has_file_name(struct index_state *istate, |
984 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
0f1e4f04 | 985 | { |
12676608 LT |
986 | int retval = 0; |
987 | int len = ce_namelen(ce); | |
b155725d | 988 | int stage = ce_stage(ce); |
12676608 | 989 | const char *name = ce->name; |
0f1e4f04 | 990 | |
4aab5b46 JH |
991 | while (pos < istate->cache_nr) { |
992 | struct cache_entry *p = istate->cache[pos++]; | |
0f1e4f04 | 993 | |
12676608 | 994 | if (len >= ce_namelen(p)) |
0f1e4f04 | 995 | break; |
12676608 LT |
996 | if (memcmp(name, p->name, len)) |
997 | break; | |
b155725d JH |
998 | if (ce_stage(p) != stage) |
999 | continue; | |
12676608 LT |
1000 | if (p->name[len] != '/') |
1001 | continue; | |
7a51ed66 | 1002 | if (p->ce_flags & CE_REMOVE) |
21cd8d00 | 1003 | continue; |
12676608 LT |
1004 | retval = -1; |
1005 | if (!ok_to_replace) | |
1006 | break; | |
4aab5b46 | 1007 | remove_index_entry_at(istate, --pos); |
0f1e4f04 | 1008 | } |
12676608 LT |
1009 | return retval; |
1010 | } | |
0f1e4f04 | 1011 | |
a6db3fbb JH |
1012 | |
1013 | /* | |
1014 | * Like strcmp(), but also return the offset of the first change. | |
1015 | * If strings are equal, return the length. | |
1016 | */ | |
1017 | int strcmp_offset(const char *s1, const char *s2, size_t *first_change) | |
1018 | { | |
1019 | size_t k; | |
1020 | ||
1021 | if (!first_change) | |
1022 | return strcmp(s1, s2); | |
1023 | ||
1024 | for (k = 0; s1[k] == s2[k]; k++) | |
1025 | if (s1[k] == '\0') | |
1026 | break; | |
1027 | ||
1028 | *first_change = k; | |
1029 | return (unsigned char)s1[k] - (unsigned char)s2[k]; | |
1030 | } | |
1031 | ||
12676608 LT |
1032 | /* |
1033 | * Do we have another file with a pathname that is a proper | |
1034 | * subset of the name we're trying to add? | |
06b6d81b JH |
1035 | * |
1036 | * That is, is there another file in the index with a path | |
1037 | * that matches a sub-directory in the given entry? | |
12676608 | 1038 | */ |
4aab5b46 JH |
1039 | static int has_dir_name(struct index_state *istate, |
1040 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
12676608 LT |
1041 | { |
1042 | int retval = 0; | |
b155725d | 1043 | int stage = ce_stage(ce); |
12676608 LT |
1044 | const char *name = ce->name; |
1045 | const char *slash = name + ce_namelen(ce); | |
06b6d81b JH |
1046 | size_t len_eq_last; |
1047 | int cmp_last = 0; | |
1048 | ||
1049 | /* | |
1050 | * We are frequently called during an iteration on a sorted | |
1051 | * list of pathnames and while building a new index. Therefore, | |
1052 | * there is a high probability that this entry will eventually | |
1053 | * be appended to the index, rather than inserted in the middle. | |
1054 | * If we can confirm that, we can avoid binary searches on the | |
1055 | * components of the pathname. | |
1056 | * | |
1057 | * Compare the entry's full path with the last path in the index. | |
1058 | */ | |
1059 | if (istate->cache_nr > 0) { | |
1060 | cmp_last = strcmp_offset(name, | |
1061 | istate->cache[istate->cache_nr - 1]->name, | |
1062 | &len_eq_last); | |
1063 | if (cmp_last > 0) { | |
1064 | if (len_eq_last == 0) { | |
1065 | /* | |
1066 | * The entry sorts AFTER the last one in the | |
1067 | * index and their paths have no common prefix, | |
1068 | * so there cannot be a F/D conflict. | |
1069 | */ | |
1070 | return retval; | |
1071 | } else { | |
1072 | /* | |
1073 | * The entry sorts AFTER the last one in the | |
1074 | * index, but has a common prefix. Fall through | |
1075 | * to the loop below to disect the entry's path | |
1076 | * and see where the difference is. | |
1077 | */ | |
1078 | } | |
1079 | } else if (cmp_last == 0) { | |
1080 | /* | |
1081 | * The entry exactly matches the last one in the | |
1082 | * index, but because of multiple stage and CE_REMOVE | |
1083 | * items, we fall through and let the regular search | |
1084 | * code handle it. | |
1085 | */ | |
1086 | } | |
1087 | } | |
0f1e4f04 | 1088 | |
12676608 | 1089 | for (;;) { |
b986df5c | 1090 | size_t len; |
0f1e4f04 | 1091 | |
12676608 LT |
1092 | for (;;) { |
1093 | if (*--slash == '/') | |
1094 | break; | |
1095 | if (slash <= ce->name) | |
1096 | return retval; | |
1097 | } | |
1098 | len = slash - name; | |
0f1e4f04 | 1099 | |
b986df5c JH |
1100 | if (cmp_last > 0) { |
1101 | /* | |
1102 | * (len + 1) is a directory boundary (including | |
1103 | * the trailing slash). And since the loop is | |
1104 | * decrementing "slash", the first iteration is | |
1105 | * the longest directory prefix; subsequent | |
1106 | * iterations consider parent directories. | |
1107 | */ | |
1108 | ||
1109 | if (len + 1 <= len_eq_last) { | |
1110 | /* | |
1111 | * The directory prefix (including the trailing | |
1112 | * slash) also appears as a prefix in the last | |
1113 | * entry, so the remainder cannot collide (because | |
1114 | * strcmp said the whole path was greater). | |
1115 | * | |
1116 | * EQ: last: xxx/A | |
1117 | * this: xxx/B | |
1118 | * | |
1119 | * LT: last: xxx/file_A | |
1120 | * this: xxx/file_B | |
1121 | */ | |
1122 | return retval; | |
1123 | } | |
1124 | ||
1125 | if (len > len_eq_last) { | |
1126 | /* | |
1127 | * This part of the directory prefix (excluding | |
1128 | * the trailing slash) is longer than the known | |
1129 | * equal portions, so this sub-directory cannot | |
1130 | * collide with a file. | |
1131 | * | |
1132 | * GT: last: xxxA | |
1133 | * this: xxxB/file | |
1134 | */ | |
1135 | return retval; | |
1136 | } | |
1137 | ||
1138 | if (istate->cache_nr > 0 && | |
1139 | ce_namelen(istate->cache[istate->cache_nr - 1]) > len) { | |
1140 | /* | |
1141 | * The directory prefix lines up with part of | |
1142 | * a longer file or directory name, but sorts | |
1143 | * after it, so this sub-directory cannot | |
1144 | * collide with a file. | |
1145 | * | |
1146 | * last: xxx/yy-file (because '-' sorts before '/') | |
1147 | * this: xxx/yy/abc | |
1148 | */ | |
1149 | return retval; | |
1150 | } | |
1151 | ||
1152 | /* | |
1153 | * This is a possible collision. Fall through and | |
1154 | * let the regular search code handle it. | |
1155 | * | |
1156 | * last: xxx | |
1157 | * this: xxx/file | |
1158 | */ | |
1159 | } | |
1160 | ||
b60e188c | 1161 | pos = index_name_stage_pos(istate, name, len, stage); |
12676608 | 1162 | if (pos >= 0) { |
21cd8d00 JH |
1163 | /* |
1164 | * Found one, but not so fast. This could | |
1165 | * be a marker that says "I was here, but | |
1166 | * I am being removed". Such an entry is | |
1167 | * not a part of the resulting tree, and | |
1168 | * it is Ok to have a directory at the same | |
1169 | * path. | |
1170 | */ | |
077c48df | 1171 | if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) { |
21cd8d00 JH |
1172 | retval = -1; |
1173 | if (!ok_to_replace) | |
1174 | break; | |
4aab5b46 | 1175 | remove_index_entry_at(istate, pos); |
21cd8d00 JH |
1176 | continue; |
1177 | } | |
12676608 | 1178 | } |
21cd8d00 JH |
1179 | else |
1180 | pos = -pos-1; | |
12676608 LT |
1181 | |
1182 | /* | |
1183 | * Trivial optimization: if we find an entry that | |
1184 | * already matches the sub-directory, then we know | |
b155725d | 1185 | * we're ok, and we can exit. |
12676608 | 1186 | */ |
4aab5b46 JH |
1187 | while (pos < istate->cache_nr) { |
1188 | struct cache_entry *p = istate->cache[pos]; | |
b155725d JH |
1189 | if ((ce_namelen(p) <= len) || |
1190 | (p->name[len] != '/') || | |
1191 | memcmp(p->name, name, len)) | |
1192 | break; /* not our subdirectory */ | |
077c48df JH |
1193 | if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE)) |
1194 | /* | |
1195 | * p is at the same stage as our entry, and | |
b155725d JH |
1196 | * is a subdirectory of what we are looking |
1197 | * at, so we cannot have conflicts at our | |
1198 | * level or anything shorter. | |
1199 | */ | |
1200 | return retval; | |
1201 | pos++; | |
192268c1 | 1202 | } |
0f1e4f04 | 1203 | } |
12676608 LT |
1204 | return retval; |
1205 | } | |
1206 | ||
1207 | /* We may be in a situation where we already have path/file and path | |
1208 | * is being added, or we already have path and path/file is being | |
1209 | * added. Either one would result in a nonsense tree that has path | |
1210 | * twice when git-write-tree tries to write it out. Prevent it. | |
a6080a0a | 1211 | * |
12676608 LT |
1212 | * If ok-to-replace is specified, we remove the conflicting entries |
1213 | * from the cache so the caller should recompute the insert position. | |
1214 | * When this happens, we return non-zero. | |
1215 | */ | |
4aab5b46 JH |
1216 | static int check_file_directory_conflict(struct index_state *istate, |
1217 | const struct cache_entry *ce, | |
1218 | int pos, int ok_to_replace) | |
12676608 | 1219 | { |
21cd8d00 JH |
1220 | int retval; |
1221 | ||
1222 | /* | |
1223 | * When ce is an "I am going away" entry, we allow it to be added | |
1224 | */ | |
7a51ed66 | 1225 | if (ce->ce_flags & CE_REMOVE) |
21cd8d00 JH |
1226 | return 0; |
1227 | ||
12676608 LT |
1228 | /* |
1229 | * We check if the path is a sub-path of a subsequent pathname | |
1230 | * first, since removing those will not change the position | |
21cd8d00 | 1231 | * in the array. |
12676608 | 1232 | */ |
4aab5b46 | 1233 | retval = has_file_name(istate, ce, pos, ok_to_replace); |
21cd8d00 | 1234 | |
12676608 LT |
1235 | /* |
1236 | * Then check if the path might have a clashing sub-directory | |
1237 | * before it. | |
1238 | */ | |
4aab5b46 | 1239 | return retval + has_dir_name(istate, ce, pos, ok_to_replace); |
0f1e4f04 JH |
1240 | } |
1241 | ||
af3785dc | 1242 | static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option) |
197ee8c9 LT |
1243 | { |
1244 | int pos; | |
192268c1 JH |
1245 | int ok_to_add = option & ADD_CACHE_OK_TO_ADD; |
1246 | int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE; | |
b155725d | 1247 | int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK; |
39425819 | 1248 | int new_only = option & ADD_CACHE_NEW_ONLY; |
5f73076c | 1249 | |
ce7c614b NTND |
1250 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1251 | cache_tree_invalidate_path(istate, ce->name); | |
e5494631 JH |
1252 | |
1253 | /* | |
1254 | * If this entry's path sorts after the last entry in the index, | |
1255 | * we can avoid searching for it. | |
1256 | */ | |
1257 | if (istate->cache_nr > 0 && | |
1258 | strcmp(ce->name, istate->cache[istate->cache_nr - 1]->name) > 0) | |
1259 | pos = -istate->cache_nr - 1; | |
1260 | else | |
1261 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce)); | |
197ee8c9 | 1262 | |
3e09cdfd | 1263 | /* existing match? Just replace it. */ |
76e7f4ec | 1264 | if (pos >= 0) { |
39425819 JH |
1265 | if (!new_only) |
1266 | replace_index_entry(istate, pos, ce); | |
197ee8c9 LT |
1267 | return 0; |
1268 | } | |
76e7f4ec | 1269 | pos = -pos-1; |
197ee8c9 | 1270 | |
ffcc9ba7 NTND |
1271 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1272 | untracked_cache_add_to_index(istate, ce->name); | |
e931371a | 1273 | |
7b937ca3 LT |
1274 | /* |
1275 | * Inserting a merged entry ("stage 0") into the index | |
1276 | * will always replace all non-merged entries.. | |
1277 | */ | |
4aab5b46 JH |
1278 | if (pos < istate->cache_nr && ce_stage(ce) == 0) { |
1279 | while (ce_same_name(istate->cache[pos], ce)) { | |
7b937ca3 | 1280 | ok_to_add = 1; |
4aab5b46 | 1281 | if (!remove_index_entry_at(istate, pos)) |
7b937ca3 LT |
1282 | break; |
1283 | } | |
1284 | } | |
1285 | ||
121481ab LT |
1286 | if (!ok_to_add) |
1287 | return -1; | |
10ecfa76 | 1288 | if (!verify_path(ce->name, ce->ce_mode)) |
7e7abea9 | 1289 | return error("Invalid path '%s'", ce->name); |
121481ab | 1290 | |
3e09cdfd | 1291 | if (!skip_df_check && |
4aab5b46 | 1292 | check_file_directory_conflict(istate, ce, pos, ok_to_replace)) { |
192268c1 | 1293 | if (!ok_to_replace) |
4aab5b46 JH |
1294 | return error("'%s' appears as both a file and as a directory", |
1295 | ce->name); | |
b60e188c | 1296 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce)); |
192268c1 JH |
1297 | pos = -pos-1; |
1298 | } | |
af3785dc JH |
1299 | return pos + 1; |
1300 | } | |
1301 | ||
1302 | int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option) | |
1303 | { | |
1304 | int pos; | |
1305 | ||
1306 | if (option & ADD_CACHE_JUST_APPEND) | |
1307 | pos = istate->cache_nr; | |
1308 | else { | |
1309 | int ret; | |
1310 | ret = add_index_entry_with_check(istate, ce, option); | |
1311 | if (ret <= 0) | |
1312 | return ret; | |
1313 | pos = ret - 1; | |
1314 | } | |
0f1e4f04 | 1315 | |
197ee8c9 | 1316 | /* Make sure the array is big enough .. */ |
999f5660 | 1317 | ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc); |
197ee8c9 LT |
1318 | |
1319 | /* Add it in.. */ | |
4aab5b46 | 1320 | istate->cache_nr++; |
af3785dc | 1321 | if (istate->cache_nr > pos + 1) |
f919ffeb SG |
1322 | MOVE_ARRAY(istate->cache + pos + 1, istate->cache + pos, |
1323 | istate->cache_nr - pos - 1); | |
cf558704 | 1324 | set_index_entry(istate, pos, ce); |
e636a7b4 | 1325 | istate->cache_changed |= CE_ENTRY_ADDED; |
197ee8c9 LT |
1326 | return 0; |
1327 | } | |
1328 | ||
405e5b2f LT |
1329 | /* |
1330 | * "refresh" does not calculate a new sha1 file or bring the | |
1331 | * cache up-to-date for mode/content changes. But what it | |
1332 | * _does_ do is to "re-match" the stat information of a file | |
1333 | * with the cache, so that you can refresh the cache for a | |
1334 | * file that hasn't been changed but where the stat entry is | |
1335 | * out of date. | |
1336 | * | |
1337 | * For example, you'd want to do this after doing a "git-read-tree", | |
1338 | * to link up the stat cache details with the proper files. | |
1339 | */ | |
4aab5b46 | 1340 | static struct cache_entry *refresh_cache_ent(struct index_state *istate, |
4bd5b7da | 1341 | struct cache_entry *ce, |
d05e6970 JK |
1342 | unsigned int options, int *err, |
1343 | int *changed_ret) | |
405e5b2f LT |
1344 | { |
1345 | struct stat st; | |
1346 | struct cache_entry *updated; | |
a849735b | 1347 | int changed; |
25762726 | 1348 | int refresh = options & CE_MATCH_REFRESH; |
4bd5b7da | 1349 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 1350 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
2e2e7ec1 | 1351 | int ignore_missing = options & CE_MATCH_IGNORE_MISSING; |
883e248b | 1352 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
405e5b2f | 1353 | |
25762726 | 1354 | if (!refresh || ce_uptodate(ce)) |
eadb5831 JH |
1355 | return ce; |
1356 | ||
883e248b BP |
1357 | if (!ignore_fsmonitor) |
1358 | refresh_fsmonitor(istate); | |
aa9349d4 | 1359 | /* |
56cac48c NTND |
1360 | * CE_VALID or CE_SKIP_WORKTREE means the user promised us |
1361 | * that the change to the work tree does not matter and told | |
1362 | * us not to worry. | |
aa9349d4 | 1363 | */ |
56cac48c NTND |
1364 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) { |
1365 | ce_mark_uptodate(ce); | |
1366 | return ce; | |
1367 | } | |
aa9349d4 MSO |
1368 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) { |
1369 | ce_mark_uptodate(ce); | |
1370 | return ce; | |
1371 | } | |
883e248b BP |
1372 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) { |
1373 | ce_mark_uptodate(ce); | |
1374 | return ce; | |
1375 | } | |
aa9349d4 | 1376 | |
ccad42d4 RS |
1377 | if (has_symlink_leading_path(ce->name, ce_namelen(ce))) { |
1378 | if (ignore_missing) | |
1379 | return ce; | |
1380 | if (err) | |
1381 | *err = ENOENT; | |
1382 | return NULL; | |
1383 | } | |
1384 | ||
8fd2cb40 | 1385 | if (lstat(ce->name, &st) < 0) { |
2e2e7ec1 BK |
1386 | if (ignore_missing && errno == ENOENT) |
1387 | return ce; | |
ec0cc704 JH |
1388 | if (err) |
1389 | *err = errno; | |
8fd2cb40 JS |
1390 | return NULL; |
1391 | } | |
405e5b2f | 1392 | |
4bd5b7da | 1393 | changed = ie_match_stat(istate, ce, &st, options); |
d05e6970 JK |
1394 | if (changed_ret) |
1395 | *changed_ret = changed; | |
405e5b2f | 1396 | if (!changed) { |
4bd5b7da JH |
1397 | /* |
1398 | * The path is unchanged. If we were told to ignore | |
1399 | * valid bit, then we did the actual stat check and | |
1400 | * found that the entry is unmodified. If the entry | |
1401 | * is not marked VALID, this is the place to mark it | |
1402 | * valid again, under "assume unchanged" mode. | |
1403 | */ | |
1404 | if (ignore_valid && assume_unchanged && | |
7a51ed66 | 1405 | !(ce->ce_flags & CE_VALID)) |
405e5b2f | 1406 | ; /* mark this one VALID again */ |
eadb5831 JH |
1407 | else { |
1408 | /* | |
1409 | * We do not mark the index itself "modified" | |
1410 | * because CE_UPTODATE flag is in-core only; | |
1411 | * we are not going to write this change out. | |
1412 | */ | |
883e248b | 1413 | if (!S_ISGITLINK(ce->ce_mode)) { |
125fd984 | 1414 | ce_mark_uptodate(ce); |
883e248b BP |
1415 | mark_fsmonitor_valid(ce); |
1416 | } | |
8fd2cb40 | 1417 | return ce; |
eadb5831 | 1418 | } |
405e5b2f LT |
1419 | } |
1420 | ||
4bd5b7da | 1421 | if (ie_modified(istate, ce, &st, options)) { |
ec0cc704 JH |
1422 | if (err) |
1423 | *err = EINVAL; | |
8fd2cb40 JS |
1424 | return NULL; |
1425 | } | |
405e5b2f | 1426 | |
a849735b | 1427 | updated = make_empty_cache_entry(istate, ce_namelen(ce)); |
0e267b7a BP |
1428 | copy_cache_entry(updated, ce); |
1429 | memcpy(updated->name, ce->name, ce->ce_namelen + 1); | |
405e5b2f | 1430 | fill_stat_cache_info(updated, &st); |
4bd5b7da JH |
1431 | /* |
1432 | * If ignore_valid is not set, we should leave CE_VALID bit | |
1433 | * alone. Otherwise, paths marked with --no-assume-unchanged | |
1434 | * (i.e. things to be edited) will reacquire CE_VALID bit | |
1435 | * automatically, which is not really what we want. | |
405e5b2f | 1436 | */ |
4bd5b7da | 1437 | if (!ignore_valid && assume_unchanged && |
7a51ed66 LT |
1438 | !(ce->ce_flags & CE_VALID)) |
1439 | updated->ce_flags &= ~CE_VALID; | |
405e5b2f | 1440 | |
e636a7b4 | 1441 | /* istate->cache_changed is updated in the caller */ |
405e5b2f LT |
1442 | return updated; |
1443 | } | |
1444 | ||
3deffc52 | 1445 | static void show_file(const char * fmt, const char * name, int in_porcelain, |
046613c5 | 1446 | int * first, const char *header_msg) |
3deffc52 MM |
1447 | { |
1448 | if (in_porcelain && *first && header_msg) { | |
1449 | printf("%s\n", header_msg); | |
cd2b8ae9 | 1450 | *first = 0; |
3deffc52 MM |
1451 | } |
1452 | printf(fmt, name); | |
1453 | } | |
1454 | ||
9b2d6149 NTND |
1455 | int refresh_index(struct index_state *istate, unsigned int flags, |
1456 | const struct pathspec *pathspec, | |
046613c5 | 1457 | char *seen, const char *header_msg) |
405e5b2f LT |
1458 | { |
1459 | int i; | |
1460 | int has_errors = 0; | |
1461 | int really = (flags & REFRESH_REALLY) != 0; | |
1462 | int allow_unmerged = (flags & REFRESH_UNMERGED) != 0; | |
1463 | int quiet = (flags & REFRESH_QUIET) != 0; | |
1464 | int not_new = (flags & REFRESH_IGNORE_MISSING) != 0; | |
5fdeacb0 | 1465 | int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0; |
3deffc52 MM |
1466 | int first = 1; |
1467 | int in_porcelain = (flags & REFRESH_IN_PORCELAIN); | |
25762726 BK |
1468 | unsigned int options = (CE_MATCH_REFRESH | |
1469 | (really ? CE_MATCH_IGNORE_VALID : 0) | | |
2e2e7ec1 | 1470 | (not_new ? CE_MATCH_IGNORE_MISSING : 0)); |
4bd4e730 | 1471 | const char *modified_fmt; |
73b7eae6 JK |
1472 | const char *deleted_fmt; |
1473 | const char *typechange_fmt; | |
1474 | const char *added_fmt; | |
4bd4e730 | 1475 | const char *unmerged_fmt; |
ca54d9ba | 1476 | uint64_t start = getnanotime(); |
405e5b2f | 1477 | |
4bd4e730 | 1478 | modified_fmt = (in_porcelain ? "M\t%s\n" : "%s: needs update\n"); |
73b7eae6 JK |
1479 | deleted_fmt = (in_porcelain ? "D\t%s\n" : "%s: needs update\n"); |
1480 | typechange_fmt = (in_porcelain ? "T\t%s\n" : "%s needs update\n"); | |
1481 | added_fmt = (in_porcelain ? "A\t%s\n" : "%s needs update\n"); | |
4bd4e730 | 1482 | unmerged_fmt = (in_porcelain ? "U\t%s\n" : "%s: needs merge\n"); |
4aab5b46 | 1483 | for (i = 0; i < istate->cache_nr; i++) { |
285c2e25 | 1484 | struct cache_entry *ce, *new_entry; |
ec0cc704 | 1485 | int cache_errno = 0; |
73b7eae6 | 1486 | int changed = 0; |
3d1f148c | 1487 | int filtered = 0; |
ec0cc704 | 1488 | |
4aab5b46 | 1489 | ce = istate->cache[i]; |
5fdeacb0 JS |
1490 | if (ignore_submodules && S_ISGITLINK(ce->ce_mode)) |
1491 | continue; | |
1492 | ||
429bb40a | 1493 | if (pathspec && !ce_path_match(ce, pathspec, seen)) |
3d1f148c JH |
1494 | filtered = 1; |
1495 | ||
405e5b2f | 1496 | if (ce_stage(ce)) { |
4aab5b46 JH |
1497 | while ((i < istate->cache_nr) && |
1498 | ! strcmp(istate->cache[i]->name, ce->name)) | |
405e5b2f LT |
1499 | i++; |
1500 | i--; | |
1501 | if (allow_unmerged) | |
1502 | continue; | |
3d1f148c JH |
1503 | if (!filtered) |
1504 | show_file(unmerged_fmt, ce->name, in_porcelain, | |
1505 | &first, header_msg); | |
405e5b2f LT |
1506 | has_errors = 1; |
1507 | continue; | |
1508 | } | |
1509 | ||
3d1f148c | 1510 | if (filtered) |
d616813d AJ |
1511 | continue; |
1512 | ||
285c2e25 BW |
1513 | new_entry = refresh_cache_ent(istate, ce, options, &cache_errno, &changed); |
1514 | if (new_entry == ce) | |
405e5b2f | 1515 | continue; |
285c2e25 | 1516 | if (!new_entry) { |
73b7eae6 JK |
1517 | const char *fmt; |
1518 | ||
8fd2cb40 | 1519 | if (really && cache_errno == EINVAL) { |
405e5b2f LT |
1520 | /* If we are doing --really-refresh that |
1521 | * means the index is not valid anymore. | |
1522 | */ | |
7a51ed66 | 1523 | ce->ce_flags &= ~CE_VALID; |
078a58e8 | 1524 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 1525 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 1526 | istate->cache_changed |= CE_ENTRY_CHANGED; |
405e5b2f LT |
1527 | } |
1528 | if (quiet) | |
1529 | continue; | |
73b7eae6 JK |
1530 | |
1531 | if (cache_errno == ENOENT) | |
1532 | fmt = deleted_fmt; | |
895ff3b2 | 1533 | else if (ce_intent_to_add(ce)) |
73b7eae6 JK |
1534 | fmt = added_fmt; /* must be before other checks */ |
1535 | else if (changed & TYPE_CHANGED) | |
1536 | fmt = typechange_fmt; | |
1537 | else | |
1538 | fmt = modified_fmt; | |
1539 | show_file(fmt, | |
1540 | ce->name, in_porcelain, &first, header_msg); | |
405e5b2f LT |
1541 | has_errors = 1; |
1542 | continue; | |
1543 | } | |
cf558704 | 1544 | |
285c2e25 | 1545 | replace_index_entry(istate, i, new_entry); |
405e5b2f | 1546 | } |
ca54d9ba | 1547 | trace_performance_since(start, "refresh index"); |
405e5b2f LT |
1548 | return has_errors; |
1549 | } | |
1550 | ||
768d7965 JM |
1551 | struct cache_entry *refresh_cache_entry(struct index_state *istate, |
1552 | struct cache_entry *ce, | |
1553 | unsigned int options) | |
ec0cc704 | 1554 | { |
768d7965 | 1555 | return refresh_cache_ent(istate, ce, options, NULL, NULL); |
ec0cc704 JH |
1556 | } |
1557 | ||
db3b313c JH |
1558 | |
1559 | /***************************************************************** | |
1560 | * Index File I/O | |
1561 | *****************************************************************/ | |
1562 | ||
9d227781 JH |
1563 | #define INDEX_FORMAT_DEFAULT 3 |
1564 | ||
136347d7 TG |
1565 | static unsigned int get_index_format_default(void) |
1566 | { | |
1567 | char *envversion = getenv("GIT_INDEX_VERSION"); | |
3c09d684 | 1568 | char *endp; |
b27a5720 | 1569 | int value; |
3c09d684 TG |
1570 | unsigned int version = INDEX_FORMAT_DEFAULT; |
1571 | ||
136347d7 | 1572 | if (!envversion) { |
b27a5720 TA |
1573 | if (!git_config_get_int("index.version", &value)) |
1574 | version = value; | |
3c09d684 TG |
1575 | if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { |
1576 | warning(_("index.version set, but the value is invalid.\n" | |
136347d7 | 1577 | "Using version %i"), INDEX_FORMAT_DEFAULT); |
3c09d684 | 1578 | return INDEX_FORMAT_DEFAULT; |
136347d7 TG |
1579 | } |
1580 | return version; | |
1581 | } | |
3c09d684 TG |
1582 | |
1583 | version = strtoul(envversion, &endp, 10); | |
1584 | if (*endp || | |
1585 | version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { | |
1586 | warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n" | |
1587 | "Using version %i"), INDEX_FORMAT_DEFAULT); | |
1588 | version = INDEX_FORMAT_DEFAULT; | |
1589 | } | |
1590 | return version; | |
136347d7 TG |
1591 | } |
1592 | ||
db3b313c JH |
1593 | /* |
1594 | * dev/ino/uid/gid/size are also just tracked to the low 32 bits | |
1595 | * Again - this is just a (very strong in practice) heuristic that | |
1596 | * the inode hasn't changed. | |
1597 | * | |
1598 | * We save the fields in big-endian order to allow using the | |
1599 | * index file over NFS transparently. | |
1600 | */ | |
1601 | struct ondisk_cache_entry { | |
1602 | struct cache_time ctime; | |
1603 | struct cache_time mtime; | |
7800c1eb TG |
1604 | uint32_t dev; |
1605 | uint32_t ino; | |
1606 | uint32_t mode; | |
1607 | uint32_t uid; | |
1608 | uint32_t gid; | |
1609 | uint32_t size; | |
db3b313c | 1610 | unsigned char sha1[20]; |
7800c1eb | 1611 | uint16_t flags; |
db3b313c JH |
1612 | char name[FLEX_ARRAY]; /* more */ |
1613 | }; | |
1614 | ||
1615 | /* | |
1616 | * This struct is used when CE_EXTENDED bit is 1 | |
1617 | * The struct must match ondisk_cache_entry exactly from | |
1618 | * ctime till flags | |
1619 | */ | |
1620 | struct ondisk_cache_entry_extended { | |
1621 | struct cache_time ctime; | |
1622 | struct cache_time mtime; | |
7800c1eb TG |
1623 | uint32_t dev; |
1624 | uint32_t ino; | |
1625 | uint32_t mode; | |
1626 | uint32_t uid; | |
1627 | uint32_t gid; | |
1628 | uint32_t size; | |
db3b313c | 1629 | unsigned char sha1[20]; |
7800c1eb TG |
1630 | uint16_t flags; |
1631 | uint16_t flags2; | |
db3b313c JH |
1632 | char name[FLEX_ARRAY]; /* more */ |
1633 | }; | |
1634 | ||
6c9cd161 | 1635 | /* These are only used for v3 or lower */ |
ce012deb | 1636 | #define align_padding_size(size, len) ((size + (len) + 8) & ~7) - (size + len) |
db3b313c JH |
1637 | #define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,name) + (len) + 8) & ~7) |
1638 | #define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len) | |
1639 | #define ondisk_cache_entry_extended_size(len) align_flex_name(ondisk_cache_entry_extended,len) | |
1640 | #define ondisk_ce_size(ce) (((ce)->ce_flags & CE_EXTENDED) ? \ | |
1641 | ondisk_cache_entry_extended_size(ce_namelen(ce)) : \ | |
1642 | ondisk_cache_entry_size(ce_namelen(ce))) | |
1643 | ||
a33fc72f JH |
1644 | /* Allow fsck to force verification of the index checksum. */ |
1645 | int verify_index_checksum; | |
1646 | ||
00ec50e5 BP |
1647 | /* Allow fsck to force verification of the cache entry order. */ |
1648 | int verify_ce_order; | |
1649 | ||
bad68ec9 | 1650 | static int verify_hdr(struct cache_header *hdr, unsigned long size) |
e83c5163 | 1651 | { |
aab61359 | 1652 | git_hash_ctx c; |
1653 | unsigned char hash[GIT_MAX_RAWSZ]; | |
0136bac9 | 1654 | int hdr_version; |
e83c5163 | 1655 | |
ccc4feb5 | 1656 | if (hdr->hdr_signature != htonl(CACHE_SIGNATURE)) |
e83c5163 | 1657 | return error("bad signature"); |
0136bac9 | 1658 | hdr_version = ntohl(hdr->hdr_version); |
b82a7b5b | 1659 | if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version) |
0136bac9 | 1660 | return error("bad index version %d", hdr_version); |
a33fc72f JH |
1661 | |
1662 | if (!verify_index_checksum) | |
1663 | return 0; | |
1664 | ||
aab61359 | 1665 | the_hash_algo->init_fn(&c); |
1666 | the_hash_algo->update_fn(&c, hdr, size - the_hash_algo->rawsz); | |
1667 | the_hash_algo->final_fn(hash, &c); | |
1668 | if (hashcmp(hash, (unsigned char *)hdr + size - the_hash_algo->rawsz)) | |
ca9be054 | 1669 | return error("bad index file sha1 signature"); |
e83c5163 LT |
1670 | return 0; |
1671 | } | |
1672 | ||
4aab5b46 JH |
1673 | static int read_index_extension(struct index_state *istate, |
1674 | const char *ext, void *data, unsigned long sz) | |
bad68ec9 JH |
1675 | { |
1676 | switch (CACHE_EXT(ext)) { | |
1677 | case CACHE_EXT_TREE: | |
4aab5b46 | 1678 | istate->cache_tree = cache_tree_read(data, sz); |
bad68ec9 | 1679 | break; |
cfc5789a JH |
1680 | case CACHE_EXT_RESOLVE_UNDO: |
1681 | istate->resolve_undo = resolve_undo_read(data, sz); | |
1682 | break; | |
5fc2fc8f NTND |
1683 | case CACHE_EXT_LINK: |
1684 | if (read_link_extension(istate, data, sz)) | |
1685 | return -1; | |
1686 | break; | |
f9e6c649 NTND |
1687 | case CACHE_EXT_UNTRACKED: |
1688 | istate->untracked = read_untracked_extension(data, sz); | |
1689 | break; | |
883e248b BP |
1690 | case CACHE_EXT_FSMONITOR: |
1691 | read_fsmonitor_extension(istate, data, sz); | |
1692 | break; | |
bad68ec9 JH |
1693 | default: |
1694 | if (*ext < 'A' || 'Z' < *ext) | |
1695 | return error("index uses %.4s extension, which we do not understand", | |
1696 | ext); | |
1697 | fprintf(stderr, "ignoring %.4s extension\n", ext); | |
1698 | break; | |
1699 | } | |
1700 | return 0; | |
1701 | } | |
1702 | ||
b3e83cc7 | 1703 | int hold_locked_index(struct lock_file *lk, int lock_flags) |
216aab1e | 1704 | { |
b3e83cc7 | 1705 | return hold_lock_file_for_update(lk, get_index_file(), lock_flags); |
216aab1e MH |
1706 | } |
1707 | ||
4aab5b46 | 1708 | int read_index(struct index_state *istate) |
8fd2cb40 | 1709 | { |
a125a223 | 1710 | return read_index_from(istate, get_index_file(), get_git_dir()); |
8fd2cb40 JS |
1711 | } |
1712 | ||
8e72d675 | 1713 | static struct cache_entry *cache_entry_from_ondisk(struct mem_pool *mem_pool, |
a849735b | 1714 | struct ondisk_cache_entry *ondisk, |
3fc22b53 JH |
1715 | unsigned int flags, |
1716 | const char *name, | |
1717 | size_t len) | |
1718 | { | |
8e72d675 | 1719 | struct cache_entry *ce = mem_pool__ce_alloc(mem_pool, len); |
3fc22b53 | 1720 | |
c3d8da57 JK |
1721 | ce->ce_stat_data.sd_ctime.sec = get_be32(&ondisk->ctime.sec); |
1722 | ce->ce_stat_data.sd_mtime.sec = get_be32(&ondisk->mtime.sec); | |
1723 | ce->ce_stat_data.sd_ctime.nsec = get_be32(&ondisk->ctime.nsec); | |
1724 | ce->ce_stat_data.sd_mtime.nsec = get_be32(&ondisk->mtime.nsec); | |
1725 | ce->ce_stat_data.sd_dev = get_be32(&ondisk->dev); | |
1726 | ce->ce_stat_data.sd_ino = get_be32(&ondisk->ino); | |
1727 | ce->ce_mode = get_be32(&ondisk->mode); | |
1728 | ce->ce_stat_data.sd_uid = get_be32(&ondisk->uid); | |
1729 | ce->ce_stat_data.sd_gid = get_be32(&ondisk->gid); | |
1730 | ce->ce_stat_data.sd_size = get_be32(&ondisk->size); | |
b60e188c TG |
1731 | ce->ce_flags = flags & ~CE_NAMEMASK; |
1732 | ce->ce_namelen = len; | |
5fc2fc8f | 1733 | ce->index = 0; |
99d1a986 | 1734 | hashcpy(ce->oid.hash, ondisk->sha1); |
3fc22b53 JH |
1735 | memcpy(ce->name, name, len); |
1736 | ce->name[len] = '\0'; | |
1737 | return ce; | |
1738 | } | |
1739 | ||
6c9cd161 JH |
1740 | /* |
1741 | * Adjacent cache entries tend to share the leading paths, so it makes | |
1742 | * sense to only store the differences in later entries. In the v4 | |
1743 | * on-disk format of the index, each on-disk cache entry stores the | |
1744 | * number of bytes to be stripped from the end of the previous name, | |
1745 | * and the bytes to append to the result, to come up with its name. | |
1746 | */ | |
1747 | static unsigned long expand_name_field(struct strbuf *name, const char *cp_) | |
1748 | { | |
1749 | const unsigned char *ep, *cp = (const unsigned char *)cp_; | |
1750 | size_t len = decode_varint(&cp); | |
1751 | ||
1752 | if (name->len < len) | |
1753 | die("malformed name field in the index"); | |
1754 | strbuf_remove(name, name->len - len, len); | |
1755 | for (ep = cp; *ep; ep++) | |
1756 | ; /* find the end */ | |
1757 | strbuf_add(name, cp, ep - cp); | |
1758 | return (const char *)ep + 1 - cp_; | |
1759 | } | |
1760 | ||
8e72d675 | 1761 | static struct cache_entry *create_from_disk(struct mem_pool *mem_pool, |
a849735b | 1762 | struct ondisk_cache_entry *ondisk, |
6c9cd161 JH |
1763 | unsigned long *ent_size, |
1764 | struct strbuf *previous_name) | |
7a51ed66 | 1765 | { |
debed2a6 | 1766 | struct cache_entry *ce; |
7fec10b7 | 1767 | size_t len; |
06aaaa0b | 1768 | const char *name; |
debed2a6 | 1769 | unsigned int flags; |
7fec10b7 | 1770 | |
7a51ed66 | 1771 | /* On-disk flags are just 16 bits */ |
c3d8da57 | 1772 | flags = get_be16(&ondisk->flags); |
debed2a6 | 1773 | len = flags & CE_NAMEMASK; |
7fec10b7 | 1774 | |
debed2a6 | 1775 | if (flags & CE_EXTENDED) { |
06aaaa0b NTND |
1776 | struct ondisk_cache_entry_extended *ondisk2; |
1777 | int extended_flags; | |
1778 | ondisk2 = (struct ondisk_cache_entry_extended *)ondisk; | |
c3d8da57 | 1779 | extended_flags = get_be16(&ondisk2->flags2) << 16; |
06aaaa0b NTND |
1780 | /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */ |
1781 | if (extended_flags & ~CE_EXTENDED_FLAGS) | |
1782 | die("Unknown index entry format %08x", extended_flags); | |
debed2a6 | 1783 | flags |= extended_flags; |
06aaaa0b NTND |
1784 | name = ondisk2->name; |
1785 | } | |
1786 | else | |
1787 | name = ondisk->name; | |
1788 | ||
6c9cd161 JH |
1789 | if (!previous_name) { |
1790 | /* v3 and earlier */ | |
1791 | if (len == CE_NAMEMASK) | |
1792 | len = strlen(name); | |
8e72d675 | 1793 | ce = cache_entry_from_ondisk(mem_pool, ondisk, flags, name, len); |
6c9cd161 JH |
1794 | |
1795 | *ent_size = ondisk_ce_size(ce); | |
1796 | } else { | |
1797 | unsigned long consumed; | |
1798 | consumed = expand_name_field(previous_name, name); | |
8e72d675 | 1799 | ce = cache_entry_from_ondisk(mem_pool, ondisk, flags, |
6c9cd161 JH |
1800 | previous_name->buf, |
1801 | previous_name->len); | |
1802 | ||
1803 | *ent_size = (name - ((char *)ondisk)) + consumed; | |
1804 | } | |
debed2a6 | 1805 | return ce; |
cf558704 LT |
1806 | } |
1807 | ||
03f15a79 | 1808 | static void check_ce_order(struct index_state *istate) |
15999d0b | 1809 | { |
03f15a79 TG |
1810 | unsigned int i; |
1811 | ||
00ec50e5 BP |
1812 | if (!verify_ce_order) |
1813 | return; | |
1814 | ||
03f15a79 TG |
1815 | for (i = 1; i < istate->cache_nr; i++) { |
1816 | struct cache_entry *ce = istate->cache[i - 1]; | |
1817 | struct cache_entry *next_ce = istate->cache[i]; | |
1818 | int name_compare = strcmp(ce->name, next_ce->name); | |
1819 | ||
1820 | if (0 < name_compare) | |
1821 | die("unordered stage entries in index"); | |
1822 | if (!name_compare) { | |
1823 | if (!ce_stage(ce)) | |
1824 | die("multiple stage entries for merged file '%s'", | |
1825 | ce->name); | |
1826 | if (ce_stage(ce) > ce_stage(next_ce)) | |
1827 | die("unordered stage entries for '%s'", | |
1828 | ce->name); | |
1829 | } | |
15999d0b JSP |
1830 | } |
1831 | } | |
1832 | ||
435ec090 CC |
1833 | static void tweak_untracked_cache(struct index_state *istate) |
1834 | { | |
1835 | switch (git_config_get_untracked_cache()) { | |
1836 | case -1: /* keep: do nothing */ | |
1837 | break; | |
1838 | case 0: /* false */ | |
1839 | remove_untracked_cache(istate); | |
1840 | break; | |
1841 | case 1: /* true */ | |
1842 | add_untracked_cache(istate); | |
1843 | break; | |
1844 | default: /* unknown value: do nothing */ | |
1845 | break; | |
1846 | } | |
1847 | } | |
1848 | ||
43925312 CC |
1849 | static void tweak_split_index(struct index_state *istate) |
1850 | { | |
1851 | switch (git_config_get_split_index()) { | |
1852 | case -1: /* unset: do nothing */ | |
1853 | break; | |
1854 | case 0: /* false */ | |
1855 | remove_split_index(istate); | |
1856 | break; | |
1857 | case 1: /* true */ | |
1858 | add_split_index(istate); | |
1859 | break; | |
1860 | default: /* unknown value: do nothing */ | |
1861 | break; | |
1862 | } | |
1863 | } | |
1864 | ||
435ec090 CC |
1865 | static void post_read_index_from(struct index_state *istate) |
1866 | { | |
1867 | check_ce_order(istate); | |
1868 | tweak_untracked_cache(istate); | |
43925312 | 1869 | tweak_split_index(istate); |
883e248b | 1870 | tweak_fsmonitor(istate); |
435ec090 CC |
1871 | } |
1872 | ||
8e72d675 JM |
1873 | static size_t estimate_cache_size_from_compressed(unsigned int entries) |
1874 | { | |
1875 | return entries * (sizeof(struct cache_entry) + CACHE_ENTRY_PATH_LENGTH); | |
1876 | } | |
1877 | ||
1878 | static size_t estimate_cache_size(size_t ondisk_size, unsigned int entries) | |
1879 | { | |
1880 | long per_entry = sizeof(struct cache_entry) - sizeof(struct ondisk_cache_entry); | |
1881 | ||
1882 | /* | |
1883 | * Account for potential alignment differences. | |
1884 | */ | |
1885 | per_entry += align_padding_size(sizeof(struct cache_entry), -sizeof(struct ondisk_cache_entry)); | |
1886 | return ondisk_size + entries * per_entry; | |
1887 | } | |
1888 | ||
8fd2cb40 | 1889 | /* remember to discard_cache() before reading a different cache! */ |
3e52f70b | 1890 | int do_read_index(struct index_state *istate, const char *path, int must_exist) |
e83c5163 LT |
1891 | { |
1892 | int fd, i; | |
1893 | struct stat st; | |
debed2a6 | 1894 | unsigned long src_offset; |
e83c5163 | 1895 | struct cache_header *hdr; |
7a51ed66 LT |
1896 | void *mmap; |
1897 | size_t mmap_size; | |
6c9cd161 | 1898 | struct strbuf previous_name_buf = STRBUF_INIT, *previous_name; |
e83c5163 | 1899 | |
913e0e99 | 1900 | if (istate->initialized) |
4aab5b46 | 1901 | return istate->cache_nr; |
5d1a5c02 | 1902 | |
fba2f38a KB |
1903 | istate->timestamp.sec = 0; |
1904 | istate->timestamp.nsec = 0; | |
8fd2cb40 | 1905 | fd = open(path, O_RDONLY); |
5d1a5c02 | 1906 | if (fd < 0) { |
5fc2fc8f | 1907 | if (!must_exist && errno == ENOENT) |
5d1a5c02 | 1908 | return 0; |
5fc2fc8f | 1909 | die_errno("%s: index file open failed", path); |
5d1a5c02 | 1910 | } |
e83c5163 | 1911 | |
3511a377 | 1912 | if (fstat(fd, &st)) |
d824cbba | 1913 | die_errno("cannot stat the open index"); |
3511a377 | 1914 | |
7a51ed66 | 1915 | mmap_size = xsize_t(st.st_size); |
aab61359 | 1916 | if (mmap_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
3511a377 LFC |
1917 | die("index file smaller than expected"); |
1918 | ||
a1293ef7 | 1919 | mmap = xmmap(NULL, mmap_size, PROT_READ, MAP_PRIVATE, fd, 0); |
7a51ed66 | 1920 | if (mmap == MAP_FAILED) |
0721c314 | 1921 | die_errno("unable to map index file"); |
57d84f8d | 1922 | close(fd); |
e83c5163 | 1923 | |
7a51ed66 LT |
1924 | hdr = mmap; |
1925 | if (verify_hdr(hdr, mmap_size) < 0) | |
e83c5163 LT |
1926 | goto unmap; |
1927 | ||
75691ea3 | 1928 | hashcpy(istate->oid.hash, (const unsigned char *)hdr + mmap_size - the_hash_algo->rawsz); |
9d227781 | 1929 | istate->version = ntohl(hdr->hdr_version); |
4aab5b46 JH |
1930 | istate->cache_nr = ntohl(hdr->hdr_entries); |
1931 | istate->cache_alloc = alloc_nr(istate->cache_nr); | |
c4aa3167 | 1932 | istate->cache = xcalloc(istate->cache_alloc, sizeof(*istate->cache)); |
913e0e99 | 1933 | istate->initialized = 1; |
7a51ed66 | 1934 | |
8e72d675 | 1935 | if (istate->version == 4) { |
6c9cd161 | 1936 | previous_name = &previous_name_buf; |
8e72d675 JM |
1937 | mem_pool_init(&istate->ce_mem_pool, |
1938 | estimate_cache_size_from_compressed(istate->cache_nr)); | |
1939 | } else { | |
6c9cd161 | 1940 | previous_name = NULL; |
8e72d675 JM |
1941 | mem_pool_init(&istate->ce_mem_pool, |
1942 | estimate_cache_size(mmap_size, istate->cache_nr)); | |
1943 | } | |
6c9cd161 | 1944 | |
7a51ed66 | 1945 | src_offset = sizeof(*hdr); |
4aab5b46 | 1946 | for (i = 0; i < istate->cache_nr; i++) { |
7a51ed66 | 1947 | struct ondisk_cache_entry *disk_ce; |
4aab5b46 | 1948 | struct cache_entry *ce; |
936f53d0 | 1949 | unsigned long consumed; |
4aab5b46 | 1950 | |
7a51ed66 | 1951 | disk_ce = (struct ondisk_cache_entry *)((char *)mmap + src_offset); |
8e72d675 | 1952 | ce = create_from_disk(istate->ce_mem_pool, disk_ce, &consumed, previous_name); |
cf558704 | 1953 | set_index_entry(istate, i, ce); |
7a51ed66 | 1954 | |
936f53d0 | 1955 | src_offset += consumed; |
e83c5163 | 1956 | } |
6c9cd161 | 1957 | strbuf_release(&previous_name_buf); |
fba2f38a | 1958 | istate->timestamp.sec = st.st_mtime; |
c06ff490 | 1959 | istate->timestamp.nsec = ST_MTIME_NSEC(st); |
fba2f38a | 1960 | |
aab61359 | 1961 | while (src_offset <= mmap_size - the_hash_algo->rawsz - 8) { |
bad68ec9 JH |
1962 | /* After an array of active_nr index entries, |
1963 | * there can be arbitrary number of extended | |
1964 | * sections, each of which is prefixed with | |
1965 | * extension name (4-byte) and section length | |
1966 | * in 4-byte network byte order. | |
1967 | */ | |
07cc8eca | 1968 | uint32_t extsize; |
7a51ed66 | 1969 | memcpy(&extsize, (char *)mmap + src_offset + 4, 4); |
bad68ec9 | 1970 | extsize = ntohl(extsize); |
4aab5b46 | 1971 | if (read_index_extension(istate, |
7a51ed66 LT |
1972 | (const char *) mmap + src_offset, |
1973 | (char *) mmap + src_offset + 8, | |
1d7f171c | 1974 | extsize) < 0) |
bad68ec9 | 1975 | goto unmap; |
7a51ed66 LT |
1976 | src_offset += 8; |
1977 | src_offset += extsize; | |
bad68ec9 | 1978 | } |
7a51ed66 | 1979 | munmap(mmap, mmap_size); |
4aab5b46 | 1980 | return istate->cache_nr; |
e83c5163 LT |
1981 | |
1982 | unmap: | |
7a51ed66 | 1983 | munmap(mmap, mmap_size); |
5d1a5c02 | 1984 | die("index file corrupt"); |
e83c5163 LT |
1985 | } |
1986 | ||
0d59ffb4 CC |
1987 | /* |
1988 | * Signal that the shared index is used by updating its mtime. | |
1989 | * | |
1990 | * This way, shared index can be removed if they have not been used | |
1991 | * for some time. | |
1992 | */ | |
a125a223 | 1993 | static void freshen_shared_index(const char *shared_index, int warn) |
0d59ffb4 | 1994 | { |
0d59ffb4 CC |
1995 | if (!check_and_freshen_file(shared_index, 1) && warn) |
1996 | warning("could not freshen shared index '%s'", shared_index); | |
1997 | } | |
1998 | ||
a125a223 TG |
1999 | int read_index_from(struct index_state *istate, const char *path, |
2000 | const char *gitdir) | |
5fc2fc8f | 2001 | { |
ca54d9ba | 2002 | uint64_t start = getnanotime(); |
5fc2fc8f NTND |
2003 | struct split_index *split_index; |
2004 | int ret; | |
2182abd9 | 2005 | char *base_oid_hex; |
a125a223 | 2006 | char *base_path; |
5fc2fc8f NTND |
2007 | |
2008 | /* istate->initialized covers both .git/index and .git/sharedindex.xxx */ | |
2009 | if (istate->initialized) | |
2010 | return istate->cache_nr; | |
2011 | ||
2012 | ret = do_read_index(istate, path, 0); | |
ca54d9ba | 2013 | trace_performance_since(start, "read cache %s", path); |
435ec090 | 2014 | |
5fc2fc8f | 2015 | split_index = istate->split_index; |
2182abd9 | 2016 | if (!split_index || is_null_oid(&split_index->base_oid)) { |
435ec090 | 2017 | post_read_index_from(istate); |
5fc2fc8f | 2018 | return ret; |
03f15a79 | 2019 | } |
5fc2fc8f NTND |
2020 | |
2021 | if (split_index->base) | |
2022 | discard_index(split_index->base); | |
2023 | else | |
2024 | split_index->base = xcalloc(1, sizeof(*split_index->base)); | |
de6ae5f9 | 2025 | |
2182abd9 | 2026 | base_oid_hex = oid_to_hex(&split_index->base_oid); |
2027 | base_path = xstrfmt("%s/sharedindex.%s", gitdir, base_oid_hex); | |
de6ae5f9 | 2028 | ret = do_read_index(split_index->base, base_path, 1); |
75691ea3 | 2029 | if (oidcmp(&split_index->base_oid, &split_index->base->oid)) |
5fc2fc8f | 2030 | die("broken index, expect %s in %s, got %s", |
2182abd9 | 2031 | base_oid_hex, base_path, |
75691ea3 | 2032 | oid_to_hex(&split_index->base->oid)); |
de6ae5f9 | 2033 | |
a125a223 | 2034 | freshen_shared_index(base_path, 0); |
5fc2fc8f | 2035 | merge_base_index(istate); |
435ec090 | 2036 | post_read_index_from(istate); |
ca54d9ba | 2037 | trace_performance_since(start, "read cache %s", base_path); |
a125a223 | 2038 | free(base_path); |
5fc2fc8f NTND |
2039 | return ret; |
2040 | } | |
2041 | ||
fa7b3c2f JH |
2042 | int is_index_unborn(struct index_state *istate) |
2043 | { | |
debed2a6 | 2044 | return (!istate->cache_nr && !istate->timestamp.sec); |
fa7b3c2f JH |
2045 | } |
2046 | ||
4aab5b46 | 2047 | int discard_index(struct index_state *istate) |
6d297f81 | 2048 | { |
8e72d675 JM |
2049 | /* |
2050 | * Cache entries in istate->cache[] should have been allocated | |
2051 | * from the memory pool associated with this index, or from an | |
2052 | * associated split_index. There is no need to free individual | |
2053 | * cache entries. | |
2054 | */ | |
debed2a6 | 2055 | |
cfc5789a | 2056 | resolve_undo_clear_index(istate); |
4aab5b46 JH |
2057 | istate->cache_nr = 0; |
2058 | istate->cache_changed = 0; | |
fba2f38a KB |
2059 | istate->timestamp.sec = 0; |
2060 | istate->timestamp.nsec = 0; | |
2092678c | 2061 | free_name_hash(istate); |
4aab5b46 | 2062 | cache_tree_free(&(istate->cache_tree)); |
913e0e99 | 2063 | istate->initialized = 0; |
6a83d902 | 2064 | FREE_AND_NULL(istate->cache); |
a0fc4db0 | 2065 | istate->cache_alloc = 0; |
5fc2fc8f | 2066 | discard_split_index(istate); |
f9e6c649 NTND |
2067 | free_untracked_cache(istate->untracked); |
2068 | istate->untracked = NULL; | |
8e72d675 JM |
2069 | |
2070 | if (istate->ce_mem_pool) { | |
2071 | mem_pool_discard(istate->ce_mem_pool); | |
2072 | istate->ce_mem_pool = NULL; | |
2073 | } | |
2074 | ||
7a51ed66 | 2075 | return 0; |
6d297f81 JS |
2076 | } |
2077 | ||
d1f128b0 | 2078 | int unmerged_index(const struct index_state *istate) |
94a5728c DB |
2079 | { |
2080 | int i; | |
2081 | for (i = 0; i < istate->cache_nr; i++) { | |
2082 | if (ce_stage(istate->cache[i])) | |
2083 | return 1; | |
2084 | } | |
2085 | return 0; | |
2086 | } | |
2087 | ||
4990aadc | 2088 | #define WRITE_BUFFER_SIZE 8192 |
bf0f910d | 2089 | static unsigned char write_buffer[WRITE_BUFFER_SIZE]; |
4990aadc LT |
2090 | static unsigned long write_buffer_len; |
2091 | ||
aab61359 | 2092 | static int ce_write_flush(git_hash_ctx *context, int fd) |
6015c28b JH |
2093 | { |
2094 | unsigned int buffered = write_buffer_len; | |
2095 | if (buffered) { | |
aab61359 | 2096 | the_hash_algo->update_fn(context, write_buffer, buffered); |
06f46f23 | 2097 | if (write_in_full(fd, write_buffer, buffered) < 0) |
6015c28b JH |
2098 | return -1; |
2099 | write_buffer_len = 0; | |
2100 | } | |
2101 | return 0; | |
2102 | } | |
2103 | ||
aab61359 | 2104 | static int ce_write(git_hash_ctx *context, int fd, void *data, unsigned int len) |
4990aadc LT |
2105 | { |
2106 | while (len) { | |
2107 | unsigned int buffered = write_buffer_len; | |
2108 | unsigned int partial = WRITE_BUFFER_SIZE - buffered; | |
2109 | if (partial > len) | |
2110 | partial = len; | |
2111 | memcpy(write_buffer + buffered, data, partial); | |
2112 | buffered += partial; | |
2113 | if (buffered == WRITE_BUFFER_SIZE) { | |
6015c28b JH |
2114 | write_buffer_len = buffered; |
2115 | if (ce_write_flush(context, fd)) | |
4990aadc LT |
2116 | return -1; |
2117 | buffered = 0; | |
2118 | } | |
2119 | write_buffer_len = buffered; | |
2120 | len -= partial; | |
1d7f171c | 2121 | data = (char *) data + partial; |
a6080a0a JH |
2122 | } |
2123 | return 0; | |
4990aadc LT |
2124 | } |
2125 | ||
aab61359 | 2126 | static int write_index_ext_header(git_hash_ctx *context, int fd, |
ac58c7b1 | 2127 | unsigned int ext, unsigned int sz) |
bad68ec9 JH |
2128 | { |
2129 | ext = htonl(ext); | |
2130 | sz = htonl(sz); | |
968a1d65 DR |
2131 | return ((ce_write(context, fd, &ext, 4) < 0) || |
2132 | (ce_write(context, fd, &sz, 4) < 0)) ? -1 : 0; | |
bad68ec9 JH |
2133 | } |
2134 | ||
aab61359 | 2135 | static int ce_flush(git_hash_ctx *context, int fd, unsigned char *hash) |
4990aadc LT |
2136 | { |
2137 | unsigned int left = write_buffer_len; | |
ca9be054 | 2138 | |
4990aadc LT |
2139 | if (left) { |
2140 | write_buffer_len = 0; | |
aab61359 | 2141 | the_hash_algo->update_fn(context, write_buffer, left); |
4990aadc | 2142 | } |
ca9be054 | 2143 | |
aab61359 | 2144 | /* Flush first if not enough space for hash signature */ |
2145 | if (left + the_hash_algo->rawsz > WRITE_BUFFER_SIZE) { | |
06f46f23 | 2146 | if (write_in_full(fd, write_buffer, left) < 0) |
2c865d9a QH |
2147 | return -1; |
2148 | left = 0; | |
2149 | } | |
2150 | ||
aab61359 | 2151 | /* Append the hash signature at the end */ |
2152 | the_hash_algo->final_fn(write_buffer + left, context); | |
2153 | hashcpy(hash, write_buffer + left); | |
2154 | left += the_hash_algo->rawsz; | |
06f46f23 | 2155 | return (write_in_full(fd, write_buffer, left) < 0) ? -1 : 0; |
4990aadc LT |
2156 | } |
2157 | ||
407c8eb0 JH |
2158 | static void ce_smudge_racily_clean_entry(struct cache_entry *ce) |
2159 | { | |
2160 | /* | |
2161 | * The only thing we care about in this function is to smudge the | |
2162 | * falsely clean entry due to touch-update-touch race, so we leave | |
2163 | * everything else as they are. We are called for entries whose | |
c21d39d7 | 2164 | * ce_stat_data.sd_mtime match the index file mtime. |
c70115b4 JH |
2165 | * |
2166 | * Note that this actually does not do much for gitlinks, for | |
2167 | * which ce_match_stat_basic() always goes to the actual | |
2168 | * contents. The caller checks with is_racy_timestamp() which | |
2169 | * always says "no" for gitlinks, so we are not called for them ;-) | |
407c8eb0 JH |
2170 | */ |
2171 | struct stat st; | |
2172 | ||
2173 | if (lstat(ce->name, &st) < 0) | |
2174 | return; | |
2175 | if (ce_match_stat_basic(ce, &st)) | |
2176 | return; | |
2177 | if (ce_modified_check_fs(ce, &st)) { | |
4b3511b0 JH |
2178 | /* This is "racily clean"; smudge it. Note that this |
2179 | * is a tricky code. At first glance, it may appear | |
2180 | * that it can break with this sequence: | |
2181 | * | |
2182 | * $ echo xyzzy >frotz | |
2183 | * $ git-update-index --add frotz | |
2184 | * $ : >frotz | |
2185 | * $ sleep 3 | |
2186 | * $ echo filfre >nitfol | |
2187 | * $ git-update-index --add nitfol | |
2188 | * | |
b7e58b17 | 2189 | * but it does not. When the second update-index runs, |
4b3511b0 JH |
2190 | * it notices that the entry "frotz" has the same timestamp |
2191 | * as index, and if we were to smudge it by resetting its | |
2192 | * size to zero here, then the object name recorded | |
2193 | * in index is the 6-byte file but the cached stat information | |
2194 | * becomes zero --- which would then match what we would | |
a6080a0a | 2195 | * obtain from the filesystem next time we stat("frotz"). |
4b3511b0 JH |
2196 | * |
2197 | * However, the second update-index, before calling | |
2198 | * this function, notices that the cached size is 6 | |
2199 | * bytes and what is on the filesystem is an empty | |
2200 | * file, and never calls us, so the cached size information | |
2201 | * for "frotz" stays 6 which does not match the filesystem. | |
2202 | */ | |
c21d39d7 | 2203 | ce->ce_stat_data.sd_size = 0; |
407c8eb0 JH |
2204 | } |
2205 | } | |
2206 | ||
f136f7bf | 2207 | /* Copy miscellaneous fields but not the name */ |
ce012deb | 2208 | static void copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk, |
f136f7bf | 2209 | struct cache_entry *ce) |
7a51ed66 | 2210 | { |
b60e188c TG |
2211 | short flags; |
2212 | ||
c21d39d7 MH |
2213 | ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec); |
2214 | ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec); | |
2215 | ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec); | |
2216 | ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec); | |
2217 | ondisk->dev = htonl(ce->ce_stat_data.sd_dev); | |
2218 | ondisk->ino = htonl(ce->ce_stat_data.sd_ino); | |
7a51ed66 | 2219 | ondisk->mode = htonl(ce->ce_mode); |
c21d39d7 MH |
2220 | ondisk->uid = htonl(ce->ce_stat_data.sd_uid); |
2221 | ondisk->gid = htonl(ce->ce_stat_data.sd_gid); | |
2222 | ondisk->size = htonl(ce->ce_stat_data.sd_size); | |
99d1a986 | 2223 | hashcpy(ondisk->sha1, ce->oid.hash); |
b60e188c | 2224 | |
ce51bf09 | 2225 | flags = ce->ce_flags & ~CE_NAMEMASK; |
b60e188c TG |
2226 | flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce)); |
2227 | ondisk->flags = htons(flags); | |
06aaaa0b NTND |
2228 | if (ce->ce_flags & CE_EXTENDED) { |
2229 | struct ondisk_cache_entry_extended *ondisk2; | |
2230 | ondisk2 = (struct ondisk_cache_entry_extended *)ondisk; | |
2231 | ondisk2->flags2 = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16); | |
f136f7bf JH |
2232 | } |
2233 | } | |
2234 | ||
aab61359 | 2235 | static int ce_write_entry(git_hash_ctx *c, int fd, struct cache_entry *ce, |
ce012deb | 2236 | struct strbuf *previous_name, struct ondisk_cache_entry *ondisk) |
f136f7bf | 2237 | { |
9d227781 | 2238 | int size; |
f136f7bf | 2239 | int result; |
00a4b035 RJ |
2240 | unsigned int saved_namelen; |
2241 | int stripped_name = 0; | |
ce012deb | 2242 | static unsigned char padding[8] = { 0x00 }; |
f136f7bf | 2243 | |
b3c96fb1 NTND |
2244 | if (ce->ce_flags & CE_STRIP_NAME) { |
2245 | saved_namelen = ce_namelen(ce); | |
2246 | ce->ce_namelen = 0; | |
00a4b035 | 2247 | stripped_name = 1; |
b3c96fb1 NTND |
2248 | } |
2249 | ||
ce012deb KW |
2250 | if (ce->ce_flags & CE_EXTENDED) |
2251 | size = offsetof(struct ondisk_cache_entry_extended, name); | |
2252 | else | |
2253 | size = offsetof(struct ondisk_cache_entry, name); | |
2254 | ||
9d227781 | 2255 | if (!previous_name) { |
ce012deb KW |
2256 | int len = ce_namelen(ce); |
2257 | copy_cache_entry_to_ondisk(ondisk, ce); | |
2258 | result = ce_write(c, fd, ondisk, size); | |
2259 | if (!result) | |
2260 | result = ce_write(c, fd, ce->name, len); | |
2261 | if (!result) | |
2262 | result = ce_write(c, fd, padding, align_padding_size(size, len)); | |
9d227781 JH |
2263 | } else { |
2264 | int common, to_remove, prefix_size; | |
2265 | unsigned char to_remove_vi[16]; | |
2266 | for (common = 0; | |
2267 | (ce->name[common] && | |
2268 | common < previous_name->len && | |
2269 | ce->name[common] == previous_name->buf[common]); | |
2270 | common++) | |
2271 | ; /* still matching */ | |
2272 | to_remove = previous_name->len - common; | |
2273 | prefix_size = encode_varint(to_remove, to_remove_vi); | |
2274 | ||
ce012deb KW |
2275 | copy_cache_entry_to_ondisk(ondisk, ce); |
2276 | result = ce_write(c, fd, ondisk, size); | |
2277 | if (!result) | |
2278 | result = ce_write(c, fd, to_remove_vi, prefix_size); | |
2279 | if (!result) | |
0b90b881 TG |
2280 | result = ce_write(c, fd, ce->name + common, ce_namelen(ce) - common); |
2281 | if (!result) | |
2282 | result = ce_write(c, fd, padding, 1); | |
9d227781 JH |
2283 | |
2284 | strbuf_splice(previous_name, common, to_remove, | |
2285 | ce->name + common, ce_namelen(ce) - common); | |
06aaaa0b | 2286 | } |
00a4b035 | 2287 | if (stripped_name) { |
b3c96fb1 NTND |
2288 | ce->ce_namelen = saved_namelen; |
2289 | ce->ce_flags &= ~CE_STRIP_NAME; | |
2290 | } | |
7a51ed66 | 2291 | |
59efba64 | 2292 | return result; |
7a51ed66 LT |
2293 | } |
2294 | ||
426ddeea YM |
2295 | /* |
2296 | * This function verifies if index_state has the correct sha1 of the | |
2297 | * index file. Don't die if we have any other failure, just return 0. | |
2298 | */ | |
2299 | static int verify_index_from(const struct index_state *istate, const char *path) | |
2300 | { | |
2301 | int fd; | |
2302 | ssize_t n; | |
2303 | struct stat st; | |
aab61359 | 2304 | unsigned char hash[GIT_MAX_RAWSZ]; |
426ddeea YM |
2305 | |
2306 | if (!istate->initialized) | |
2307 | return 0; | |
2308 | ||
2309 | fd = open(path, O_RDONLY); | |
2310 | if (fd < 0) | |
2311 | return 0; | |
2312 | ||
2313 | if (fstat(fd, &st)) | |
2314 | goto out; | |
2315 | ||
aab61359 | 2316 | if (st.st_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
426ddeea YM |
2317 | goto out; |
2318 | ||
aab61359 | 2319 | n = pread_in_full(fd, hash, the_hash_algo->rawsz, st.st_size - the_hash_algo->rawsz); |
2320 | if (n != the_hash_algo->rawsz) | |
426ddeea YM |
2321 | goto out; |
2322 | ||
75691ea3 | 2323 | if (hashcmp(istate->oid.hash, hash)) |
426ddeea YM |
2324 | goto out; |
2325 | ||
2326 | close(fd); | |
2327 | return 1; | |
2328 | ||
2329 | out: | |
2330 | close(fd); | |
2331 | return 0; | |
2332 | } | |
2333 | ||
2334 | static int verify_index(const struct index_state *istate) | |
2335 | { | |
2336 | return verify_index_from(istate, get_index_file()); | |
2337 | } | |
2338 | ||
483fbe2b JH |
2339 | static int has_racy_timestamp(struct index_state *istate) |
2340 | { | |
2341 | int entries = istate->cache_nr; | |
2342 | int i; | |
2343 | ||
2344 | for (i = 0; i < entries; i++) { | |
2345 | struct cache_entry *ce = istate->cache[i]; | |
2346 | if (is_racy_timestamp(istate, ce)) | |
2347 | return 1; | |
2348 | } | |
2349 | return 0; | |
2350 | } | |
2351 | ||
ccdc4ec3 JH |
2352 | void update_index_if_able(struct index_state *istate, struct lock_file *lockfile) |
2353 | { | |
483fbe2b | 2354 | if ((istate->cache_changed || has_racy_timestamp(istate)) && |
df60cf57 MÅ |
2355 | verify_index(istate)) |
2356 | write_locked_index(istate, lockfile, COMMIT_LOCK); | |
b74c90fb MÅ |
2357 | else |
2358 | rollback_lock_file(lockfile); | |
ccdc4ec3 JH |
2359 | } |
2360 | ||
812d6b00 MÅ |
2361 | /* |
2362 | * On success, `tempfile` is closed. If it is the temporary file | |
2363 | * of a `struct lock_file`, we will therefore effectively perform | |
2364 | * a 'close_lock_file_gently()`. Since that is an implementation | |
2365 | * detail of lockfiles, callers of `do_write_index()` should not | |
2366 | * rely on it. | |
2367 | */ | |
9f41c7a6 | 2368 | static int do_write_index(struct index_state *istate, struct tempfile *tempfile, |
c18b80a0 | 2369 | int strip_extensions) |
197ee8c9 | 2370 | { |
ca54d9ba | 2371 | uint64_t start = getnanotime(); |
9f41c7a6 | 2372 | int newfd = tempfile->fd; |
aab61359 | 2373 | git_hash_ctx c; |
197ee8c9 | 2374 | struct cache_header hdr; |
b50386c7 | 2375 | int i, err = 0, removed, extended, hdr_version; |
4aab5b46 JH |
2376 | struct cache_entry **cache = istate->cache; |
2377 | int entries = istate->cache_nr; | |
e1afca4f | 2378 | struct stat st; |
ce012deb | 2379 | struct ondisk_cache_entry_extended ondisk; |
9d227781 | 2380 | struct strbuf previous_name_buf = STRBUF_INIT, *previous_name; |
4bddd983 | 2381 | int drop_cache_tree = istate->drop_cache_tree; |
025a0709 | 2382 | |
06aaaa0b | 2383 | for (i = removed = extended = 0; i < entries; i++) { |
7a51ed66 | 2384 | if (cache[i]->ce_flags & CE_REMOVE) |
025a0709 | 2385 | removed++; |
197ee8c9 | 2386 | |
06aaaa0b NTND |
2387 | /* reduce extended entries if possible */ |
2388 | cache[i]->ce_flags &= ~CE_EXTENDED; | |
2389 | if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) { | |
2390 | extended++; | |
2391 | cache[i]->ce_flags |= CE_EXTENDED; | |
2392 | } | |
2393 | } | |
2394 | ||
d6e3c181 | 2395 | if (!istate->version) { |
136347d7 | 2396 | istate->version = get_index_format_default(); |
4c2db938 | 2397 | if (git_env_bool("GIT_TEST_SPLIT_INDEX", 0)) |
d6e3c181 NTND |
2398 | init_split_index(istate); |
2399 | } | |
9d227781 JH |
2400 | |
2401 | /* demote version 3 to version 2 when the latter suffices */ | |
2402 | if (istate->version == 3 || istate->version == 2) | |
2403 | istate->version = extended ? 3 : 2; | |
2404 | ||
2405 | hdr_version = istate->version; | |
2406 | ||
ccc4feb5 | 2407 | hdr.hdr_signature = htonl(CACHE_SIGNATURE); |
9d227781 | 2408 | hdr.hdr_version = htonl(hdr_version); |
025a0709 | 2409 | hdr.hdr_entries = htonl(entries - removed); |
197ee8c9 | 2410 | |
aab61359 | 2411 | the_hash_algo->init_fn(&c); |
ca9be054 | 2412 | if (ce_write(&c, newfd, &hdr, sizeof(hdr)) < 0) |
197ee8c9 LT |
2413 | return -1; |
2414 | ||
9d227781 | 2415 | previous_name = (hdr_version == 4) ? &previous_name_buf : NULL; |
ce012deb | 2416 | |
197ee8c9 LT |
2417 | for (i = 0; i < entries; i++) { |
2418 | struct cache_entry *ce = cache[i]; | |
7a51ed66 | 2419 | if (ce->ce_flags & CE_REMOVE) |
aa16021e | 2420 | continue; |
e06c43c7 | 2421 | if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce)) |
407c8eb0 | 2422 | ce_smudge_racily_clean_entry(ce); |
99d1a986 | 2423 | if (is_null_oid(&ce->oid)) { |
83bd7437 JK |
2424 | static const char msg[] = "cache entry has null sha1: %s"; |
2425 | static int allow = -1; | |
2426 | ||
2427 | if (allow < 0) | |
2428 | allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0); | |
2429 | if (allow) | |
2430 | warning(msg, ce->name); | |
2431 | else | |
b50386c7 | 2432 | err = error(msg, ce->name); |
a96d3cc3 JK |
2433 | |
2434 | drop_cache_tree = 1; | |
83bd7437 | 2435 | } |
ce012deb | 2436 | if (ce_write_entry(&c, newfd, ce, previous_name, (struct ondisk_cache_entry *)&ondisk) < 0) |
b50386c7 KW |
2437 | err = -1; |
2438 | ||
2439 | if (err) | |
2440 | break; | |
197ee8c9 | 2441 | } |
9d227781 | 2442 | strbuf_release(&previous_name_buf); |
1af1c2b6 | 2443 | |
b50386c7 KW |
2444 | if (err) |
2445 | return err; | |
2446 | ||
bad68ec9 | 2447 | /* Write extension data here */ |
c18b80a0 | 2448 | if (!strip_extensions && istate->split_index) { |
5fc2fc8f NTND |
2449 | struct strbuf sb = STRBUF_INIT; |
2450 | ||
2451 | err = write_link_extension(&sb, istate) < 0 || | |
2452 | write_index_ext_header(&c, newfd, CACHE_EXT_LINK, | |
2453 | sb.len) < 0 || | |
2454 | ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2455 | strbuf_release(&sb); | |
2456 | if (err) | |
2457 | return -1; | |
2458 | } | |
a96d3cc3 | 2459 | if (!strip_extensions && !drop_cache_tree && istate->cache_tree) { |
f285a2d7 | 2460 | struct strbuf sb = STRBUF_INIT; |
1dffb8fa | 2461 | |
1dffb8fa PH |
2462 | cache_tree_write(&sb, istate->cache_tree); |
2463 | err = write_index_ext_header(&c, newfd, CACHE_EXT_TREE, sb.len) < 0 | |
2464 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2465 | strbuf_release(&sb); | |
2466 | if (err) | |
bad68ec9 | 2467 | return -1; |
bad68ec9 | 2468 | } |
c18b80a0 | 2469 | if (!strip_extensions && istate->resolve_undo) { |
cfc5789a JH |
2470 | struct strbuf sb = STRBUF_INIT; |
2471 | ||
2472 | resolve_undo_write(&sb, istate->resolve_undo); | |
2473 | err = write_index_ext_header(&c, newfd, CACHE_EXT_RESOLVE_UNDO, | |
2474 | sb.len) < 0 | |
2475 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2476 | strbuf_release(&sb); | |
2477 | if (err) | |
2478 | return -1; | |
2479 | } | |
83c094ad NTND |
2480 | if (!strip_extensions && istate->untracked) { |
2481 | struct strbuf sb = STRBUF_INIT; | |
2482 | ||
2483 | write_untracked_extension(&sb, istate->untracked); | |
2484 | err = write_index_ext_header(&c, newfd, CACHE_EXT_UNTRACKED, | |
2485 | sb.len) < 0 || | |
2486 | ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2487 | strbuf_release(&sb); | |
2488 | if (err) | |
2489 | return -1; | |
2490 | } | |
883e248b BP |
2491 | if (!strip_extensions && istate->fsmonitor_last_update) { |
2492 | struct strbuf sb = STRBUF_INIT; | |
2493 | ||
2494 | write_fsmonitor_extension(&sb, istate); | |
2495 | err = write_index_ext_header(&c, newfd, CACHE_EXT_FSMONITOR, sb.len) < 0 | |
2496 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2497 | strbuf_release(&sb); | |
2498 | if (err) | |
2499 | return -1; | |
2500 | } | |
e1afca4f | 2501 | |
75691ea3 | 2502 | if (ce_flush(&c, newfd, istate->oid.hash)) |
9f41c7a6 | 2503 | return -1; |
49bd0fc2 JK |
2504 | if (close_tempfile_gently(tempfile)) { |
2505 | error(_("could not close '%s'"), tempfile->filename.buf); | |
49bd0fc2 JK |
2506 | return -1; |
2507 | } | |
9f41c7a6 | 2508 | if (stat(tempfile->filename.buf, &st)) |
e1afca4f | 2509 | return -1; |
5bcf109c KB |
2510 | istate->timestamp.sec = (unsigned int)st.st_mtime; |
2511 | istate->timestamp.nsec = ST_MTIME_NSEC(st); | |
ca54d9ba | 2512 | trace_performance_since(start, "write index, changed mask = %x", istate->cache_changed); |
e1afca4f | 2513 | return 0; |
197ee8c9 | 2514 | } |
e46bbcf6 | 2515 | |
626f35c8 NTND |
2516 | void set_alternate_index_output(const char *name) |
2517 | { | |
2518 | alternate_index_output = name; | |
2519 | } | |
2520 | ||
2521 | static int commit_locked_index(struct lock_file *lk) | |
2522 | { | |
751baced MH |
2523 | if (alternate_index_output) |
2524 | return commit_lock_file_to(lk, alternate_index_output); | |
2525 | else | |
626f35c8 | 2526 | return commit_lock_file(lk); |
626f35c8 NTND |
2527 | } |
2528 | ||
03b86647 NTND |
2529 | static int do_write_locked_index(struct index_state *istate, struct lock_file *lock, |
2530 | unsigned flags) | |
2531 | { | |
076aa2cb | 2532 | int ret = do_write_index(istate, lock->tempfile, 0); |
03b86647 NTND |
2533 | if (ret) |
2534 | return ret; | |
03b86647 NTND |
2535 | if (flags & COMMIT_LOCK) |
2536 | return commit_locked_index(lock); | |
812d6b00 | 2537 | return close_lock_file_gently(lock); |
03b86647 NTND |
2538 | } |
2539 | ||
5fc2fc8f NTND |
2540 | static int write_split_index(struct index_state *istate, |
2541 | struct lock_file *lock, | |
2542 | unsigned flags) | |
2543 | { | |
2544 | int ret; | |
2545 | prepare_to_write_split_index(istate); | |
2546 | ret = do_write_locked_index(istate, lock, flags); | |
2547 | finish_writing_split_index(istate); | |
2548 | return ret; | |
2549 | } | |
2550 | ||
b9683722 CC |
2551 | static const char *shared_index_expire = "2.weeks.ago"; |
2552 | ||
2553 | static unsigned long get_shared_index_expire_date(void) | |
2554 | { | |
2555 | static unsigned long shared_index_expire_date; | |
2556 | static int shared_index_expire_date_prepared; | |
2557 | ||
2558 | if (!shared_index_expire_date_prepared) { | |
2559 | git_config_get_expiry("splitindex.sharedindexexpire", | |
2560 | &shared_index_expire); | |
2561 | shared_index_expire_date = approxidate(shared_index_expire); | |
2562 | shared_index_expire_date_prepared = 1; | |
2563 | } | |
2564 | ||
2565 | return shared_index_expire_date; | |
2566 | } | |
2567 | ||
2568 | static int should_delete_shared_index(const char *shared_index_path) | |
2569 | { | |
2570 | struct stat st; | |
2571 | unsigned long expiration; | |
2572 | ||
2573 | /* Check timestamp */ | |
2574 | expiration = get_shared_index_expire_date(); | |
2575 | if (!expiration) | |
2576 | return 0; | |
2577 | if (stat(shared_index_path, &st)) | |
78bde923 | 2578 | return error_errno(_("could not stat '%s'"), shared_index_path); |
b9683722 CC |
2579 | if (st.st_mtime > expiration) |
2580 | return 0; | |
2581 | ||
2582 | return 1; | |
2583 | } | |
2584 | ||
2585 | static int clean_shared_index_files(const char *current_hex) | |
2586 | { | |
2587 | struct dirent *de; | |
2588 | DIR *dir = opendir(get_git_dir()); | |
2589 | ||
2590 | if (!dir) | |
2591 | return error_errno(_("unable to open git dir: %s"), get_git_dir()); | |
2592 | ||
2593 | while ((de = readdir(dir)) != NULL) { | |
2594 | const char *sha1_hex; | |
2595 | const char *shared_index_path; | |
2596 | if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex)) | |
2597 | continue; | |
2598 | if (!strcmp(sha1_hex, current_hex)) | |
2599 | continue; | |
2600 | shared_index_path = git_path("%s", de->d_name); | |
2601 | if (should_delete_shared_index(shared_index_path) > 0 && | |
2602 | unlink(shared_index_path)) | |
2603 | warning_errno(_("unable to unlink: %s"), shared_index_path); | |
2604 | } | |
2605 | closedir(dir); | |
2606 | ||
2607 | return 0; | |
2608 | } | |
2609 | ||
a0a96756 | 2610 | static int write_shared_index(struct index_state *istate, |
59f9d2dd | 2611 | struct tempfile **temp) |
c18b80a0 NTND |
2612 | { |
2613 | struct split_index *si = istate->split_index; | |
076aa2cb | 2614 | int ret; |
c18b80a0 | 2615 | |
c18b80a0 | 2616 | move_cache_to_base_index(istate); |
7db2d08c | 2617 | ret = do_write_index(si->base, *temp, 1); |
59f9d2dd | 2618 | if (ret) |
c18b80a0 | 2619 | return ret; |
7db2d08c | 2620 | ret = adjust_shared_perm(get_tempfile_path(*temp)); |
df801f3f | 2621 | if (ret) { |
7db2d08c | 2622 | error("cannot fix permission bits on %s", get_tempfile_path(*temp)); |
df801f3f CC |
2623 | return ret; |
2624 | } | |
7db2d08c | 2625 | ret = rename_tempfile(temp, |
75691ea3 | 2626 | git_path("sharedindex.%s", oid_to_hex(&si->base->oid))); |
b9683722 | 2627 | if (!ret) { |
75691ea3 | 2628 | oidcpy(&si->base_oid, &si->base->oid); |
2629 | clean_shared_index_files(oid_to_hex(&si->base->oid)); | |
b9683722 CC |
2630 | } |
2631 | ||
c18b80a0 NTND |
2632 | return ret; |
2633 | } | |
2634 | ||
e6a1dd77 CC |
2635 | static const int default_max_percent_split_change = 20; |
2636 | ||
2637 | static int too_many_not_shared_entries(struct index_state *istate) | |
2638 | { | |
2639 | int i, not_shared = 0; | |
2640 | int max_split = git_config_get_max_percent_split_change(); | |
2641 | ||
2642 | switch (max_split) { | |
2643 | case -1: | |
2644 | /* not or badly configured: use the default value */ | |
2645 | max_split = default_max_percent_split_change; | |
2646 | break; | |
2647 | case 0: | |
2648 | return 1; /* 0% means always write a new shared index */ | |
2649 | case 100: | |
2650 | return 0; /* 100% means never write a new shared index */ | |
2651 | default: | |
2652 | break; /* just use the configured value */ | |
2653 | } | |
2654 | ||
2655 | /* Count not shared entries */ | |
2656 | for (i = 0; i < istate->cache_nr; i++) { | |
2657 | struct cache_entry *ce = istate->cache[i]; | |
2658 | if (!ce->index) | |
2659 | not_shared++; | |
2660 | } | |
2661 | ||
2662 | return (int64_t)istate->cache_nr * max_split < (int64_t)not_shared * 100; | |
2663 | } | |
2664 | ||
03b86647 NTND |
2665 | int write_locked_index(struct index_state *istate, struct lock_file *lock, |
2666 | unsigned flags) | |
2667 | { | |
0d59ffb4 | 2668 | int new_shared_index, ret; |
5fc2fc8f NTND |
2669 | struct split_index *si = istate->split_index; |
2670 | ||
61000814 MÅ |
2671 | if ((flags & SKIP_IF_UNCHANGED) && !istate->cache_changed) { |
2672 | if (flags & COMMIT_LOCK) | |
2673 | rollback_lock_file(lock); | |
2674 | return 0; | |
2675 | } | |
2676 | ||
3bd28eb2 AV |
2677 | if (istate->fsmonitor_last_update) |
2678 | fill_fsmonitor_bitmap(istate); | |
2679 | ||
5165dd59 NTND |
2680 | if (!si || alternate_index_output || |
2681 | (istate->cache_changed & ~EXTMASK)) { | |
5fc2fc8f | 2682 | if (si) |
2182abd9 | 2683 | oidclr(&si->base_oid); |
df60cf57 MÅ |
2684 | ret = do_write_locked_index(istate, lock, flags); |
2685 | goto out; | |
5fc2fc8f NTND |
2686 | } |
2687 | ||
4c2db938 | 2688 | if (git_env_bool("GIT_TEST_SPLIT_INDEX", 0)) { |
2182abd9 | 2689 | int v = si->base_oid.hash[0]; |
d6e3c181 NTND |
2690 | if ((v & 15) < 6) |
2691 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
2692 | } | |
e6a1dd77 CC |
2693 | if (too_many_not_shared_entries(istate)) |
2694 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
0d59ffb4 CC |
2695 | |
2696 | new_shared_index = istate->cache_changed & SPLIT_INDEX_ORDERED; | |
2697 | ||
2698 | if (new_shared_index) { | |
59f9d2dd NTND |
2699 | struct tempfile *temp; |
2700 | int saved_errno; | |
2701 | ||
2702 | temp = mks_tempfile(git_path("sharedindex_XXXXXX")); | |
2703 | if (!temp) { | |
2182abd9 | 2704 | oidclr(&si->base_oid); |
59f9d2dd | 2705 | ret = do_write_locked_index(istate, lock, flags); |
ef5b3a6c NTND |
2706 | goto out; |
2707 | } | |
2708 | ret = write_shared_index(istate, &temp); | |
59f9d2dd NTND |
2709 | |
2710 | saved_errno = errno; | |
2711 | if (is_tempfile_active(temp)) | |
2712 | delete_tempfile(&temp); | |
2713 | errno = saved_errno; | |
2714 | ||
c18b80a0 | 2715 | if (ret) |
df60cf57 | 2716 | goto out; |
c18b80a0 NTND |
2717 | } |
2718 | ||
0d59ffb4 CC |
2719 | ret = write_split_index(istate, lock, flags); |
2720 | ||
2721 | /* Freshen the shared index only if the split-index was written */ | |
a125a223 TG |
2722 | if (!ret && !new_shared_index) { |
2723 | const char *shared_index = git_path("sharedindex.%s", | |
2182abd9 | 2724 | oid_to_hex(&si->base_oid)); |
a125a223 TG |
2725 | freshen_shared_index(shared_index, 1); |
2726 | } | |
0d59ffb4 | 2727 | |
df60cf57 MÅ |
2728 | out: |
2729 | if (flags & COMMIT_LOCK) | |
2730 | rollback_lock_file(lock); | |
0d59ffb4 | 2731 | return ret; |
03b86647 NTND |
2732 | } |
2733 | ||
e46bbcf6 MV |
2734 | /* |
2735 | * Read the index file that is potentially unmerged into given | |
63e8dc5b | 2736 | * index_state, dropping any unmerged entries. Returns true if |
e46bbcf6 MV |
2737 | * the index is unmerged. Callers who want to refuse to work |
2738 | * from an unmerged state can call this and check its return value, | |
2739 | * instead of calling read_cache(). | |
2740 | */ | |
2741 | int read_index_unmerged(struct index_state *istate) | |
2742 | { | |
2743 | int i; | |
d1a43f2a | 2744 | int unmerged = 0; |
e46bbcf6 MV |
2745 | |
2746 | read_index(istate); | |
e46bbcf6 MV |
2747 | for (i = 0; i < istate->cache_nr; i++) { |
2748 | struct cache_entry *ce = istate->cache[i]; | |
d1a43f2a | 2749 | struct cache_entry *new_ce; |
a849735b | 2750 | int len; |
d1a43f2a JH |
2751 | |
2752 | if (!ce_stage(ce)) | |
e46bbcf6 | 2753 | continue; |
d1a43f2a | 2754 | unmerged = 1; |
68c4f6a5 | 2755 | len = ce_namelen(ce); |
a849735b | 2756 | new_ce = make_empty_cache_entry(istate, len); |
d1a43f2a | 2757 | memcpy(new_ce->name, ce->name, len); |
b60e188c TG |
2758 | new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED; |
2759 | new_ce->ce_namelen = len; | |
d1a43f2a JH |
2760 | new_ce->ce_mode = ce->ce_mode; |
2761 | if (add_index_entry(istate, new_ce, 0)) | |
2762 | return error("%s: cannot drop to stage #0", | |
5699d17e | 2763 | new_ce->name); |
e46bbcf6 | 2764 | } |
d1a43f2a | 2765 | return unmerged; |
e46bbcf6 | 2766 | } |
041aee31 | 2767 | |
98fa4738 JK |
2768 | /* |
2769 | * Returns 1 if the path is an "other" path with respect to | |
2770 | * the index; that is, the path is not mentioned in the index at all, | |
2771 | * either as a file, a directory with some files in the index, | |
2772 | * or as an unmerged entry. | |
2773 | * | |
2774 | * We helpfully remove a trailing "/" from directories so that | |
2775 | * the output of read_directory can be used as-is. | |
2776 | */ | |
2777 | int index_name_is_other(const struct index_state *istate, const char *name, | |
2778 | int namelen) | |
2779 | { | |
2780 | int pos; | |
2781 | if (namelen && name[namelen - 1] == '/') | |
2782 | namelen--; | |
2783 | pos = index_name_pos(istate, name, namelen); | |
2784 | if (0 <= pos) | |
2785 | return 0; /* exact match */ | |
2786 | pos = -pos - 1; | |
2787 | if (pos < istate->cache_nr) { | |
2788 | struct cache_entry *ce = istate->cache[pos]; | |
2789 | if (ce_namelen(ce) == namelen && | |
2790 | !memcmp(ce->name, name, namelen)) | |
2791 | return 0; /* Yup, this one exists unmerged */ | |
2792 | } | |
2793 | return 1; | |
2794 | } | |
29fb37b2 | 2795 | |
87542508 BW |
2796 | void *read_blob_data_from_index(const struct index_state *istate, |
2797 | const char *path, unsigned long *size) | |
29fb37b2 LF |
2798 | { |
2799 | int pos, len; | |
2800 | unsigned long sz; | |
2801 | enum object_type type; | |
2802 | void *data; | |
2803 | ||
2804 | len = strlen(path); | |
2805 | pos = index_name_pos(istate, path, len); | |
2806 | if (pos < 0) { | |
2807 | /* | |
2808 | * We might be in the middle of a merge, in which | |
2809 | * case we would read stage #2 (ours). | |
2810 | */ | |
2811 | int i; | |
2812 | for (i = -pos - 1; | |
2813 | (pos < 0 && i < istate->cache_nr && | |
2814 | !strcmp(istate->cache[i]->name, path)); | |
2815 | i++) | |
2816 | if (ce_stage(istate->cache[i]) == 2) | |
2817 | pos = i; | |
2818 | } | |
2819 | if (pos < 0) | |
2820 | return NULL; | |
b4f5aca4 | 2821 | data = read_object_file(&istate->cache[pos]->oid, &type, &sz); |
29fb37b2 LF |
2822 | if (!data || type != OBJ_BLOB) { |
2823 | free(data); | |
2824 | return NULL; | |
2825 | } | |
ff366825 LF |
2826 | if (size) |
2827 | *size = sz; | |
29fb37b2 LF |
2828 | return data; |
2829 | } | |
38612532 MH |
2830 | |
2831 | void stat_validity_clear(struct stat_validity *sv) | |
2832 | { | |
6a83d902 | 2833 | FREE_AND_NULL(sv->sd); |
38612532 MH |
2834 | } |
2835 | ||
2836 | int stat_validity_check(struct stat_validity *sv, const char *path) | |
2837 | { | |
2838 | struct stat st; | |
2839 | ||
2840 | if (stat(path, &st) < 0) | |
2841 | return sv->sd == NULL; | |
2842 | if (!sv->sd) | |
2843 | return 0; | |
2844 | return S_ISREG(st.st_mode) && !match_stat_data(sv->sd, &st); | |
2845 | } | |
2846 | ||
2847 | void stat_validity_update(struct stat_validity *sv, int fd) | |
2848 | { | |
2849 | struct stat st; | |
2850 | ||
2851 | if (fstat(fd, &st) < 0 || !S_ISREG(st.st_mode)) | |
2852 | stat_validity_clear(sv); | |
2853 | else { | |
2854 | if (!sv->sd) | |
2855 | sv->sd = xcalloc(1, sizeof(struct stat_data)); | |
2856 | fill_stat_data(sv->sd, &st); | |
2857 | } | |
2858 | } | |
edf3b905 DT |
2859 | |
2860 | void move_index_extensions(struct index_state *dst, struct index_state *src) | |
2861 | { | |
2862 | dst->untracked = src->untracked; | |
2863 | src->untracked = NULL; | |
2864 | } | |
a849735b | 2865 | |
8e72d675 JM |
2866 | struct cache_entry *dup_cache_entry(const struct cache_entry *ce, |
2867 | struct index_state *istate) | |
2868 | { | |
2869 | unsigned int size = ce_size(ce); | |
2870 | int mem_pool_allocated; | |
2871 | struct cache_entry *new_entry = make_empty_cache_entry(istate, ce_namelen(ce)); | |
2872 | mem_pool_allocated = new_entry->mem_pool_allocated; | |
2873 | ||
2874 | memcpy(new_entry, ce, size); | |
2875 | new_entry->mem_pool_allocated = mem_pool_allocated; | |
2876 | return new_entry; | |
2877 | } | |
2878 | ||
a849735b JM |
2879 | void discard_cache_entry(struct cache_entry *ce) |
2880 | { | |
8e72d675 JM |
2881 | if (ce && ce->mem_pool_allocated) |
2882 | return; | |
2883 | ||
a849735b JM |
2884 | free(ce); |
2885 | } |