]>
Commit | Line | Data |
---|---|---|
8bc9a0c7 LT |
1 | /* |
2 | * GIT - The information manager from hell | |
3 | * | |
4 | * Copyright (C) Linus Torvalds, 2005 | |
5 | */ | |
e83c5163 | 6 | #include "cache.h" |
b2141fc1 | 7 | #include "config.h" |
cffbfad5 EN |
8 | #include "diff.h" |
9 | #include "diffcore.h" | |
f6ecc62d | 10 | #include "tempfile.h" |
697cc8ef | 11 | #include "lockfile.h" |
bad68ec9 | 12 | #include "cache-tree.h" |
f35a6d3b | 13 | #include "refs.h" |
d616813d | 14 | #include "dir.h" |
cbd53a21 | 15 | #include "object-store.h" |
041aee31 JH |
16 | #include "tree.h" |
17 | #include "commit.h" | |
39425819 | 18 | #include "blob.h" |
cfc5789a | 19 | #include "resolve-undo.h" |
1956ecd0 | 20 | #include "run-command.h" |
6c9cd161 JH |
21 | #include "strbuf.h" |
22 | #include "varint.h" | |
5fc2fc8f | 23 | #include "split-index.h" |
a42643aa | 24 | #include "utf8.h" |
883e248b | 25 | #include "fsmonitor.h" |
abb4bb83 | 26 | #include "thread-utils.h" |
ae9af122 | 27 | #include "progress.h" |
6e773527 | 28 | #include "sparse-index.h" |
410334ed | 29 | #include "csum-file.h" |
b2896d27 | 30 | #include "promisor-remote.h" |
dbb1c613 | 31 | #include "hook.h" |
bad68ec9 | 32 | |
b60e188c TG |
33 | /* Mask for the name length in ce_flags in the on-disk index */ |
34 | ||
35 | #define CE_NAMEMASK (0x0fff) | |
36 | ||
bad68ec9 JH |
37 | /* Index extensions. |
38 | * | |
39 | * The first letter should be 'A'..'Z' for extensions that are not | |
40 | * necessary for a correct operation (i.e. optimization data). | |
41 | * When new extensions are added that _needs_ to be understood in | |
42 | * order to correctly interpret the index file, pick character that | |
43 | * is outside the range, to cause the reader to abort. | |
44 | */ | |
45 | ||
46 | #define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) ) | |
47 | #define CACHE_EXT_TREE 0x54524545 /* "TREE" */ | |
b659b49b | 48 | #define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */ |
5fc2fc8f | 49 | #define CACHE_EXT_LINK 0x6c696e6b /* "link" */ |
83c094ad | 50 | #define CACHE_EXT_UNTRACKED 0x554E5452 /* "UNTR" */ |
883e248b | 51 | #define CACHE_EXT_FSMONITOR 0x46534D4E /* "FSMN" */ |
3b1d9e04 | 52 | #define CACHE_EXT_ENDOFINDEXENTRIES 0x454F4945 /* "EOIE" */ |
3255089a | 53 | #define CACHE_EXT_INDEXENTRYOFFSETTABLE 0x49454F54 /* "IEOT" */ |
cd42415f | 54 | #define CACHE_EXT_SPARSE_DIRECTORIES 0x73646972 /* "sdir" */ |
5fc2fc8f NTND |
55 | |
56 | /* changes that can be kept in $GIT_DIR/index (basically all extensions) */ | |
e0cf0d7d | 57 | #define EXTMASK (RESOLVE_UNDO_CHANGED | CACHE_TREE_CHANGED | \ |
c18b80a0 | 58 | CE_ENTRY_ADDED | CE_ENTRY_REMOVED | CE_ENTRY_CHANGED | \ |
883e248b | 59 | SPLIT_INDEX_ORDERED | UNTRACKED_CHANGED | FSMONITOR_CHANGED) |
e83c5163 | 60 | |
8e72d675 JM |
61 | |
62 | /* | |
63 | * This is an estimate of the pathname length in the index. We use | |
64 | * this for V4 index files to guess the un-deltafied size of the index | |
65 | * in memory because of pathname deltafication. This is not required | |
66 | * for V2/V3 index formats because their pathnames are not compressed. | |
67 | * If the initial amount of memory set aside is not sufficient, the | |
68 | * mem pool will allocate extra memory. | |
69 | */ | |
70 | #define CACHE_ENTRY_PATH_LENGTH 80 | |
71 | ||
20ec2d03 VD |
72 | enum index_search_mode { |
73 | NO_EXPAND_SPARSE = 0, | |
74 | EXPAND_SPARSE = 1 | |
75 | }; | |
76 | ||
8e72d675 JM |
77 | static inline struct cache_entry *mem_pool__ce_alloc(struct mem_pool *mem_pool, size_t len) |
78 | { | |
79 | struct cache_entry *ce; | |
80 | ce = mem_pool_alloc(mem_pool, cache_entry_size(len)); | |
81 | ce->mem_pool_allocated = 1; | |
82 | return ce; | |
83 | } | |
84 | ||
85 | static inline struct cache_entry *mem_pool__ce_calloc(struct mem_pool *mem_pool, size_t len) | |
86 | { | |
87 | struct cache_entry * ce; | |
88 | ce = mem_pool_calloc(mem_pool, 1, cache_entry_size(len)); | |
89 | ce->mem_pool_allocated = 1; | |
90 | return ce; | |
91 | } | |
92 | ||
93 | static struct mem_pool *find_mem_pool(struct index_state *istate) | |
94 | { | |
95 | struct mem_pool **pool_ptr; | |
96 | ||
97 | if (istate->split_index && istate->split_index->base) | |
98 | pool_ptr = &istate->split_index->base->ce_mem_pool; | |
99 | else | |
100 | pool_ptr = &istate->ce_mem_pool; | |
101 | ||
44c7e1a7 EN |
102 | if (!*pool_ptr) { |
103 | *pool_ptr = xmalloc(sizeof(**pool_ptr)); | |
104 | mem_pool_init(*pool_ptr, 0); | |
105 | } | |
8e72d675 JM |
106 | |
107 | return *pool_ptr; | |
108 | } | |
109 | ||
626f35c8 | 110 | static const char *alternate_index_output; |
8fd2cb40 | 111 | |
9cb76b8c JH |
112 | static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
113 | { | |
4300f844 | 114 | if (S_ISSPARSEDIR(ce->ce_mode)) |
9fadb373 | 115 | istate->sparse_index = INDEX_COLLAPSED; |
4300f844 | 116 | |
9cb76b8c | 117 | istate->cache[nr] = ce; |
96872bc2 | 118 | add_name_hash(istate, ce); |
9cb76b8c JH |
119 | } |
120 | ||
cf558704 LT |
121 | static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
122 | { | |
123 | struct cache_entry *old = istate->cache[nr]; | |
124 | ||
078a58e8 | 125 | replace_index_entry_in_base(istate, old, ce); |
2092678c | 126 | remove_name_hash(istate, old); |
a849735b | 127 | discard_cache_entry(old); |
0e267b7a | 128 | ce->ce_flags &= ~CE_HASHED; |
a22c6371 | 129 | set_index_entry(istate, nr, ce); |
078a58e8 | 130 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 131 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 132 | istate->cache_changed |= CE_ENTRY_CHANGED; |
cf558704 LT |
133 | } |
134 | ||
81dc2307 PB |
135 | void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name) |
136 | { | |
b7f9130a | 137 | struct cache_entry *old_entry = istate->cache[nr], *new_entry, *refreshed; |
81dc2307 PB |
138 | int namelen = strlen(new_name); |
139 | ||
a849735b | 140 | new_entry = make_empty_cache_entry(istate, namelen); |
285c2e25 BW |
141 | copy_cache_entry(new_entry, old_entry); |
142 | new_entry->ce_flags &= ~CE_HASHED; | |
143 | new_entry->ce_namelen = namelen; | |
144 | new_entry->index = 0; | |
145 | memcpy(new_entry->name, new_name, namelen + 1); | |
81dc2307 | 146 | |
285c2e25 BW |
147 | cache_tree_invalidate_path(istate, old_entry->name); |
148 | untracked_cache_remove_from_index(istate, old_entry->name); | |
81dc2307 | 149 | remove_index_entry_at(istate, nr); |
b7f9130a VD |
150 | |
151 | /* | |
152 | * Refresh the new index entry. Using 'refresh_cache_entry' ensures | |
153 | * we only update stat info if the entry is otherwise up-to-date (i.e., | |
154 | * the contents/mode haven't changed). This ensures that we reflect the | |
155 | * 'ctime' of the rename in the index without (incorrectly) updating | |
156 | * the cached stat info to reflect unstaged changes on disk. | |
157 | */ | |
158 | refreshed = refresh_cache_entry(istate, new_entry, CE_MATCH_REFRESH); | |
159 | if (refreshed && refreshed != new_entry) { | |
160 | add_index_entry(istate, refreshed, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); | |
161 | discard_cache_entry(new_entry); | |
162 | } else | |
163 | add_index_entry(istate, new_entry, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); | |
81dc2307 PB |
164 | } |
165 | ||
c21d39d7 MH |
166 | void fill_stat_data(struct stat_data *sd, struct stat *st) |
167 | { | |
168 | sd->sd_ctime.sec = (unsigned int)st->st_ctime; | |
169 | sd->sd_mtime.sec = (unsigned int)st->st_mtime; | |
170 | sd->sd_ctime.nsec = ST_CTIME_NSEC(*st); | |
171 | sd->sd_mtime.nsec = ST_MTIME_NSEC(*st); | |
172 | sd->sd_dev = st->st_dev; | |
173 | sd->sd_ino = st->st_ino; | |
174 | sd->sd_uid = st->st_uid; | |
175 | sd->sd_gid = st->st_gid; | |
176 | sd->sd_size = st->st_size; | |
177 | } | |
178 | ||
179 | int match_stat_data(const struct stat_data *sd, struct stat *st) | |
180 | { | |
181 | int changed = 0; | |
182 | ||
183 | if (sd->sd_mtime.sec != (unsigned int)st->st_mtime) | |
184 | changed |= MTIME_CHANGED; | |
185 | if (trust_ctime && check_stat && | |
186 | sd->sd_ctime.sec != (unsigned int)st->st_ctime) | |
187 | changed |= CTIME_CHANGED; | |
188 | ||
189 | #ifdef USE_NSEC | |
190 | if (check_stat && sd->sd_mtime.nsec != ST_MTIME_NSEC(*st)) | |
191 | changed |= MTIME_CHANGED; | |
192 | if (trust_ctime && check_stat && | |
193 | sd->sd_ctime.nsec != ST_CTIME_NSEC(*st)) | |
194 | changed |= CTIME_CHANGED; | |
195 | #endif | |
196 | ||
197 | if (check_stat) { | |
198 | if (sd->sd_uid != (unsigned int) st->st_uid || | |
199 | sd->sd_gid != (unsigned int) st->st_gid) | |
200 | changed |= OWNER_CHANGED; | |
201 | if (sd->sd_ino != (unsigned int) st->st_ino) | |
202 | changed |= INODE_CHANGED; | |
203 | } | |
204 | ||
205 | #ifdef USE_STDEV | |
206 | /* | |
207 | * st_dev breaks on network filesystems where different | |
208 | * clients will have different views of what "device" | |
209 | * the filesystem is on | |
210 | */ | |
211 | if (check_stat && sd->sd_dev != (unsigned int) st->st_dev) | |
212 | changed |= INODE_CHANGED; | |
213 | #endif | |
214 | ||
215 | if (sd->sd_size != (unsigned int) st->st_size) | |
216 | changed |= DATA_CHANGED; | |
217 | ||
218 | return changed; | |
219 | } | |
220 | ||
415e96c8 JH |
221 | /* |
222 | * This only updates the "non-critical" parts of the directory | |
223 | * cache, ie the parts that aren't tracked by GIT, and only used | |
224 | * to validate the cache. | |
225 | */ | |
d4c0a3ac | 226 | void fill_stat_cache_info(struct index_state *istate, struct cache_entry *ce, struct stat *st) |
415e96c8 | 227 | { |
c21d39d7 | 228 | fill_stat_data(&ce->ce_stat_data, st); |
5f73076c JH |
229 | |
230 | if (assume_unchanged) | |
7a51ed66 | 231 | ce->ce_flags |= CE_VALID; |
eadb5831 | 232 | |
883e248b | 233 | if (S_ISREG(st->st_mode)) { |
eadb5831 | 234 | ce_mark_uptodate(ce); |
b5a81697 | 235 | mark_fsmonitor_valid(istate, ce); |
883e248b | 236 | } |
415e96c8 JH |
237 | } |
238 | ||
58bf2a4c NTND |
239 | static int ce_compare_data(struct index_state *istate, |
240 | const struct cache_entry *ce, | |
241 | struct stat *st) | |
29e4d363 JH |
242 | { |
243 | int match = -1; | |
1b8ac5ea | 244 | int fd = git_open_cloexec(ce->name, O_RDONLY); |
29e4d363 JH |
245 | |
246 | if (fd >= 0) { | |
bebfecb9 | 247 | struct object_id oid; |
58bf2a4c | 248 | if (!index_fd(istate, &oid, fd, st, OBJ_BLOB, ce->name, 0)) |
6a29d7b7 | 249 | match = !oideq(&oid, &ce->oid); |
7f8508e8 | 250 | /* index_fd() closed the file descriptor already */ |
29e4d363 JH |
251 | } |
252 | return match; | |
253 | } | |
254 | ||
21a6b9fa | 255 | static int ce_compare_link(const struct cache_entry *ce, size_t expected_size) |
29e4d363 JH |
256 | { |
257 | int match = -1; | |
29e4d363 JH |
258 | void *buffer; |
259 | unsigned long size; | |
21666f1a | 260 | enum object_type type; |
a60272b3 | 261 | struct strbuf sb = STRBUF_INIT; |
29e4d363 | 262 | |
a60272b3 | 263 | if (strbuf_readlink(&sb, ce->name, expected_size)) |
29e4d363 | 264 | return -1; |
a60272b3 | 265 | |
b4f5aca4 | 266 | buffer = read_object_file(&ce->oid, &type, &size); |
a60272b3 LT |
267 | if (buffer) { |
268 | if (size == sb.len) | |
269 | match = memcmp(buffer, sb.buf, size); | |
270 | free(buffer); | |
29e4d363 | 271 | } |
a60272b3 | 272 | strbuf_release(&sb); |
29e4d363 JH |
273 | return match; |
274 | } | |
275 | ||
21a6b9fa | 276 | static int ce_compare_gitlink(const struct cache_entry *ce) |
f35a6d3b | 277 | { |
1053fe82 | 278 | struct object_id oid; |
f35a6d3b LT |
279 | |
280 | /* | |
281 | * We don't actually require that the .git directory | |
302b9282 | 282 | * under GITLINK directory be a valid git directory. It |
f35a6d3b LT |
283 | * might even be missing (in case nobody populated that |
284 | * sub-project). | |
285 | * | |
286 | * If so, we consider it always to match. | |
287 | */ | |
a98e6101 | 288 | if (resolve_gitlink_ref(ce->name, "HEAD", &oid) < 0) |
f35a6d3b | 289 | return 0; |
6a29d7b7 | 290 | return !oideq(&oid, &ce->oid); |
f35a6d3b LT |
291 | } |
292 | ||
58bf2a4c NTND |
293 | static int ce_modified_check_fs(struct index_state *istate, |
294 | const struct cache_entry *ce, | |
295 | struct stat *st) | |
29e4d363 JH |
296 | { |
297 | switch (st->st_mode & S_IFMT) { | |
298 | case S_IFREG: | |
58bf2a4c | 299 | if (ce_compare_data(istate, ce, st)) |
29e4d363 JH |
300 | return DATA_CHANGED; |
301 | break; | |
302 | case S_IFLNK: | |
dc49cd76 | 303 | if (ce_compare_link(ce, xsize_t(st->st_size))) |
29e4d363 JH |
304 | return DATA_CHANGED; |
305 | break; | |
a8ee75bc | 306 | case S_IFDIR: |
7a51ed66 | 307 | if (S_ISGITLINK(ce->ce_mode)) |
c70115b4 | 308 | return ce_compare_gitlink(ce) ? DATA_CHANGED : 0; |
1cf01a34 | 309 | /* else fallthrough */ |
29e4d363 JH |
310 | default: |
311 | return TYPE_CHANGED; | |
312 | } | |
313 | return 0; | |
314 | } | |
315 | ||
21a6b9fa | 316 | static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st) |
734aab75 LT |
317 | { |
318 | unsigned int changed = 0; | |
319 | ||
7a51ed66 LT |
320 | if (ce->ce_flags & CE_REMOVE) |
321 | return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED; | |
322 | ||
323 | switch (ce->ce_mode & S_IFMT) { | |
8ae0a8c5 KS |
324 | case S_IFREG: |
325 | changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0; | |
3e09cdfd JH |
326 | /* We consider only the owner x bit to be relevant for |
327 | * "mode changes" | |
328 | */ | |
329 | if (trust_executable_bit && | |
7a51ed66 | 330 | (0100 & (ce->ce_mode ^ st->st_mode))) |
ffbe1add | 331 | changed |= MODE_CHANGED; |
8ae0a8c5 KS |
332 | break; |
333 | case S_IFLNK: | |
78a8d641 JS |
334 | if (!S_ISLNK(st->st_mode) && |
335 | (has_symlinks || !S_ISREG(st->st_mode))) | |
336 | changed |= TYPE_CHANGED; | |
8ae0a8c5 | 337 | break; |
302b9282 | 338 | case S_IFGITLINK: |
c70115b4 | 339 | /* We ignore most of the st_xxx fields for gitlinks */ |
f35a6d3b LT |
340 | if (!S_ISDIR(st->st_mode)) |
341 | changed |= TYPE_CHANGED; | |
342 | else if (ce_compare_gitlink(ce)) | |
343 | changed |= DATA_CHANGED; | |
a8ee75bc | 344 | return changed; |
8ae0a8c5 | 345 | default: |
391408e5 | 346 | BUG("unsupported ce_mode: %o", ce->ce_mode); |
8ae0a8c5 | 347 | } |
2cb45e95 | 348 | |
c21d39d7 | 349 | changed |= match_stat_data(&ce->ce_stat_data, st); |
b0391890 | 350 | |
f49c2c22 | 351 | /* Racily smudged entry? */ |
c21d39d7 | 352 | if (!ce->ce_stat_data.sd_size) { |
99d1a986 | 353 | if (!is_empty_blob_sha1(ce->oid.hash)) |
f49c2c22 LT |
354 | changed |= DATA_CHANGED; |
355 | } | |
356 | ||
407c8eb0 JH |
357 | return changed; |
358 | } | |
359 | ||
2bb4cda1 NTND |
360 | static int is_racy_stat(const struct index_state *istate, |
361 | const struct stat_data *sd) | |
6d91da6d | 362 | { |
2bb4cda1 | 363 | return (istate->timestamp.sec && |
fba2f38a KB |
364 | #ifdef USE_NSEC |
365 | /* nanosecond timestamped files can also be racy! */ | |
2bb4cda1 NTND |
366 | (istate->timestamp.sec < sd->sd_mtime.sec || |
367 | (istate->timestamp.sec == sd->sd_mtime.sec && | |
368 | istate->timestamp.nsec <= sd->sd_mtime.nsec)) | |
fba2f38a | 369 | #else |
2bb4cda1 | 370 | istate->timestamp.sec <= sd->sd_mtime.sec |
fba2f38a | 371 | #endif |
2bb4cda1 NTND |
372 | ); |
373 | } | |
374 | ||
5581a019 | 375 | int is_racy_timestamp(const struct index_state *istate, |
2bb4cda1 NTND |
376 | const struct cache_entry *ce) |
377 | { | |
378 | return (!S_ISGITLINK(ce->ce_mode) && | |
379 | is_racy_stat(istate, &ce->ce_stat_data)); | |
6d91da6d JH |
380 | } |
381 | ||
ed4efab1 NTND |
382 | int match_stat_data_racy(const struct index_state *istate, |
383 | const struct stat_data *sd, struct stat *st) | |
384 | { | |
385 | if (is_racy_stat(istate, sd)) | |
386 | return MTIME_CHANGED; | |
387 | return match_stat_data(sd, st); | |
6d91da6d JH |
388 | } |
389 | ||
883e248b | 390 | int ie_match_stat(struct index_state *istate, |
21a6b9fa | 391 | const struct cache_entry *ce, struct stat *st, |
4bd5b7da | 392 | unsigned int options) |
407c8eb0 | 393 | { |
5f73076c | 394 | unsigned int changed; |
4bd5b7da | 395 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 396 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
4bd5b7da | 397 | int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY; |
883e248b | 398 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
5f73076c | 399 | |
883e248b BP |
400 | if (!ignore_fsmonitor) |
401 | refresh_fsmonitor(istate); | |
5f73076c JH |
402 | /* |
403 | * If it's marked as always valid in the index, it's | |
404 | * valid whatever the checked-out copy says. | |
56cac48c NTND |
405 | * |
406 | * skip-worktree has the same effect with higher precedence | |
5f73076c | 407 | */ |
56cac48c NTND |
408 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) |
409 | return 0; | |
7a51ed66 | 410 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) |
5f73076c | 411 | return 0; |
883e248b BP |
412 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) |
413 | return 0; | |
5f73076c | 414 | |
331fcb59 JH |
415 | /* |
416 | * Intent-to-add entries have not been added, so the index entry | |
417 | * by definition never matches what is in the work tree until it | |
418 | * actually gets added. | |
419 | */ | |
895ff3b2 | 420 | if (ce_intent_to_add(ce)) |
331fcb59 JH |
421 | return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED; |
422 | ||
5f73076c | 423 | changed = ce_match_stat_basic(ce, st); |
407c8eb0 | 424 | |
29e4d363 JH |
425 | /* |
426 | * Within 1 second of this sequence: | |
427 | * echo xyzzy >file && git-update-index --add file | |
428 | * running this command: | |
429 | * echo frotz >file | |
430 | * would give a falsely clean cache entry. The mtime and | |
431 | * length match the cache, and other stat fields do not change. | |
432 | * | |
433 | * We could detect this at update-index time (the cache entry | |
434 | * being registered/updated records the same time as "now") | |
435 | * and delay the return from git-update-index, but that would | |
436 | * effectively mean we can make at most one commit per second, | |
437 | * which is not acceptable. Instead, we check cache entries | |
438 | * whose mtime are the same as the index file timestamp more | |
5f73076c | 439 | * carefully than others. |
29e4d363 | 440 | */ |
6d91da6d | 441 | if (!changed && is_racy_timestamp(istate, ce)) { |
42f77406 JH |
442 | if (assume_racy_is_modified) |
443 | changed |= DATA_CHANGED; | |
444 | else | |
58bf2a4c | 445 | changed |= ce_modified_check_fs(istate, ce, st); |
42f77406 | 446 | } |
b0391890 | 447 | |
29e4d363 | 448 | return changed; |
b0391890 JH |
449 | } |
450 | ||
883e248b | 451 | int ie_modified(struct index_state *istate, |
21a6b9fa RS |
452 | const struct cache_entry *ce, |
453 | struct stat *st, unsigned int options) | |
b0391890 | 454 | { |
29e4d363 | 455 | int changed, changed_fs; |
4bd5b7da JH |
456 | |
457 | changed = ie_match_stat(istate, ce, st, options); | |
b0391890 JH |
458 | if (!changed) |
459 | return 0; | |
b0391890 JH |
460 | /* |
461 | * If the mode or type has changed, there's no point in trying | |
462 | * to refresh the entry - it's not going to match | |
463 | */ | |
464 | if (changed & (MODE_CHANGED | TYPE_CHANGED)) | |
465 | return changed; | |
466 | ||
c70115b4 JH |
467 | /* |
468 | * Immediately after read-tree or update-index --cacheinfo, | |
469 | * the length field is zero, as we have never even read the | |
470 | * lstat(2) information once, and we cannot trust DATA_CHANGED | |
471 | * returned by ie_match_stat() which in turn was returned by | |
472 | * ce_match_stat_basic() to signal that the filesize of the | |
473 | * blob changed. We have to actually go to the filesystem to | |
474 | * see if the contents match, and if so, should answer "unchanged". | |
475 | * | |
476 | * The logic does not apply to gitlinks, as ce_match_stat_basic() | |
477 | * already has checked the actual HEAD from the filesystem in the | |
478 | * subproject. If ie_match_stat() already said it is different, | |
479 | * then we know it is. | |
b0391890 | 480 | */ |
c70115b4 | 481 | if ((changed & DATA_CHANGED) && |
c21d39d7 | 482 | (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0)) |
b0391890 JH |
483 | return changed; |
484 | ||
58bf2a4c | 485 | changed_fs = ce_modified_check_fs(istate, ce, st); |
29e4d363 JH |
486 | if (changed_fs) |
487 | return changed | changed_fs; | |
b0391890 JH |
488 | return 0; |
489 | } | |
490 | ||
1b4a38d7 RS |
491 | int base_name_compare(const char *name1, size_t len1, int mode1, |
492 | const char *name2, size_t len2, int mode2) | |
958ba6c9 LT |
493 | { |
494 | unsigned char c1, c2; | |
1b4a38d7 | 495 | size_t len = len1 < len2 ? len1 : len2; |
958ba6c9 LT |
496 | int cmp; |
497 | ||
498 | cmp = memcmp(name1, name2, len); | |
499 | if (cmp) | |
500 | return cmp; | |
501 | c1 = name1[len]; | |
502 | c2 = name2[len]; | |
1833a925 | 503 | if (!c1 && S_ISDIR(mode1)) |
958ba6c9 | 504 | c1 = '/'; |
1833a925 | 505 | if (!c2 && S_ISDIR(mode2)) |
958ba6c9 LT |
506 | c2 = '/'; |
507 | return (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0; | |
508 | } | |
509 | ||
0ab9e1e8 LT |
510 | /* |
511 | * df_name_compare() is identical to base_name_compare(), except it | |
512 | * compares conflicting directory/file entries as equal. Note that | |
513 | * while a directory name compares as equal to a regular file, they | |
514 | * then individually compare _differently_ to a filename that has | |
515 | * a dot after the basename (because '\0' < '.' < '/'). | |
516 | * | |
517 | * This is used by routines that want to traverse the git namespace | |
518 | * but then handle conflicting entries together when possible. | |
519 | */ | |
1b4a38d7 RS |
520 | int df_name_compare(const char *name1, size_t len1, int mode1, |
521 | const char *name2, size_t len2, int mode2) | |
0ab9e1e8 | 522 | { |
0ab9e1e8 | 523 | unsigned char c1, c2; |
1b4a38d7 RS |
524 | size_t len = len1 < len2 ? len1 : len2; |
525 | int cmp; | |
0ab9e1e8 LT |
526 | |
527 | cmp = memcmp(name1, name2, len); | |
528 | if (cmp) | |
529 | return cmp; | |
530 | /* Directories and files compare equal (same length, same name) */ | |
531 | if (len1 == len2) | |
532 | return 0; | |
533 | c1 = name1[len]; | |
534 | if (!c1 && S_ISDIR(mode1)) | |
535 | c1 = '/'; | |
536 | c2 = name2[len]; | |
537 | if (!c2 && S_ISDIR(mode2)) | |
538 | c2 = '/'; | |
539 | if (c1 == '/' && !c2) | |
540 | return 0; | |
541 | if (c2 == '/' && !c1) | |
542 | return 0; | |
543 | return c1 - c2; | |
544 | } | |
545 | ||
ccdd4a0f | 546 | int name_compare(const char *name1, size_t len1, const char *name2, size_t len2) |
eb38c22f | 547 | { |
ccdd4a0f JM |
548 | size_t min_len = (len1 < len2) ? len1 : len2; |
549 | int cmp = memcmp(name1, name2, min_len); | |
eb38c22f LT |
550 | if (cmp) |
551 | return cmp; | |
552 | if (len1 < len2) | |
553 | return -1; | |
554 | if (len1 > len2) | |
555 | return 1; | |
ccdd4a0f JM |
556 | return 0; |
557 | } | |
558 | ||
559 | int cache_name_stage_compare(const char *name1, int len1, int stage1, const char *name2, int len2, int stage2) | |
560 | { | |
561 | int cmp; | |
562 | ||
563 | cmp = name_compare(name1, len1, name2, len2); | |
564 | if (cmp) | |
565 | return cmp; | |
5f73076c | 566 | |
b60e188c | 567 | if (stage1 < stage2) |
95fd5bf8 | 568 | return -1; |
b60e188c | 569 | if (stage1 > stage2) |
95fd5bf8 | 570 | return 1; |
eb38c22f LT |
571 | return 0; |
572 | } | |
573 | ||
20ec2d03 VD |
574 | static int index_name_stage_pos(struct index_state *istate, |
575 | const char *name, int namelen, | |
576 | int stage, | |
577 | enum index_search_mode search_mode) | |
eb38c22f LT |
578 | { |
579 | int first, last; | |
580 | ||
581 | first = 0; | |
4aab5b46 | 582 | last = istate->cache_nr; |
eb38c22f | 583 | while (last > first) { |
568a05c5 | 584 | int next = first + ((last - first) >> 1); |
4aab5b46 | 585 | struct cache_entry *ce = istate->cache[next]; |
b60e188c | 586 | int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce)); |
eb38c22f | 587 | if (!cmp) |
76e7f4ec | 588 | return next; |
eb38c22f LT |
589 | if (cmp < 0) { |
590 | last = next; | |
591 | continue; | |
592 | } | |
593 | first = next+1; | |
594 | } | |
95e0321c | 595 | |
20ec2d03 | 596 | if (search_mode == EXPAND_SPARSE && istate->sparse_index && |
95e0321c DS |
597 | first > 0) { |
598 | /* Note: first <= istate->cache_nr */ | |
599 | struct cache_entry *ce = istate->cache[first - 1]; | |
600 | ||
601 | /* | |
602 | * If we are in a sparse-index _and_ the entry before the | |
603 | * insertion position is a sparse-directory entry that is | |
604 | * an ancestor of 'name', then we need to expand the index | |
605 | * and search again. This will only trigger once, because | |
606 | * thereafter the index is fully expanded. | |
607 | */ | |
608 | if (S_ISSPARSEDIR(ce->ce_mode) && | |
609 | ce_namelen(ce) < namelen && | |
610 | !strncmp(name, ce->name, ce_namelen(ce))) { | |
611 | ensure_full_index(istate); | |
20ec2d03 | 612 | return index_name_stage_pos(istate, name, namelen, stage, search_mode); |
95e0321c DS |
613 | } |
614 | } | |
615 | ||
76e7f4ec | 616 | return -first-1; |
eb38c22f LT |
617 | } |
618 | ||
847a9e5d | 619 | int index_name_pos(struct index_state *istate, const char *name, int namelen) |
b60e188c | 620 | { |
20ec2d03 VD |
621 | return index_name_stage_pos(istate, name, namelen, 0, EXPAND_SPARSE); |
622 | } | |
623 | ||
9553aa0f VD |
624 | int index_name_pos_sparse(struct index_state *istate, const char *name, int namelen) |
625 | { | |
626 | return index_name_stage_pos(istate, name, namelen, 0, NO_EXPAND_SPARSE); | |
627 | } | |
628 | ||
20ec2d03 VD |
629 | int index_entry_exists(struct index_state *istate, const char *name, int namelen) |
630 | { | |
631 | return index_name_stage_pos(istate, name, namelen, 0, NO_EXPAND_SPARSE) >= 0; | |
b60e188c TG |
632 | } |
633 | ||
4aab5b46 | 634 | int remove_index_entry_at(struct index_state *istate, int pos) |
7b937ca3 | 635 | { |
cf558704 LT |
636 | struct cache_entry *ce = istate->cache[pos]; |
637 | ||
cfc5789a | 638 | record_resolve_undo(istate, ce); |
2092678c | 639 | remove_name_hash(istate, ce); |
045113a5 | 640 | save_or_free_index_entry(istate, ce); |
e636a7b4 | 641 | istate->cache_changed |= CE_ENTRY_REMOVED; |
4aab5b46 JH |
642 | istate->cache_nr--; |
643 | if (pos >= istate->cache_nr) | |
7b937ca3 | 644 | return 0; |
f331ab9d RS |
645 | MOVE_ARRAY(istate->cache + pos, istate->cache + pos + 1, |
646 | istate->cache_nr - pos); | |
7b937ca3 LT |
647 | return 1; |
648 | } | |
649 | ||
36419c8e | 650 | /* |
98e023de | 651 | * Remove all cache entries marked for removal, that is where |
36419c8e KB |
652 | * CE_REMOVE is set in ce_flags. This is much more effective than |
653 | * calling remove_index_entry_at() for each entry to be removed. | |
654 | */ | |
6fdc2057 | 655 | void remove_marked_cache_entries(struct index_state *istate, int invalidate) |
36419c8e KB |
656 | { |
657 | struct cache_entry **ce_array = istate->cache; | |
658 | unsigned int i, j; | |
659 | ||
660 | for (i = j = 0; i < istate->cache_nr; i++) { | |
5699d17e | 661 | if (ce_array[i]->ce_flags & CE_REMOVE) { |
6fdc2057 TG |
662 | if (invalidate) { |
663 | cache_tree_invalidate_path(istate, | |
664 | ce_array[i]->name); | |
665 | untracked_cache_remove_from_index(istate, | |
666 | ce_array[i]->name); | |
667 | } | |
2092678c | 668 | remove_name_hash(istate, ce_array[i]); |
045113a5 | 669 | save_or_free_index_entry(istate, ce_array[i]); |
5699d17e | 670 | } |
36419c8e KB |
671 | else |
672 | ce_array[j++] = ce_array[i]; | |
673 | } | |
ad837d9e NTND |
674 | if (j == istate->cache_nr) |
675 | return; | |
e636a7b4 | 676 | istate->cache_changed |= CE_ENTRY_REMOVED; |
36419c8e KB |
677 | istate->cache_nr = j; |
678 | } | |
679 | ||
4aab5b46 | 680 | int remove_file_from_index(struct index_state *istate, const char *path) |
197ee8c9 | 681 | { |
4aab5b46 | 682 | int pos = index_name_pos(istate, path, strlen(path)); |
c4e3cca1 JH |
683 | if (pos < 0) |
684 | pos = -pos-1; | |
a5400efe | 685 | cache_tree_invalidate_path(istate, path); |
e931371a | 686 | untracked_cache_remove_from_index(istate, path); |
4aab5b46 JH |
687 | while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path)) |
688 | remove_index_entry_at(istate, pos); | |
197ee8c9 LT |
689 | return 0; |
690 | } | |
691 | ||
20314271 JS |
692 | static int compare_name(struct cache_entry *ce, const char *path, int namelen) |
693 | { | |
694 | return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen); | |
695 | } | |
696 | ||
697 | static int index_name_pos_also_unmerged(struct index_state *istate, | |
698 | const char *path, int namelen) | |
699 | { | |
700 | int pos = index_name_pos(istate, path, namelen); | |
701 | struct cache_entry *ce; | |
702 | ||
703 | if (pos >= 0) | |
704 | return pos; | |
705 | ||
706 | /* maybe unmerged? */ | |
707 | pos = -1 - pos; | |
708 | if (pos >= istate->cache_nr || | |
709 | compare_name((ce = istate->cache[pos]), path, namelen)) | |
710 | return -1; | |
711 | ||
712 | /* order of preference: stage 2, 1, 3 */ | |
713 | if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr && | |
714 | ce_stage((ce = istate->cache[pos + 1])) == 2 && | |
715 | !compare_name(ce, path, namelen)) | |
716 | pos++; | |
717 | return pos; | |
718 | } | |
719 | ||
1102952b LT |
720 | static int different_name(struct cache_entry *ce, struct cache_entry *alias) |
721 | { | |
722 | int len = ce_namelen(ce); | |
723 | return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len); | |
724 | } | |
725 | ||
726 | /* | |
727 | * If we add a filename that aliases in the cache, we will use the | |
728 | * name that we already have - but we don't want to update the same | |
729 | * alias twice, because that implies that there were actually two | |
730 | * different files with aliasing names! | |
731 | * | |
732 | * So we use the CE_ADDED flag to verify that the alias was an old | |
733 | * one before we accept it as | |
734 | */ | |
045113a5 NTND |
735 | static struct cache_entry *create_alias_ce(struct index_state *istate, |
736 | struct cache_entry *ce, | |
737 | struct cache_entry *alias) | |
1102952b LT |
738 | { |
739 | int len; | |
285c2e25 | 740 | struct cache_entry *new_entry; |
1102952b LT |
741 | |
742 | if (alias->ce_flags & CE_ADDED) | |
9d0a9e90 NTND |
743 | die(_("will not add file alias '%s' ('%s' already exists in index)"), |
744 | ce->name, alias->name); | |
1102952b LT |
745 | |
746 | /* Ok, create the new entry using the name of the existing alias */ | |
747 | len = ce_namelen(alias); | |
a849735b | 748 | new_entry = make_empty_cache_entry(istate, len); |
285c2e25 BW |
749 | memcpy(new_entry->name, alias->name, len); |
750 | copy_cache_entry(new_entry, ce); | |
045113a5 | 751 | save_or_free_index_entry(istate, ce); |
285c2e25 | 752 | return new_entry; |
1102952b LT |
753 | } |
754 | ||
b4b313f9 | 755 | void set_object_name_for_intent_to_add_entry(struct cache_entry *ce) |
39425819 | 756 | { |
a09c985e | 757 | struct object_id oid; |
c80d226a | 758 | if (write_object_file("", 0, OBJ_BLOB, &oid)) |
9d0a9e90 | 759 | die(_("cannot create an empty blob in the object database")); |
a09c985e | 760 | oidcpy(&ce->oid, &oid); |
39425819 JH |
761 | } |
762 | ||
610d55af | 763 | int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags) |
11be42a4 | 764 | { |
a849735b | 765 | int namelen, was_same; |
d177cab0 | 766 | mode_t st_mode = st->st_mode; |
9472935d | 767 | struct cache_entry *ce, *alias = NULL; |
56cac48c | 768 | unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY; |
38ed1d89 JH |
769 | int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND); |
770 | int pretend = flags & ADD_CACHE_PRETEND; | |
39425819 JH |
771 | int intent_only = flags & ADD_CACHE_INTENT; |
772 | int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE| | |
773 | (intent_only ? ADD_CACHE_NEW_ONLY : 0)); | |
e578d031 | 774 | unsigned hash_flags = pretend ? 0 : HASH_WRITE_OBJECT; |
f937bc2f | 775 | struct object_id oid; |
9472935d | 776 | |
9e5da3d0 JK |
777 | if (flags & ADD_CACHE_RENORMALIZE) |
778 | hash_flags |= HASH_RENORMALIZE; | |
11be42a4 | 779 | |
d177cab0 | 780 | if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode)) |
9d0a9e90 | 781 | return error(_("%s: can only add regular files, symbolic links or git-directories"), path); |
11be42a4 JS |
782 | |
783 | namelen = strlen(path); | |
d177cab0 | 784 | if (S_ISDIR(st_mode)) { |
f937bc2f KM |
785 | if (resolve_gitlink_ref(path, "HEAD", &oid) < 0) |
786 | return error(_("'%s' does not have a commit checked out"), path); | |
09595258 LT |
787 | while (namelen && path[namelen-1] == '/') |
788 | namelen--; | |
789 | } | |
a849735b | 790 | ce = make_empty_cache_entry(istate, namelen); |
11be42a4 | 791 | memcpy(ce->name, path, namelen); |
b60e188c | 792 | ce->ce_namelen = namelen; |
39425819 | 793 | if (!intent_only) |
d4c0a3ac | 794 | fill_stat_cache_info(istate, ce, st); |
388b2acd JH |
795 | else |
796 | ce->ce_flags |= CE_INTENT_TO_ADD; | |
11be42a4 | 797 | |
610d55af TG |
798 | |
799 | if (trust_executable_bit && has_symlinks) { | |
d177cab0 | 800 | ce->ce_mode = create_ce_mode(st_mode); |
610d55af | 801 | } else { |
78a8d641 JS |
802 | /* If there is an existing entry, pick the mode bits and type |
803 | * from it, otherwise assume unexecutable regular file. | |
11be42a4 | 804 | */ |
185c975f | 805 | struct cache_entry *ent; |
20314271 | 806 | int pos = index_name_pos_also_unmerged(istate, path, namelen); |
185c975f | 807 | |
4aab5b46 | 808 | ent = (0 <= pos) ? istate->cache[pos] : NULL; |
d177cab0 | 809 | ce->ce_mode = ce_mode_from_stat(ent, st_mode); |
11be42a4 JS |
810 | } |
811 | ||
dc1ae704 JJ |
812 | /* When core.ignorecase=true, determine if a directory of the same name but differing |
813 | * case already exists within the Git repository. If it does, ensure the directory | |
814 | * case of the file being added to the repository matches (is folded into) the existing | |
815 | * entry's directory case. | |
816 | */ | |
817 | if (ignore_case) { | |
41284eb0 | 818 | adjust_dirname_case(istate, ce->name); |
dc1ae704 | 819 | } |
e2c2a375 | 820 | if (!(flags & ADD_CACHE_RENORMALIZE)) { |
9472935d TB |
821 | alias = index_file_exists(istate, ce->name, |
822 | ce_namelen(ce), ignore_case); | |
823 | if (alias && | |
824 | !ce_stage(alias) && | |
825 | !ie_match_stat(istate, alias, st, ce_option)) { | |
826 | /* Nothing changed, really */ | |
827 | if (!S_ISGITLINK(alias->ce_mode)) | |
828 | ce_mark_uptodate(alias); | |
829 | alias->ce_flags |= CE_ADDED; | |
dc1ae704 | 830 | |
a849735b | 831 | discard_cache_entry(ce); |
9472935d TB |
832 | return 0; |
833 | } | |
0781b8a9 | 834 | } |
39425819 | 835 | if (!intent_only) { |
1c418243 | 836 | if (index_path(istate, &ce->oid, path, st, hash_flags)) { |
a849735b | 837 | discard_cache_entry(ce); |
9d0a9e90 | 838 | return error(_("unable to index file '%s'"), path); |
2d9426b0 | 839 | } |
39425819 | 840 | } else |
b4b313f9 | 841 | set_object_name_for_intent_to_add_entry(ce); |
39425819 | 842 | |
1102952b | 843 | if (ignore_case && alias && different_name(ce, alias)) |
045113a5 | 844 | ce = create_alias_ce(istate, ce, alias); |
1102952b | 845 | ce->ce_flags |= CE_ADDED; |
38ed1d89 | 846 | |
3bf0dd1f | 847 | /* It was suspected to be racily clean, but it turns out to be Ok */ |
38ed1d89 JH |
848 | was_same = (alias && |
849 | !ce_stage(alias) && | |
4a7e27e9 | 850 | oideq(&alias->oid, &ce->oid) && |
38ed1d89 JH |
851 | ce->ce_mode == alias->ce_mode); |
852 | ||
853 | if (pretend) | |
a849735b | 854 | discard_cache_entry(ce); |
067178ed | 855 | else if (add_index_entry(istate, ce, add_option)) { |
a849735b | 856 | discard_cache_entry(ce); |
9d0a9e90 | 857 | return error(_("unable to add '%s' to index"), path); |
067178ed | 858 | } |
38ed1d89 | 859 | if (verbose && !was_same) |
11be42a4 | 860 | printf("add '%s'\n", path); |
11be42a4 JS |
861 | return 0; |
862 | } | |
863 | ||
610d55af | 864 | int add_file_to_index(struct index_state *istate, const char *path, int flags) |
d177cab0 LT |
865 | { |
866 | struct stat st; | |
867 | if (lstat(path, &st)) | |
9d0a9e90 | 868 | die_errno(_("unable to stat '%s'"), path); |
610d55af | 869 | return add_to_index(istate, path, &st, flags); |
d177cab0 LT |
870 | } |
871 | ||
a849735b JM |
872 | struct cache_entry *make_empty_cache_entry(struct index_state *istate, size_t len) |
873 | { | |
8e72d675 | 874 | return mem_pool__ce_calloc(find_mem_pool(istate), len); |
a849735b JM |
875 | } |
876 | ||
96168827 MT |
877 | struct cache_entry *make_empty_transient_cache_entry(size_t len, |
878 | struct mem_pool *ce_mem_pool) | |
a849735b | 879 | { |
96168827 MT |
880 | if (ce_mem_pool) |
881 | return mem_pool__ce_calloc(ce_mem_pool, len); | |
a849735b JM |
882 | return xcalloc(1, cache_entry_size(len)); |
883 | } | |
884 | ||
2a1ae649 RS |
885 | enum verify_path_result { |
886 | PATH_OK, | |
887 | PATH_INVALID, | |
888 | PATH_DIR_WITH_SEP, | |
889 | }; | |
890 | ||
891 | static enum verify_path_result verify_path_internal(const char *, unsigned); | |
892 | ||
893 | int verify_path(const char *path, unsigned mode) | |
894 | { | |
c8ad9d04 | 895 | return verify_path_internal(path, mode) == PATH_OK; |
2a1ae649 RS |
896 | } |
897 | ||
a849735b JM |
898 | struct cache_entry *make_cache_entry(struct index_state *istate, |
899 | unsigned int mode, | |
825ed4d9 JM |
900 | const struct object_id *oid, |
901 | const char *path, | |
902 | int stage, | |
903 | unsigned int refresh_options) | |
6640f881 | 904 | { |
bc1c2caa | 905 | struct cache_entry *ce, *ret; |
a849735b | 906 | int len; |
6640f881 | 907 | |
c8ad9d04 | 908 | if (verify_path_internal(path, mode) == PATH_INVALID) { |
9d0a9e90 | 909 | error(_("invalid path '%s'"), path); |
6640f881 | 910 | return NULL; |
7e7abea9 | 911 | } |
6640f881 CR |
912 | |
913 | len = strlen(path); | |
a849735b | 914 | ce = make_empty_cache_entry(istate, len); |
6640f881 | 915 | |
825ed4d9 | 916 | oidcpy(&ce->oid, oid); |
6640f881 | 917 | memcpy(ce->name, path, len); |
b60e188c TG |
918 | ce->ce_flags = create_ce_flags(stage); |
919 | ce->ce_namelen = len; | |
6640f881 CR |
920 | ce->ce_mode = create_ce_mode(mode); |
921 | ||
d7b665c3 | 922 | ret = refresh_cache_entry(istate, ce, refresh_options); |
915e44c6 | 923 | if (ret != ce) |
a849735b | 924 | discard_cache_entry(ce); |
915e44c6 | 925 | return ret; |
6640f881 CR |
926 | } |
927 | ||
96168827 MT |
928 | struct cache_entry *make_transient_cache_entry(unsigned int mode, |
929 | const struct object_id *oid, | |
930 | const char *path, | |
931 | int stage, | |
932 | struct mem_pool *ce_mem_pool) | |
a849735b JM |
933 | { |
934 | struct cache_entry *ce; | |
935 | int len; | |
936 | ||
937 | if (!verify_path(path, mode)) { | |
9d0a9e90 | 938 | error(_("invalid path '%s'"), path); |
a849735b JM |
939 | return NULL; |
940 | } | |
941 | ||
942 | len = strlen(path); | |
96168827 | 943 | ce = make_empty_transient_cache_entry(len, ce_mem_pool); |
a849735b JM |
944 | |
945 | oidcpy(&ce->oid, oid); | |
946 | memcpy(ce->name, path, len); | |
947 | ce->ce_flags = create_ce_flags(stage); | |
948 | ce->ce_namelen = len; | |
949 | ce->ce_mode = create_ce_mode(mode); | |
950 | ||
951 | return ce; | |
952 | } | |
953 | ||
d9d70966 TG |
954 | /* |
955 | * Chmod an index entry with either +x or -x. | |
956 | * | |
957 | * Returns -1 if the chmod for the particular cache entry failed (if it's | |
958 | * not a regular file), -2 if an invalid flip argument is passed in, 0 | |
959 | * otherwise. | |
960 | */ | |
961 | int chmod_index_entry(struct index_state *istate, struct cache_entry *ce, | |
962 | char flip) | |
963 | { | |
964 | if (!S_ISREG(ce->ce_mode)) | |
965 | return -1; | |
966 | switch (flip) { | |
967 | case '+': | |
968 | ce->ce_mode |= 0111; | |
969 | break; | |
970 | case '-': | |
971 | ce->ce_mode &= ~0111; | |
972 | break; | |
973 | default: | |
974 | return -2; | |
975 | } | |
976 | cache_tree_invalidate_path(istate, ce->name); | |
977 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
883e248b | 978 | mark_fsmonitor_invalid(istate, ce); |
d9d70966 TG |
979 | istate->cache_changed |= CE_ENTRY_CHANGED; |
980 | ||
981 | return 0; | |
982 | } | |
983 | ||
9c5e6c80 | 984 | int ce_same_name(const struct cache_entry *a, const struct cache_entry *b) |
7b937ca3 LT |
985 | { |
986 | int len = ce_namelen(a); | |
987 | return ce_namelen(b) == len && !memcmp(a->name, b->name, len); | |
988 | } | |
989 | ||
8dcf39c4 LT |
990 | /* |
991 | * We fundamentally don't like some paths: we don't want | |
992 | * dot or dot-dot anywhere, and for obvious reasons don't | |
993 | * want to recurse into ".git" either. | |
994 | * | |
995 | * Also, we don't want double slashes or slashes at the | |
996 | * end that can make pathnames ambiguous. | |
997 | */ | |
10ecfa76 | 998 | static int verify_dotfile(const char *rest, unsigned mode) |
8dcf39c4 LT |
999 | { |
1000 | /* | |
1001 | * The first character was '.', but that | |
1002 | * has already been discarded, we now test | |
1003 | * the rest. | |
1004 | */ | |
e0f530ff | 1005 | |
8dcf39c4 | 1006 | /* "." is not allowed */ |
e0f530ff | 1007 | if (*rest == '\0' || is_dir_sep(*rest)) |
8dcf39c4 LT |
1008 | return 0; |
1009 | ||
e0f530ff | 1010 | switch (*rest) { |
8dcf39c4 | 1011 | /* |
641084b6 JK |
1012 | * ".git" followed by NUL or slash is bad. Note that we match |
1013 | * case-insensitively here, even if ignore_case is not set. | |
1014 | * This outlaws ".GIT" everywhere out of an abundance of caution, | |
1015 | * since there's really no good reason to allow it. | |
10ecfa76 JK |
1016 | * |
1017 | * Once we've seen ".git", we can also find ".gitmodules", etc (also | |
1018 | * case-insensitively). | |
8dcf39c4 LT |
1019 | */ |
1020 | case 'g': | |
cc2fc7c2 JK |
1021 | case 'G': |
1022 | if (rest[1] != 'i' && rest[1] != 'I') | |
8dcf39c4 | 1023 | break; |
cc2fc7c2 | 1024 | if (rest[2] != 't' && rest[2] != 'T') |
8dcf39c4 | 1025 | break; |
e19e5e66 JK |
1026 | if (rest[3] == '\0' || is_dir_sep(rest[3])) |
1027 | return 0; | |
10ecfa76 JK |
1028 | if (S_ISLNK(mode)) { |
1029 | rest += 3; | |
1030 | if (skip_iprefix(rest, "modules", &rest) && | |
1031 | (*rest == '\0' || is_dir_sep(*rest))) | |
1032 | return 0; | |
1033 | } | |
e19e5e66 | 1034 | break; |
8dcf39c4 | 1035 | case '.': |
e0f530ff | 1036 | if (rest[1] == '\0' || is_dir_sep(rest[1])) |
8dcf39c4 LT |
1037 | return 0; |
1038 | } | |
1039 | return 1; | |
1040 | } | |
1041 | ||
2a1ae649 RS |
1042 | static enum verify_path_result verify_path_internal(const char *path, |
1043 | unsigned mode) | |
8dcf39c4 | 1044 | { |
49e268e2 | 1045 | char c = 0; |
8dcf39c4 | 1046 | |
56948cb6 | 1047 | if (has_dos_drive_prefix(path)) |
2a1ae649 | 1048 | return PATH_INVALID; |
56948cb6 | 1049 | |
d2c84dad | 1050 | if (!is_valid_path(path)) |
2a1ae649 | 1051 | return PATH_INVALID; |
d2c84dad | 1052 | |
8dcf39c4 LT |
1053 | goto inside; |
1054 | for (;;) { | |
1055 | if (!c) | |
2a1ae649 | 1056 | return PATH_OK; |
56948cb6 | 1057 | if (is_dir_sep(c)) { |
8dcf39c4 | 1058 | inside: |
10ecfa76 | 1059 | if (protect_hfs) { |
49e268e2 | 1060 | |
10ecfa76 | 1061 | if (is_hfs_dotgit(path)) |
2a1ae649 | 1062 | return PATH_INVALID; |
10ecfa76 JK |
1063 | if (S_ISLNK(mode)) { |
1064 | if (is_hfs_dotgitmodules(path)) | |
2a1ae649 | 1065 | return PATH_INVALID; |
10ecfa76 JK |
1066 | } |
1067 | } | |
1068 | if (protect_ntfs) { | |
bccc37fd | 1069 | #if defined GIT_WINDOWS_NATIVE || defined __CYGWIN__ |
49e268e2 | 1070 | if (c == '\\') |
2a1ae649 | 1071 | return PATH_INVALID; |
49e268e2 | 1072 | #endif |
10ecfa76 | 1073 | if (is_ntfs_dotgit(path)) |
2a1ae649 | 1074 | return PATH_INVALID; |
10ecfa76 JK |
1075 | if (S_ISLNK(mode)) { |
1076 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 1077 | return PATH_INVALID; |
10ecfa76 JK |
1078 | } |
1079 | } | |
1080 | ||
8dcf39c4 | 1081 | c = *path++; |
10ecfa76 | 1082 | if ((c == '.' && !verify_dotfile(path, mode)) || |
6e773527 | 1083 | is_dir_sep(c)) |
2a1ae649 | 1084 | return PATH_INVALID; |
6e773527 DS |
1085 | /* |
1086 | * allow terminating directory separators for | |
1087 | * sparse directory entries. | |
1088 | */ | |
1089 | if (c == '\0') | |
2a1ae649 RS |
1090 | return S_ISDIR(mode) ? PATH_DIR_WITH_SEP : |
1091 | PATH_INVALID; | |
288a74bc JS |
1092 | } else if (c == '\\' && protect_ntfs) { |
1093 | if (is_ntfs_dotgit(path)) | |
2a1ae649 | 1094 | return PATH_INVALID; |
288a74bc JS |
1095 | if (S_ISLNK(mode)) { |
1096 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 1097 | return PATH_INVALID; |
288a74bc | 1098 | } |
8dcf39c4 | 1099 | } |
288a74bc | 1100 | |
8dcf39c4 LT |
1101 | c = *path++; |
1102 | } | |
1103 | } | |
1104 | ||
12676608 LT |
1105 | /* |
1106 | * Do we have another file that has the beginning components being a | |
1107 | * proper superset of the name we're trying to add? | |
0f1e4f04 | 1108 | */ |
4aab5b46 JH |
1109 | static int has_file_name(struct index_state *istate, |
1110 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
0f1e4f04 | 1111 | { |
12676608 LT |
1112 | int retval = 0; |
1113 | int len = ce_namelen(ce); | |
b155725d | 1114 | int stage = ce_stage(ce); |
12676608 | 1115 | const char *name = ce->name; |
0f1e4f04 | 1116 | |
4aab5b46 JH |
1117 | while (pos < istate->cache_nr) { |
1118 | struct cache_entry *p = istate->cache[pos++]; | |
0f1e4f04 | 1119 | |
12676608 | 1120 | if (len >= ce_namelen(p)) |
0f1e4f04 | 1121 | break; |
12676608 LT |
1122 | if (memcmp(name, p->name, len)) |
1123 | break; | |
b155725d JH |
1124 | if (ce_stage(p) != stage) |
1125 | continue; | |
12676608 LT |
1126 | if (p->name[len] != '/') |
1127 | continue; | |
7a51ed66 | 1128 | if (p->ce_flags & CE_REMOVE) |
21cd8d00 | 1129 | continue; |
12676608 LT |
1130 | retval = -1; |
1131 | if (!ok_to_replace) | |
1132 | break; | |
4aab5b46 | 1133 | remove_index_entry_at(istate, --pos); |
0f1e4f04 | 1134 | } |
12676608 LT |
1135 | return retval; |
1136 | } | |
0f1e4f04 | 1137 | |
a6db3fbb JH |
1138 | |
1139 | /* | |
1140 | * Like strcmp(), but also return the offset of the first change. | |
1141 | * If strings are equal, return the length. | |
1142 | */ | |
1143 | int strcmp_offset(const char *s1, const char *s2, size_t *first_change) | |
1144 | { | |
1145 | size_t k; | |
1146 | ||
1147 | if (!first_change) | |
1148 | return strcmp(s1, s2); | |
1149 | ||
1150 | for (k = 0; s1[k] == s2[k]; k++) | |
1151 | if (s1[k] == '\0') | |
1152 | break; | |
1153 | ||
1154 | *first_change = k; | |
1155 | return (unsigned char)s1[k] - (unsigned char)s2[k]; | |
1156 | } | |
1157 | ||
12676608 LT |
1158 | /* |
1159 | * Do we have another file with a pathname that is a proper | |
1160 | * subset of the name we're trying to add? | |
06b6d81b JH |
1161 | * |
1162 | * That is, is there another file in the index with a path | |
1163 | * that matches a sub-directory in the given entry? | |
12676608 | 1164 | */ |
4aab5b46 JH |
1165 | static int has_dir_name(struct index_state *istate, |
1166 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
12676608 LT |
1167 | { |
1168 | int retval = 0; | |
b155725d | 1169 | int stage = ce_stage(ce); |
12676608 LT |
1170 | const char *name = ce->name; |
1171 | const char *slash = name + ce_namelen(ce); | |
06b6d81b JH |
1172 | size_t len_eq_last; |
1173 | int cmp_last = 0; | |
1174 | ||
1175 | /* | |
1176 | * We are frequently called during an iteration on a sorted | |
1177 | * list of pathnames and while building a new index. Therefore, | |
1178 | * there is a high probability that this entry will eventually | |
1179 | * be appended to the index, rather than inserted in the middle. | |
1180 | * If we can confirm that, we can avoid binary searches on the | |
1181 | * components of the pathname. | |
1182 | * | |
1183 | * Compare the entry's full path with the last path in the index. | |
1184 | */ | |
1185 | if (istate->cache_nr > 0) { | |
1186 | cmp_last = strcmp_offset(name, | |
1187 | istate->cache[istate->cache_nr - 1]->name, | |
1188 | &len_eq_last); | |
1189 | if (cmp_last > 0) { | |
1190 | if (len_eq_last == 0) { | |
1191 | /* | |
1192 | * The entry sorts AFTER the last one in the | |
1193 | * index and their paths have no common prefix, | |
1194 | * so there cannot be a F/D conflict. | |
1195 | */ | |
1196 | return retval; | |
1197 | } else { | |
1198 | /* | |
1199 | * The entry sorts AFTER the last one in the | |
1200 | * index, but has a common prefix. Fall through | |
1201 | * to the loop below to disect the entry's path | |
1202 | * and see where the difference is. | |
1203 | */ | |
1204 | } | |
1205 | } else if (cmp_last == 0) { | |
1206 | /* | |
1207 | * The entry exactly matches the last one in the | |
1208 | * index, but because of multiple stage and CE_REMOVE | |
1209 | * items, we fall through and let the regular search | |
1210 | * code handle it. | |
1211 | */ | |
1212 | } | |
1213 | } | |
0f1e4f04 | 1214 | |
12676608 | 1215 | for (;;) { |
b986df5c | 1216 | size_t len; |
0f1e4f04 | 1217 | |
12676608 LT |
1218 | for (;;) { |
1219 | if (*--slash == '/') | |
1220 | break; | |
1221 | if (slash <= ce->name) | |
1222 | return retval; | |
1223 | } | |
1224 | len = slash - name; | |
0f1e4f04 | 1225 | |
b986df5c JH |
1226 | if (cmp_last > 0) { |
1227 | /* | |
1228 | * (len + 1) is a directory boundary (including | |
1229 | * the trailing slash). And since the loop is | |
1230 | * decrementing "slash", the first iteration is | |
1231 | * the longest directory prefix; subsequent | |
1232 | * iterations consider parent directories. | |
1233 | */ | |
1234 | ||
1235 | if (len + 1 <= len_eq_last) { | |
1236 | /* | |
1237 | * The directory prefix (including the trailing | |
1238 | * slash) also appears as a prefix in the last | |
1239 | * entry, so the remainder cannot collide (because | |
1240 | * strcmp said the whole path was greater). | |
1241 | * | |
1242 | * EQ: last: xxx/A | |
1243 | * this: xxx/B | |
1244 | * | |
1245 | * LT: last: xxx/file_A | |
1246 | * this: xxx/file_B | |
1247 | */ | |
1248 | return retval; | |
1249 | } | |
1250 | ||
1251 | if (len > len_eq_last) { | |
1252 | /* | |
1253 | * This part of the directory prefix (excluding | |
1254 | * the trailing slash) is longer than the known | |
1255 | * equal portions, so this sub-directory cannot | |
1256 | * collide with a file. | |
1257 | * | |
1258 | * GT: last: xxxA | |
1259 | * this: xxxB/file | |
1260 | */ | |
1261 | return retval; | |
1262 | } | |
1263 | ||
b986df5c JH |
1264 | /* |
1265 | * This is a possible collision. Fall through and | |
1266 | * let the regular search code handle it. | |
1267 | * | |
1268 | * last: xxx | |
1269 | * this: xxx/file | |
1270 | */ | |
1271 | } | |
1272 | ||
20ec2d03 | 1273 | pos = index_name_stage_pos(istate, name, len, stage, EXPAND_SPARSE); |
12676608 | 1274 | if (pos >= 0) { |
21cd8d00 JH |
1275 | /* |
1276 | * Found one, but not so fast. This could | |
1277 | * be a marker that says "I was here, but | |
1278 | * I am being removed". Such an entry is | |
1279 | * not a part of the resulting tree, and | |
1280 | * it is Ok to have a directory at the same | |
1281 | * path. | |
1282 | */ | |
077c48df | 1283 | if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) { |
21cd8d00 JH |
1284 | retval = -1; |
1285 | if (!ok_to_replace) | |
1286 | break; | |
4aab5b46 | 1287 | remove_index_entry_at(istate, pos); |
21cd8d00 JH |
1288 | continue; |
1289 | } | |
12676608 | 1290 | } |
21cd8d00 JH |
1291 | else |
1292 | pos = -pos-1; | |
12676608 LT |
1293 | |
1294 | /* | |
1295 | * Trivial optimization: if we find an entry that | |
1296 | * already matches the sub-directory, then we know | |
b155725d | 1297 | * we're ok, and we can exit. |
12676608 | 1298 | */ |
4aab5b46 JH |
1299 | while (pos < istate->cache_nr) { |
1300 | struct cache_entry *p = istate->cache[pos]; | |
b155725d JH |
1301 | if ((ce_namelen(p) <= len) || |
1302 | (p->name[len] != '/') || | |
1303 | memcmp(p->name, name, len)) | |
1304 | break; /* not our subdirectory */ | |
077c48df JH |
1305 | if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE)) |
1306 | /* | |
1307 | * p is at the same stage as our entry, and | |
b155725d JH |
1308 | * is a subdirectory of what we are looking |
1309 | * at, so we cannot have conflicts at our | |
1310 | * level or anything shorter. | |
1311 | */ | |
1312 | return retval; | |
1313 | pos++; | |
192268c1 | 1314 | } |
0f1e4f04 | 1315 | } |
12676608 LT |
1316 | return retval; |
1317 | } | |
1318 | ||
1319 | /* We may be in a situation where we already have path/file and path | |
1320 | * is being added, or we already have path and path/file is being | |
1321 | * added. Either one would result in a nonsense tree that has path | |
1322 | * twice when git-write-tree tries to write it out. Prevent it. | |
a6080a0a | 1323 | * |
12676608 LT |
1324 | * If ok-to-replace is specified, we remove the conflicting entries |
1325 | * from the cache so the caller should recompute the insert position. | |
1326 | * When this happens, we return non-zero. | |
1327 | */ | |
4aab5b46 JH |
1328 | static int check_file_directory_conflict(struct index_state *istate, |
1329 | const struct cache_entry *ce, | |
1330 | int pos, int ok_to_replace) | |
12676608 | 1331 | { |
21cd8d00 JH |
1332 | int retval; |
1333 | ||
1334 | /* | |
1335 | * When ce is an "I am going away" entry, we allow it to be added | |
1336 | */ | |
7a51ed66 | 1337 | if (ce->ce_flags & CE_REMOVE) |
21cd8d00 JH |
1338 | return 0; |
1339 | ||
12676608 LT |
1340 | /* |
1341 | * We check if the path is a sub-path of a subsequent pathname | |
1342 | * first, since removing those will not change the position | |
21cd8d00 | 1343 | * in the array. |
12676608 | 1344 | */ |
4aab5b46 | 1345 | retval = has_file_name(istate, ce, pos, ok_to_replace); |
21cd8d00 | 1346 | |
12676608 LT |
1347 | /* |
1348 | * Then check if the path might have a clashing sub-directory | |
1349 | * before it. | |
1350 | */ | |
4aab5b46 | 1351 | return retval + has_dir_name(istate, ce, pos, ok_to_replace); |
0f1e4f04 JH |
1352 | } |
1353 | ||
af3785dc | 1354 | static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option) |
197ee8c9 LT |
1355 | { |
1356 | int pos; | |
192268c1 JH |
1357 | int ok_to_add = option & ADD_CACHE_OK_TO_ADD; |
1358 | int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE; | |
b155725d | 1359 | int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK; |
39425819 | 1360 | int new_only = option & ADD_CACHE_NEW_ONLY; |
5f73076c | 1361 | |
e5494631 JH |
1362 | /* |
1363 | * If this entry's path sorts after the last entry in the index, | |
1364 | * we can avoid searching for it. | |
1365 | */ | |
1366 | if (istate->cache_nr > 0 && | |
1367 | strcmp(ce->name, istate->cache[istate->cache_nr - 1]->name) > 0) | |
c097b95a | 1368 | pos = index_pos_to_insert_pos(istate->cache_nr); |
e5494631 | 1369 | else |
20ec2d03 | 1370 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce), EXPAND_SPARSE); |
197ee8c9 | 1371 | |
c35e9f5e VD |
1372 | /* |
1373 | * Cache tree path should be invalidated only after index_name_stage_pos, | |
1374 | * in case it expands a sparse index. | |
1375 | */ | |
1376 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) | |
1377 | cache_tree_invalidate_path(istate, ce->name); | |
1378 | ||
3e09cdfd | 1379 | /* existing match? Just replace it. */ |
76e7f4ec | 1380 | if (pos >= 0) { |
39425819 JH |
1381 | if (!new_only) |
1382 | replace_index_entry(istate, pos, ce); | |
197ee8c9 LT |
1383 | return 0; |
1384 | } | |
76e7f4ec | 1385 | pos = -pos-1; |
197ee8c9 | 1386 | |
ffcc9ba7 NTND |
1387 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1388 | untracked_cache_add_to_index(istate, ce->name); | |
e931371a | 1389 | |
7b937ca3 LT |
1390 | /* |
1391 | * Inserting a merged entry ("stage 0") into the index | |
1392 | * will always replace all non-merged entries.. | |
1393 | */ | |
4aab5b46 JH |
1394 | if (pos < istate->cache_nr && ce_stage(ce) == 0) { |
1395 | while (ce_same_name(istate->cache[pos], ce)) { | |
7b937ca3 | 1396 | ok_to_add = 1; |
4aab5b46 | 1397 | if (!remove_index_entry_at(istate, pos)) |
7b937ca3 LT |
1398 | break; |
1399 | } | |
1400 | } | |
1401 | ||
121481ab LT |
1402 | if (!ok_to_add) |
1403 | return -1; | |
c8ad9d04 | 1404 | if (verify_path_internal(ce->name, ce->ce_mode) == PATH_INVALID) |
9d0a9e90 | 1405 | return error(_("invalid path '%s'"), ce->name); |
121481ab | 1406 | |
3e09cdfd | 1407 | if (!skip_df_check && |
4aab5b46 | 1408 | check_file_directory_conflict(istate, ce, pos, ok_to_replace)) { |
192268c1 | 1409 | if (!ok_to_replace) |
9d0a9e90 | 1410 | return error(_("'%s' appears as both a file and as a directory"), |
4aab5b46 | 1411 | ce->name); |
20ec2d03 | 1412 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce), EXPAND_SPARSE); |
192268c1 JH |
1413 | pos = -pos-1; |
1414 | } | |
af3785dc JH |
1415 | return pos + 1; |
1416 | } | |
1417 | ||
1418 | int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option) | |
1419 | { | |
1420 | int pos; | |
1421 | ||
1422 | if (option & ADD_CACHE_JUST_APPEND) | |
1423 | pos = istate->cache_nr; | |
1424 | else { | |
1425 | int ret; | |
1426 | ret = add_index_entry_with_check(istate, ce, option); | |
1427 | if (ret <= 0) | |
1428 | return ret; | |
1429 | pos = ret - 1; | |
1430 | } | |
0f1e4f04 | 1431 | |
197ee8c9 | 1432 | /* Make sure the array is big enough .. */ |
999f5660 | 1433 | ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc); |
197ee8c9 LT |
1434 | |
1435 | /* Add it in.. */ | |
4aab5b46 | 1436 | istate->cache_nr++; |
af3785dc | 1437 | if (istate->cache_nr > pos + 1) |
f919ffeb SG |
1438 | MOVE_ARRAY(istate->cache + pos + 1, istate->cache + pos, |
1439 | istate->cache_nr - pos - 1); | |
cf558704 | 1440 | set_index_entry(istate, pos, ce); |
e636a7b4 | 1441 | istate->cache_changed |= CE_ENTRY_ADDED; |
197ee8c9 LT |
1442 | return 0; |
1443 | } | |
1444 | ||
405e5b2f LT |
1445 | /* |
1446 | * "refresh" does not calculate a new sha1 file or bring the | |
1447 | * cache up-to-date for mode/content changes. But what it | |
1448 | * _does_ do is to "re-match" the stat information of a file | |
1449 | * with the cache, so that you can refresh the cache for a | |
1450 | * file that hasn't been changed but where the stat entry is | |
1451 | * out of date. | |
1452 | * | |
1453 | * For example, you'd want to do this after doing a "git-read-tree", | |
1454 | * to link up the stat cache details with the proper files. | |
1455 | */ | |
4aab5b46 | 1456 | static struct cache_entry *refresh_cache_ent(struct index_state *istate, |
4bd5b7da | 1457 | struct cache_entry *ce, |
d05e6970 | 1458 | unsigned int options, int *err, |
a98e0f2d | 1459 | int *changed_ret, |
15268d12 JH |
1460 | int *t2_did_lstat, |
1461 | int *t2_did_scan) | |
405e5b2f LT |
1462 | { |
1463 | struct stat st; | |
1464 | struct cache_entry *updated; | |
a849735b | 1465 | int changed; |
25762726 | 1466 | int refresh = options & CE_MATCH_REFRESH; |
4bd5b7da | 1467 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 1468 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
2e2e7ec1 | 1469 | int ignore_missing = options & CE_MATCH_IGNORE_MISSING; |
883e248b | 1470 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
405e5b2f | 1471 | |
25762726 | 1472 | if (!refresh || ce_uptodate(ce)) |
eadb5831 JH |
1473 | return ce; |
1474 | ||
883e248b BP |
1475 | if (!ignore_fsmonitor) |
1476 | refresh_fsmonitor(istate); | |
aa9349d4 | 1477 | /* |
56cac48c NTND |
1478 | * CE_VALID or CE_SKIP_WORKTREE means the user promised us |
1479 | * that the change to the work tree does not matter and told | |
1480 | * us not to worry. | |
aa9349d4 | 1481 | */ |
56cac48c NTND |
1482 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) { |
1483 | ce_mark_uptodate(ce); | |
1484 | return ce; | |
1485 | } | |
aa9349d4 MSO |
1486 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) { |
1487 | ce_mark_uptodate(ce); | |
1488 | return ce; | |
1489 | } | |
883e248b BP |
1490 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) { |
1491 | ce_mark_uptodate(ce); | |
1492 | return ce; | |
1493 | } | |
aa9349d4 | 1494 | |
ccad42d4 RS |
1495 | if (has_symlink_leading_path(ce->name, ce_namelen(ce))) { |
1496 | if (ignore_missing) | |
1497 | return ce; | |
1498 | if (err) | |
1499 | *err = ENOENT; | |
1500 | return NULL; | |
1501 | } | |
1502 | ||
a98e0f2d JH |
1503 | if (t2_did_lstat) |
1504 | *t2_did_lstat = 1; | |
8fd2cb40 | 1505 | if (lstat(ce->name, &st) < 0) { |
2e2e7ec1 BK |
1506 | if (ignore_missing && errno == ENOENT) |
1507 | return ce; | |
ec0cc704 JH |
1508 | if (err) |
1509 | *err = errno; | |
8fd2cb40 JS |
1510 | return NULL; |
1511 | } | |
405e5b2f | 1512 | |
4bd5b7da | 1513 | changed = ie_match_stat(istate, ce, &st, options); |
d05e6970 JK |
1514 | if (changed_ret) |
1515 | *changed_ret = changed; | |
405e5b2f | 1516 | if (!changed) { |
4bd5b7da JH |
1517 | /* |
1518 | * The path is unchanged. If we were told to ignore | |
1519 | * valid bit, then we did the actual stat check and | |
1520 | * found that the entry is unmodified. If the entry | |
1521 | * is not marked VALID, this is the place to mark it | |
1522 | * valid again, under "assume unchanged" mode. | |
1523 | */ | |
1524 | if (ignore_valid && assume_unchanged && | |
7a51ed66 | 1525 | !(ce->ce_flags & CE_VALID)) |
405e5b2f | 1526 | ; /* mark this one VALID again */ |
eadb5831 JH |
1527 | else { |
1528 | /* | |
1529 | * We do not mark the index itself "modified" | |
1530 | * because CE_UPTODATE flag is in-core only; | |
1531 | * we are not going to write this change out. | |
1532 | */ | |
883e248b | 1533 | if (!S_ISGITLINK(ce->ce_mode)) { |
125fd984 | 1534 | ce_mark_uptodate(ce); |
b5a81697 | 1535 | mark_fsmonitor_valid(istate, ce); |
883e248b | 1536 | } |
8fd2cb40 | 1537 | return ce; |
eadb5831 | 1538 | } |
405e5b2f LT |
1539 | } |
1540 | ||
15268d12 JH |
1541 | if (t2_did_scan) |
1542 | *t2_did_scan = 1; | |
4bd5b7da | 1543 | if (ie_modified(istate, ce, &st, options)) { |
ec0cc704 JH |
1544 | if (err) |
1545 | *err = EINVAL; | |
8fd2cb40 JS |
1546 | return NULL; |
1547 | } | |
405e5b2f | 1548 | |
a849735b | 1549 | updated = make_empty_cache_entry(istate, ce_namelen(ce)); |
0e267b7a BP |
1550 | copy_cache_entry(updated, ce); |
1551 | memcpy(updated->name, ce->name, ce->ce_namelen + 1); | |
d4c0a3ac | 1552 | fill_stat_cache_info(istate, updated, &st); |
4bd5b7da JH |
1553 | /* |
1554 | * If ignore_valid is not set, we should leave CE_VALID bit | |
1555 | * alone. Otherwise, paths marked with --no-assume-unchanged | |
1556 | * (i.e. things to be edited) will reacquire CE_VALID bit | |
1557 | * automatically, which is not really what we want. | |
405e5b2f | 1558 | */ |
4bd5b7da | 1559 | if (!ignore_valid && assume_unchanged && |
7a51ed66 LT |
1560 | !(ce->ce_flags & CE_VALID)) |
1561 | updated->ce_flags &= ~CE_VALID; | |
405e5b2f | 1562 | |
e636a7b4 | 1563 | /* istate->cache_changed is updated in the caller */ |
405e5b2f LT |
1564 | return updated; |
1565 | } | |
1566 | ||
3deffc52 | 1567 | static void show_file(const char * fmt, const char * name, int in_porcelain, |
046613c5 | 1568 | int * first, const char *header_msg) |
3deffc52 MM |
1569 | { |
1570 | if (in_porcelain && *first && header_msg) { | |
1571 | printf("%s\n", header_msg); | |
cd2b8ae9 | 1572 | *first = 0; |
3deffc52 MM |
1573 | } |
1574 | printf(fmt, name); | |
1575 | } | |
1576 | ||
22184497 TG |
1577 | int repo_refresh_and_write_index(struct repository *repo, |
1578 | unsigned int refresh_flags, | |
1579 | unsigned int write_flags, | |
1580 | int gentle, | |
1581 | const struct pathspec *pathspec, | |
1582 | char *seen, const char *header_msg) | |
1583 | { | |
1584 | struct lock_file lock_file = LOCK_INIT; | |
1585 | int fd, ret = 0; | |
1586 | ||
1587 | fd = repo_hold_locked_index(repo, &lock_file, 0); | |
1588 | if (!gentle && fd < 0) | |
1589 | return -1; | |
1590 | if (refresh_index(repo->index, refresh_flags, pathspec, seen, header_msg)) | |
1591 | ret = 1; | |
1592 | if (0 <= fd && write_locked_index(repo->index, &lock_file, COMMIT_LOCK | write_flags)) | |
1593 | ret = -1; | |
1594 | return ret; | |
1595 | } | |
1596 | ||
1597 | ||
9b2d6149 NTND |
1598 | int refresh_index(struct index_state *istate, unsigned int flags, |
1599 | const struct pathspec *pathspec, | |
046613c5 | 1600 | char *seen, const char *header_msg) |
405e5b2f LT |
1601 | { |
1602 | int i; | |
1603 | int has_errors = 0; | |
1604 | int really = (flags & REFRESH_REALLY) != 0; | |
1605 | int allow_unmerged = (flags & REFRESH_UNMERGED) != 0; | |
1606 | int quiet = (flags & REFRESH_QUIET) != 0; | |
1607 | int not_new = (flags & REFRESH_IGNORE_MISSING) != 0; | |
5fdeacb0 | 1608 | int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0; |
b243012c | 1609 | int ignore_skip_worktree = (flags & REFRESH_IGNORE_SKIP_WORKTREE) != 0; |
3deffc52 MM |
1610 | int first = 1; |
1611 | int in_porcelain = (flags & REFRESH_IN_PORCELAIN); | |
25762726 BK |
1612 | unsigned int options = (CE_MATCH_REFRESH | |
1613 | (really ? CE_MATCH_IGNORE_VALID : 0) | | |
2e2e7ec1 | 1614 | (not_new ? CE_MATCH_IGNORE_MISSING : 0)); |
4bd4e730 | 1615 | const char *modified_fmt; |
73b7eae6 JK |
1616 | const char *deleted_fmt; |
1617 | const char *typechange_fmt; | |
1618 | const char *added_fmt; | |
4bd4e730 | 1619 | const char *unmerged_fmt; |
ae9af122 | 1620 | struct progress *progress = NULL; |
a98e0f2d | 1621 | int t2_sum_lstat = 0; |
15268d12 | 1622 | int t2_sum_scan = 0; |
ae9af122 NTND |
1623 | |
1624 | if (flags & REFRESH_PROGRESS && isatty(2)) | |
1625 | progress = start_delayed_progress(_("Refresh index"), | |
1626 | istate->cache_nr); | |
405e5b2f | 1627 | |
c46c406a | 1628 | trace_performance_enter(); |
a71806a7 NTND |
1629 | modified_fmt = in_porcelain ? "M\t%s\n" : "%s: needs update\n"; |
1630 | deleted_fmt = in_porcelain ? "D\t%s\n" : "%s: needs update\n"; | |
1631 | typechange_fmt = in_porcelain ? "T\t%s\n" : "%s: needs update\n"; | |
1632 | added_fmt = in_porcelain ? "A\t%s\n" : "%s: needs update\n"; | |
1633 | unmerged_fmt = in_porcelain ? "U\t%s\n" : "%s: needs merge\n"; | |
99ce720c BP |
1634 | /* |
1635 | * Use the multi-threaded preload_index() to refresh most of the | |
1636 | * cache entries quickly then in the single threaded loop below, | |
1637 | * we only have to do the special cases that are left. | |
1638 | */ | |
1639 | preload_index(istate, pathspec, 0); | |
a98e0f2d | 1640 | trace2_region_enter("index", "refresh", NULL); |
d76723ee | 1641 | |
4aab5b46 | 1642 | for (i = 0; i < istate->cache_nr; i++) { |
285c2e25 | 1643 | struct cache_entry *ce, *new_entry; |
ec0cc704 | 1644 | int cache_errno = 0; |
73b7eae6 | 1645 | int changed = 0; |
3d1f148c | 1646 | int filtered = 0; |
a98e0f2d | 1647 | int t2_did_lstat = 0; |
15268d12 | 1648 | int t2_did_scan = 0; |
ec0cc704 | 1649 | |
4aab5b46 | 1650 | ce = istate->cache[i]; |
5fdeacb0 JS |
1651 | if (ignore_submodules && S_ISGITLINK(ce->ce_mode)) |
1652 | continue; | |
b243012c MT |
1653 | if (ignore_skip_worktree && ce_skip_worktree(ce)) |
1654 | continue; | |
5fdeacb0 | 1655 | |
d76723ee DS |
1656 | /* |
1657 | * If this entry is a sparse directory, then there isn't | |
1658 | * any stat() information to update. Ignore the entry. | |
1659 | */ | |
1660 | if (S_ISSPARSEDIR(ce->ce_mode)) | |
1661 | continue; | |
1662 | ||
d7b665c3 | 1663 | if (pathspec && !ce_path_match(istate, ce, pathspec, seen)) |
3d1f148c JH |
1664 | filtered = 1; |
1665 | ||
405e5b2f | 1666 | if (ce_stage(ce)) { |
4aab5b46 JH |
1667 | while ((i < istate->cache_nr) && |
1668 | ! strcmp(istate->cache[i]->name, ce->name)) | |
405e5b2f LT |
1669 | i++; |
1670 | i--; | |
1671 | if (allow_unmerged) | |
1672 | continue; | |
3d1f148c JH |
1673 | if (!filtered) |
1674 | show_file(unmerged_fmt, ce->name, in_porcelain, | |
1675 | &first, header_msg); | |
405e5b2f LT |
1676 | has_errors = 1; |
1677 | continue; | |
1678 | } | |
1679 | ||
3d1f148c | 1680 | if (filtered) |
d616813d AJ |
1681 | continue; |
1682 | ||
a98e0f2d JH |
1683 | new_entry = refresh_cache_ent(istate, ce, options, |
1684 | &cache_errno, &changed, | |
15268d12 | 1685 | &t2_did_lstat, &t2_did_scan); |
a98e0f2d | 1686 | t2_sum_lstat += t2_did_lstat; |
15268d12 | 1687 | t2_sum_scan += t2_did_scan; |
285c2e25 | 1688 | if (new_entry == ce) |
405e5b2f | 1689 | continue; |
b7b793d1 | 1690 | display_progress(progress, i); |
285c2e25 | 1691 | if (!new_entry) { |
73b7eae6 JK |
1692 | const char *fmt; |
1693 | ||
8fd2cb40 | 1694 | if (really && cache_errno == EINVAL) { |
405e5b2f LT |
1695 | /* If we are doing --really-refresh that |
1696 | * means the index is not valid anymore. | |
1697 | */ | |
7a51ed66 | 1698 | ce->ce_flags &= ~CE_VALID; |
078a58e8 | 1699 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 1700 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 1701 | istate->cache_changed |= CE_ENTRY_CHANGED; |
405e5b2f LT |
1702 | } |
1703 | if (quiet) | |
1704 | continue; | |
73b7eae6 JK |
1705 | |
1706 | if (cache_errno == ENOENT) | |
1707 | fmt = deleted_fmt; | |
895ff3b2 | 1708 | else if (ce_intent_to_add(ce)) |
73b7eae6 JK |
1709 | fmt = added_fmt; /* must be before other checks */ |
1710 | else if (changed & TYPE_CHANGED) | |
1711 | fmt = typechange_fmt; | |
1712 | else | |
1713 | fmt = modified_fmt; | |
1714 | show_file(fmt, | |
1715 | ce->name, in_porcelain, &first, header_msg); | |
405e5b2f LT |
1716 | has_errors = 1; |
1717 | continue; | |
1718 | } | |
cf558704 | 1719 | |
285c2e25 | 1720 | replace_index_entry(istate, i, new_entry); |
405e5b2f | 1721 | } |
a98e0f2d | 1722 | trace2_data_intmax("index", NULL, "refresh/sum_lstat", t2_sum_lstat); |
15268d12 | 1723 | trace2_data_intmax("index", NULL, "refresh/sum_scan", t2_sum_scan); |
a98e0f2d | 1724 | trace2_region_leave("index", "refresh", NULL); |
b7b793d1 ÆAB |
1725 | display_progress(progress, istate->cache_nr); |
1726 | stop_progress(&progress); | |
c46c406a | 1727 | trace_performance_leave("refresh index"); |
405e5b2f LT |
1728 | return has_errors; |
1729 | } | |
1730 | ||
768d7965 JM |
1731 | struct cache_entry *refresh_cache_entry(struct index_state *istate, |
1732 | struct cache_entry *ce, | |
1733 | unsigned int options) | |
ec0cc704 | 1734 | { |
15268d12 | 1735 | return refresh_cache_ent(istate, ce, options, NULL, NULL, NULL, NULL); |
ec0cc704 JH |
1736 | } |
1737 | ||
db3b313c JH |
1738 | |
1739 | /***************************************************************** | |
1740 | * Index File I/O | |
1741 | *****************************************************************/ | |
1742 | ||
9d227781 JH |
1743 | #define INDEX_FORMAT_DEFAULT 3 |
1744 | ||
7211b9e7 | 1745 | static unsigned int get_index_format_default(struct repository *r) |
136347d7 TG |
1746 | { |
1747 | char *envversion = getenv("GIT_INDEX_VERSION"); | |
3c09d684 TG |
1748 | char *endp; |
1749 | unsigned int version = INDEX_FORMAT_DEFAULT; | |
1750 | ||
136347d7 | 1751 | if (!envversion) { |
7211b9e7 DS |
1752 | prepare_repo_settings(r); |
1753 | ||
1754 | if (r->settings.index_version >= 0) | |
1755 | version = r->settings.index_version; | |
3c09d684 TG |
1756 | if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { |
1757 | warning(_("index.version set, but the value is invalid.\n" | |
136347d7 | 1758 | "Using version %i"), INDEX_FORMAT_DEFAULT); |
3c09d684 | 1759 | return INDEX_FORMAT_DEFAULT; |
136347d7 TG |
1760 | } |
1761 | return version; | |
1762 | } | |
3c09d684 TG |
1763 | |
1764 | version = strtoul(envversion, &endp, 10); | |
1765 | if (*endp || | |
1766 | version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { | |
1767 | warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n" | |
1768 | "Using version %i"), INDEX_FORMAT_DEFAULT); | |
1769 | version = INDEX_FORMAT_DEFAULT; | |
1770 | } | |
1771 | return version; | |
136347d7 TG |
1772 | } |
1773 | ||
db3b313c JH |
1774 | /* |
1775 | * dev/ino/uid/gid/size are also just tracked to the low 32 bits | |
1776 | * Again - this is just a (very strong in practice) heuristic that | |
1777 | * the inode hasn't changed. | |
1778 | * | |
1779 | * We save the fields in big-endian order to allow using the | |
1780 | * index file over NFS transparently. | |
1781 | */ | |
1782 | struct ondisk_cache_entry { | |
1783 | struct cache_time ctime; | |
1784 | struct cache_time mtime; | |
7800c1eb TG |
1785 | uint32_t dev; |
1786 | uint32_t ino; | |
1787 | uint32_t mode; | |
1788 | uint32_t uid; | |
1789 | uint32_t gid; | |
1790 | uint32_t size; | |
575fa8a3 | 1791 | /* |
1792 | * unsigned char hash[hashsz]; | |
1793 | * uint16_t flags; | |
1794 | * if (flags & CE_EXTENDED) | |
1795 | * uint16_t flags2; | |
1796 | */ | |
1797 | unsigned char data[GIT_MAX_RAWSZ + 2 * sizeof(uint16_t)]; | |
1798 | char name[FLEX_ARRAY]; | |
db3b313c JH |
1799 | }; |
1800 | ||
6c9cd161 | 1801 | /* These are only used for v3 or lower */ |
ce012deb | 1802 | #define align_padding_size(size, len) ((size + (len) + 8) & ~7) - (size + len) |
575fa8a3 | 1803 | #define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,data) + (len) + 8) & ~7) |
db3b313c | 1804 | #define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len) |
575fa8a3 | 1805 | #define ondisk_data_size(flags, len) (the_hash_algo->rawsz + \ |
1806 | ((flags & CE_EXTENDED) ? 2 : 1) * sizeof(uint16_t) + len) | |
1807 | #define ondisk_data_size_max(len) (ondisk_data_size(CE_EXTENDED, len)) | |
1808 | #define ondisk_ce_size(ce) (ondisk_cache_entry_size(ondisk_data_size((ce)->ce_flags, ce_namelen(ce)))) | |
db3b313c | 1809 | |
a33fc72f JH |
1810 | /* Allow fsck to force verification of the index checksum. */ |
1811 | int verify_index_checksum; | |
1812 | ||
00ec50e5 BP |
1813 | /* Allow fsck to force verification of the cache entry order. */ |
1814 | int verify_ce_order; | |
1815 | ||
371ed0de | 1816 | static int verify_hdr(const struct cache_header *hdr, unsigned long size) |
e83c5163 | 1817 | { |
aab61359 | 1818 | git_hash_ctx c; |
1819 | unsigned char hash[GIT_MAX_RAWSZ]; | |
0136bac9 | 1820 | int hdr_version; |
ee1f0c24 DS |
1821 | unsigned char *start, *end; |
1822 | struct object_id oid; | |
e83c5163 | 1823 | |
ccc4feb5 | 1824 | if (hdr->hdr_signature != htonl(CACHE_SIGNATURE)) |
9d0a9e90 | 1825 | return error(_("bad signature 0x%08x"), hdr->hdr_signature); |
0136bac9 | 1826 | hdr_version = ntohl(hdr->hdr_version); |
b82a7b5b | 1827 | if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version) |
9d0a9e90 | 1828 | return error(_("bad index version %d"), hdr_version); |
a33fc72f JH |
1829 | |
1830 | if (!verify_index_checksum) | |
1831 | return 0; | |
1832 | ||
ee1f0c24 DS |
1833 | end = (unsigned char *)hdr + size; |
1834 | start = end - the_hash_algo->rawsz; | |
1835 | oidread(&oid, start); | |
1836 | if (oideq(&oid, null_oid())) | |
1837 | return 0; | |
1838 | ||
aab61359 | 1839 | the_hash_algo->init_fn(&c); |
1840 | the_hash_algo->update_fn(&c, hdr, size - the_hash_algo->rawsz); | |
1841 | the_hash_algo->final_fn(hash, &c); | |
ee1f0c24 | 1842 | if (!hasheq(hash, start)) |
9d0a9e90 | 1843 | return error(_("bad index file sha1 signature")); |
e83c5163 LT |
1844 | return 0; |
1845 | } | |
1846 | ||
4aab5b46 | 1847 | static int read_index_extension(struct index_state *istate, |
371ed0de | 1848 | const char *ext, const char *data, unsigned long sz) |
bad68ec9 JH |
1849 | { |
1850 | switch (CACHE_EXT(ext)) { | |
1851 | case CACHE_EXT_TREE: | |
4aab5b46 | 1852 | istate->cache_tree = cache_tree_read(data, sz); |
bad68ec9 | 1853 | break; |
cfc5789a JH |
1854 | case CACHE_EXT_RESOLVE_UNDO: |
1855 | istate->resolve_undo = resolve_undo_read(data, sz); | |
1856 | break; | |
5fc2fc8f NTND |
1857 | case CACHE_EXT_LINK: |
1858 | if (read_link_extension(istate, data, sz)) | |
1859 | return -1; | |
1860 | break; | |
f9e6c649 NTND |
1861 | case CACHE_EXT_UNTRACKED: |
1862 | istate->untracked = read_untracked_extension(data, sz); | |
1863 | break; | |
883e248b BP |
1864 | case CACHE_EXT_FSMONITOR: |
1865 | read_fsmonitor_extension(istate, data, sz); | |
1866 | break; | |
3b1d9e04 | 1867 | case CACHE_EXT_ENDOFINDEXENTRIES: |
3255089a | 1868 | case CACHE_EXT_INDEXENTRYOFFSETTABLE: |
3b1d9e04 BP |
1869 | /* already handled in do_read_index() */ |
1870 | break; | |
cd42415f DS |
1871 | case CACHE_EXT_SPARSE_DIRECTORIES: |
1872 | /* no content, only an indicator */ | |
9fadb373 | 1873 | istate->sparse_index = INDEX_COLLAPSED; |
cd42415f | 1874 | break; |
bad68ec9 JH |
1875 | default: |
1876 | if (*ext < 'A' || 'Z' < *ext) | |
9d0a9e90 | 1877 | return error(_("index uses %.4s extension, which we do not understand"), |
bad68ec9 | 1878 | ext); |
9d0a9e90 | 1879 | fprintf_ln(stderr, _("ignoring %.4s extension"), ext); |
bad68ec9 JH |
1880 | break; |
1881 | } | |
1882 | return 0; | |
1883 | } | |
1884 | ||
4a6ed30f VD |
1885 | /* |
1886 | * Parses the contents of the cache entry contained within the 'ondisk' buffer | |
1887 | * into a new incore 'cache_entry'. | |
1888 | * | |
1889 | * Note that 'char *ondisk' may not be aligned to a 4-byte address interval in | |
1890 | * index v4, so we cannot cast it to 'struct ondisk_cache_entry *' and access | |
1891 | * its members. Instead, we use the byte offsets of members within the struct to | |
1892 | * identify where 'get_be16()', 'get_be32()', and 'oidread()' (which can all | |
1893 | * read from an unaligned memory buffer) should read from the 'ondisk' buffer | |
1894 | * into the corresponding incore 'cache_entry' members. | |
1895 | */ | |
77ff1127 BP |
1896 | static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool, |
1897 | unsigned int version, | |
4a6ed30f | 1898 | const char *ondisk, |
6c9cd161 | 1899 | unsigned long *ent_size, |
252d079c | 1900 | const struct cache_entry *previous_ce) |
7a51ed66 | 1901 | { |
debed2a6 | 1902 | struct cache_entry *ce; |
7fec10b7 | 1903 | size_t len; |
06aaaa0b | 1904 | const char *name; |
575fa8a3 | 1905 | const unsigned hashsz = the_hash_algo->rawsz; |
4a6ed30f | 1906 | const char *flagsp = ondisk + offsetof(struct ondisk_cache_entry, data) + hashsz; |
debed2a6 | 1907 | unsigned int flags; |
f5c4a9af | 1908 | size_t copy_len = 0; |
252d079c NTND |
1909 | /* |
1910 | * Adjacent cache entries tend to share the leading paths, so it makes | |
1911 | * sense to only store the differences in later entries. In the v4 | |
1912 | * on-disk format of the index, each on-disk cache entry stores the | |
1913 | * number of bytes to be stripped from the end of the previous name, | |
1914 | * and the bytes to append to the result, to come up with its name. | |
1915 | */ | |
77ff1127 | 1916 | int expand_name_field = version == 4; |
7fec10b7 | 1917 | |
7a51ed66 | 1918 | /* On-disk flags are just 16 bits */ |
575fa8a3 | 1919 | flags = get_be16(flagsp); |
debed2a6 | 1920 | len = flags & CE_NAMEMASK; |
7fec10b7 | 1921 | |
debed2a6 | 1922 | if (flags & CE_EXTENDED) { |
06aaaa0b | 1923 | int extended_flags; |
4a6ed30f | 1924 | extended_flags = get_be16(flagsp + sizeof(uint16_t)) << 16; |
06aaaa0b NTND |
1925 | /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */ |
1926 | if (extended_flags & ~CE_EXTENDED_FLAGS) | |
9d0a9e90 | 1927 | die(_("unknown index entry format 0x%08x"), extended_flags); |
debed2a6 | 1928 | flags |= extended_flags; |
4a6ed30f | 1929 | name = (const char *)(flagsp + 2 * sizeof(uint16_t)); |
06aaaa0b NTND |
1930 | } |
1931 | else | |
4a6ed30f | 1932 | name = (const char *)(flagsp + sizeof(uint16_t)); |
06aaaa0b | 1933 | |
252d079c NTND |
1934 | if (expand_name_field) { |
1935 | const unsigned char *cp = (const unsigned char *)name; | |
1936 | size_t strip_len, previous_len; | |
6c9cd161 | 1937 | |
15beaaa3 | 1938 | /* If we're at the beginning of a block, ignore the previous name */ |
252d079c | 1939 | strip_len = decode_varint(&cp); |
77ff1127 BP |
1940 | if (previous_ce) { |
1941 | previous_len = previous_ce->ce_namelen; | |
1942 | if (previous_len < strip_len) | |
252d079c | 1943 | die(_("malformed name field in the index, near path '%s'"), |
77ff1127 BP |
1944 | previous_ce->name); |
1945 | copy_len = previous_len - strip_len; | |
252d079c | 1946 | } |
252d079c NTND |
1947 | name = (const char *)cp; |
1948 | } | |
6c9cd161 | 1949 | |
252d079c NTND |
1950 | if (len == CE_NAMEMASK) { |
1951 | len = strlen(name); | |
1952 | if (expand_name_field) | |
1953 | len += copy_len; | |
1954 | } | |
6c9cd161 | 1955 | |
77ff1127 | 1956 | ce = mem_pool__ce_alloc(ce_mem_pool, len); |
252d079c | 1957 | |
4a6ed30f VD |
1958 | /* |
1959 | * NEEDSWORK: using 'offsetof()' is cumbersome and should be replaced | |
1960 | * with something more akin to 'load_bitmap_entries_v1()'s use of | |
1961 | * 'read_be16'/'read_be32'. For consistency with the corresponding | |
1962 | * ondisk entry write function ('copy_cache_entry_to_ondisk()'), this | |
1963 | * should be done at the same time as removing references to | |
1964 | * 'ondisk_cache_entry' there. | |
1965 | */ | |
1966 | ce->ce_stat_data.sd_ctime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime) | |
1967 | + offsetof(struct cache_time, sec)); | |
1968 | ce->ce_stat_data.sd_mtime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime) | |
1969 | + offsetof(struct cache_time, sec)); | |
1970 | ce->ce_stat_data.sd_ctime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime) | |
1971 | + offsetof(struct cache_time, nsec)); | |
1972 | ce->ce_stat_data.sd_mtime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime) | |
1973 | + offsetof(struct cache_time, nsec)); | |
1974 | ce->ce_stat_data.sd_dev = get_be32(ondisk + offsetof(struct ondisk_cache_entry, dev)); | |
1975 | ce->ce_stat_data.sd_ino = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ino)); | |
1976 | ce->ce_mode = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mode)); | |
1977 | ce->ce_stat_data.sd_uid = get_be32(ondisk + offsetof(struct ondisk_cache_entry, uid)); | |
1978 | ce->ce_stat_data.sd_gid = get_be32(ondisk + offsetof(struct ondisk_cache_entry, gid)); | |
1979 | ce->ce_stat_data.sd_size = get_be32(ondisk + offsetof(struct ondisk_cache_entry, size)); | |
252d079c NTND |
1980 | ce->ce_flags = flags & ~CE_NAMEMASK; |
1981 | ce->ce_namelen = len; | |
1982 | ce->index = 0; | |
4a6ed30f | 1983 | oidread(&ce->oid, (const unsigned char *)ondisk + offsetof(struct ondisk_cache_entry, data)); |
6c9cd161 | 1984 | |
252d079c NTND |
1985 | if (expand_name_field) { |
1986 | if (copy_len) | |
1987 | memcpy(ce->name, previous_ce->name, copy_len); | |
1988 | memcpy(ce->name + copy_len, name, len + 1 - copy_len); | |
1989 | *ent_size = (name - ((char *)ondisk)) + len + 1 - copy_len; | |
1990 | } else { | |
1991 | memcpy(ce->name, name, len + 1); | |
1992 | *ent_size = ondisk_ce_size(ce); | |
6c9cd161 | 1993 | } |
debed2a6 | 1994 | return ce; |
cf558704 LT |
1995 | } |
1996 | ||
03f15a79 | 1997 | static void check_ce_order(struct index_state *istate) |
15999d0b | 1998 | { |
03f15a79 TG |
1999 | unsigned int i; |
2000 | ||
00ec50e5 BP |
2001 | if (!verify_ce_order) |
2002 | return; | |
2003 | ||
03f15a79 TG |
2004 | for (i = 1; i < istate->cache_nr; i++) { |
2005 | struct cache_entry *ce = istate->cache[i - 1]; | |
2006 | struct cache_entry *next_ce = istate->cache[i]; | |
2007 | int name_compare = strcmp(ce->name, next_ce->name); | |
2008 | ||
2009 | if (0 < name_compare) | |
9d0a9e90 | 2010 | die(_("unordered stage entries in index")); |
03f15a79 TG |
2011 | if (!name_compare) { |
2012 | if (!ce_stage(ce)) | |
9d0a9e90 | 2013 | die(_("multiple stage entries for merged file '%s'"), |
03f15a79 TG |
2014 | ce->name); |
2015 | if (ce_stage(ce) > ce_stage(next_ce)) | |
9d0a9e90 | 2016 | die(_("unordered stage entries for '%s'"), |
03f15a79 TG |
2017 | ce->name); |
2018 | } | |
15999d0b JSP |
2019 | } |
2020 | } | |
2021 | ||
435ec090 CC |
2022 | static void tweak_untracked_cache(struct index_state *istate) |
2023 | { | |
ad0fb659 DS |
2024 | struct repository *r = the_repository; |
2025 | ||
2026 | prepare_repo_settings(r); | |
2027 | ||
f1bee828 ÆAB |
2028 | switch (r->settings.core_untracked_cache) { |
2029 | case UNTRACKED_CACHE_REMOVE: | |
435ec090 | 2030 | remove_untracked_cache(istate); |
f1bee828 ÆAB |
2031 | break; |
2032 | case UNTRACKED_CACHE_WRITE: | |
ad0fb659 | 2033 | add_untracked_cache(istate); |
f1bee828 ÆAB |
2034 | break; |
2035 | case UNTRACKED_CACHE_KEEP: | |
3050b6df ÆAB |
2036 | /* |
2037 | * Either an explicit "core.untrackedCache=keep", the | |
2038 | * default if "core.untrackedCache" isn't configured, | |
2039 | * or a fallback on an unknown "core.untrackedCache" | |
2040 | * value. | |
2041 | */ | |
f1bee828 | 2042 | break; |
f1bee828 | 2043 | } |
435ec090 CC |
2044 | } |
2045 | ||
43925312 CC |
2046 | static void tweak_split_index(struct index_state *istate) |
2047 | { | |
2048 | switch (git_config_get_split_index()) { | |
2049 | case -1: /* unset: do nothing */ | |
2050 | break; | |
2051 | case 0: /* false */ | |
2052 | remove_split_index(istate); | |
2053 | break; | |
2054 | case 1: /* true */ | |
2055 | add_split_index(istate); | |
2056 | break; | |
2057 | default: /* unknown value: do nothing */ | |
2058 | break; | |
2059 | } | |
2060 | } | |
2061 | ||
435ec090 CC |
2062 | static void post_read_index_from(struct index_state *istate) |
2063 | { | |
2064 | check_ce_order(istate); | |
2065 | tweak_untracked_cache(istate); | |
43925312 | 2066 | tweak_split_index(istate); |
883e248b | 2067 | tweak_fsmonitor(istate); |
435ec090 CC |
2068 | } |
2069 | ||
8e72d675 JM |
2070 | static size_t estimate_cache_size_from_compressed(unsigned int entries) |
2071 | { | |
2072 | return entries * (sizeof(struct cache_entry) + CACHE_ENTRY_PATH_LENGTH); | |
2073 | } | |
2074 | ||
2075 | static size_t estimate_cache_size(size_t ondisk_size, unsigned int entries) | |
2076 | { | |
2077 | long per_entry = sizeof(struct cache_entry) - sizeof(struct ondisk_cache_entry); | |
2078 | ||
2079 | /* | |
2080 | * Account for potential alignment differences. | |
2081 | */ | |
c097b95a | 2082 | per_entry += align_padding_size(per_entry, 0); |
8e72d675 JM |
2083 | return ondisk_size + entries * per_entry; |
2084 | } | |
2085 | ||
3255089a BP |
2086 | struct index_entry_offset |
2087 | { | |
2088 | /* starting byte offset into index file, count of index entries in this block */ | |
2089 | int offset, nr; | |
2090 | }; | |
2091 | ||
2092 | struct index_entry_offset_table | |
2093 | { | |
2094 | int nr; | |
2095 | struct index_entry_offset entries[FLEX_ARRAY]; | |
2096 | }; | |
2097 | ||
3255089a BP |
2098 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset); |
2099 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot); | |
3255089a | 2100 | |
3b1d9e04 BP |
2101 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size); |
2102 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset); | |
2103 | ||
abb4bb83 BP |
2104 | struct load_index_extensions |
2105 | { | |
abb4bb83 | 2106 | pthread_t pthread; |
abb4bb83 BP |
2107 | struct index_state *istate; |
2108 | const char *mmap; | |
2109 | size_t mmap_size; | |
2110 | unsigned long src_offset; | |
2111 | }; | |
2112 | ||
2113 | static void *load_index_extensions(void *_data) | |
2114 | { | |
2115 | struct load_index_extensions *p = _data; | |
2116 | unsigned long src_offset = p->src_offset; | |
2117 | ||
2118 | while (src_offset <= p->mmap_size - the_hash_algo->rawsz - 8) { | |
2119 | /* After an array of active_nr index entries, | |
2120 | * there can be arbitrary number of extended | |
2121 | * sections, each of which is prefixed with | |
2122 | * extension name (4-byte) and section length | |
2123 | * in 4-byte network byte order. | |
2124 | */ | |
2125 | uint32_t extsize = get_be32(p->mmap + src_offset + 4); | |
2126 | if (read_index_extension(p->istate, | |
2127 | p->mmap + src_offset, | |
2128 | p->mmap + src_offset + 8, | |
2129 | extsize) < 0) { | |
2130 | munmap((void *)p->mmap, p->mmap_size); | |
2131 | die(_("index file corrupt")); | |
2132 | } | |
2133 | src_offset += 8; | |
2134 | src_offset += extsize; | |
2135 | } | |
2136 | ||
2137 | return NULL; | |
2138 | } | |
2139 | ||
77ff1127 BP |
2140 | /* |
2141 | * A helper function that will load the specified range of cache entries | |
2142 | * from the memory mapped file and add them to the given index. | |
2143 | */ | |
2144 | static unsigned long load_cache_entry_block(struct index_state *istate, | |
2145 | struct mem_pool *ce_mem_pool, int offset, int nr, const char *mmap, | |
2146 | unsigned long start_offset, const struct cache_entry *previous_ce) | |
2147 | { | |
2148 | int i; | |
2149 | unsigned long src_offset = start_offset; | |
2150 | ||
2151 | for (i = offset; i < offset + nr; i++) { | |
77ff1127 BP |
2152 | struct cache_entry *ce; |
2153 | unsigned long consumed; | |
2154 | ||
4a6ed30f VD |
2155 | ce = create_from_disk(ce_mem_pool, istate->version, |
2156 | mmap + src_offset, | |
2157 | &consumed, previous_ce); | |
77ff1127 BP |
2158 | set_index_entry(istate, i, ce); |
2159 | ||
2160 | src_offset += consumed; | |
2161 | previous_ce = ce; | |
2162 | } | |
2163 | return src_offset - start_offset; | |
2164 | } | |
2165 | ||
2166 | static unsigned long load_all_cache_entries(struct index_state *istate, | |
2167 | const char *mmap, size_t mmap_size, unsigned long src_offset) | |
2168 | { | |
2169 | unsigned long consumed; | |
2170 | ||
44c7e1a7 | 2171 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2172 | if (istate->version == 4) { |
44c7e1a7 | 2173 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2174 | estimate_cache_size_from_compressed(istate->cache_nr)); |
2175 | } else { | |
44c7e1a7 | 2176 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2177 | estimate_cache_size(mmap_size, istate->cache_nr)); |
2178 | } | |
2179 | ||
2180 | consumed = load_cache_entry_block(istate, istate->ce_mem_pool, | |
2181 | 0, istate->cache_nr, mmap, src_offset, NULL); | |
2182 | return consumed; | |
2183 | } | |
2184 | ||
3255089a BP |
2185 | /* |
2186 | * Mostly randomly chosen maximum thread counts: we | |
2187 | * cap the parallelism to online_cpus() threads, and we want | |
2188 | * to have at least 10000 cache entries per thread for it to | |
2189 | * be worth starting a thread. | |
2190 | */ | |
2191 | ||
2192 | #define THREAD_COST (10000) | |
2193 | ||
77ff1127 BP |
2194 | struct load_cache_entries_thread_data |
2195 | { | |
2196 | pthread_t pthread; | |
2197 | struct index_state *istate; | |
2198 | struct mem_pool *ce_mem_pool; | |
2199 | int offset; | |
2200 | const char *mmap; | |
2201 | struct index_entry_offset_table *ieot; | |
2202 | int ieot_start; /* starting index into the ieot array */ | |
2203 | int ieot_blocks; /* count of ieot entries to process */ | |
2204 | unsigned long consumed; /* return # of bytes in index file processed */ | |
2205 | }; | |
2206 | ||
2207 | /* | |
2208 | * A thread proc to run the load_cache_entries() computation | |
2209 | * across multiple background threads. | |
2210 | */ | |
2211 | static void *load_cache_entries_thread(void *_data) | |
2212 | { | |
2213 | struct load_cache_entries_thread_data *p = _data; | |
2214 | int i; | |
2215 | ||
2216 | /* iterate across all ieot blocks assigned to this thread */ | |
2217 | for (i = p->ieot_start; i < p->ieot_start + p->ieot_blocks; i++) { | |
2218 | p->consumed += load_cache_entry_block(p->istate, p->ce_mem_pool, | |
2219 | p->offset, p->ieot->entries[i].nr, p->mmap, p->ieot->entries[i].offset, NULL); | |
2220 | p->offset += p->ieot->entries[i].nr; | |
2221 | } | |
2222 | return NULL; | |
2223 | } | |
2224 | ||
2225 | static unsigned long load_cache_entries_threaded(struct index_state *istate, const char *mmap, size_t mmap_size, | |
7bd9631b | 2226 | int nr_threads, struct index_entry_offset_table *ieot) |
77ff1127 BP |
2227 | { |
2228 | int i, offset, ieot_blocks, ieot_start, err; | |
2229 | struct load_cache_entries_thread_data *data; | |
2230 | unsigned long consumed = 0; | |
2231 | ||
2232 | /* a little sanity checking */ | |
2233 | if (istate->name_hash_initialized) | |
2234 | BUG("the name hash isn't thread safe"); | |
2235 | ||
44c7e1a7 EN |
2236 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
2237 | mem_pool_init(istate->ce_mem_pool, 0); | |
77ff1127 BP |
2238 | |
2239 | /* ensure we have no more threads than we have blocks to process */ | |
2240 | if (nr_threads > ieot->nr) | |
2241 | nr_threads = ieot->nr; | |
ca56dadb | 2242 | CALLOC_ARRAY(data, nr_threads); |
77ff1127 BP |
2243 | |
2244 | offset = ieot_start = 0; | |
2245 | ieot_blocks = DIV_ROUND_UP(ieot->nr, nr_threads); | |
2246 | for (i = 0; i < nr_threads; i++) { | |
2247 | struct load_cache_entries_thread_data *p = &data[i]; | |
2248 | int nr, j; | |
2249 | ||
2250 | if (ieot_start + ieot_blocks > ieot->nr) | |
2251 | ieot_blocks = ieot->nr - ieot_start; | |
2252 | ||
2253 | p->istate = istate; | |
2254 | p->offset = offset; | |
2255 | p->mmap = mmap; | |
2256 | p->ieot = ieot; | |
2257 | p->ieot_start = ieot_start; | |
2258 | p->ieot_blocks = ieot_blocks; | |
2259 | ||
2260 | /* create a mem_pool for each thread */ | |
2261 | nr = 0; | |
2262 | for (j = p->ieot_start; j < p->ieot_start + p->ieot_blocks; j++) | |
2263 | nr += p->ieot->entries[j].nr; | |
bcd2c5ee | 2264 | p->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2265 | if (istate->version == 4) { |
44c7e1a7 | 2266 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2267 | estimate_cache_size_from_compressed(nr)); |
2268 | } else { | |
44c7e1a7 | 2269 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2270 | estimate_cache_size(mmap_size, nr)); |
2271 | } | |
2272 | ||
2273 | err = pthread_create(&p->pthread, NULL, load_cache_entries_thread, p); | |
2274 | if (err) | |
2275 | die(_("unable to create load_cache_entries thread: %s"), strerror(err)); | |
2276 | ||
2277 | /* increment by the number of cache entries in the ieot block being processed */ | |
2278 | for (j = 0; j < ieot_blocks; j++) | |
2279 | offset += ieot->entries[ieot_start + j].nr; | |
2280 | ieot_start += ieot_blocks; | |
2281 | } | |
2282 | ||
2283 | for (i = 0; i < nr_threads; i++) { | |
2284 | struct load_cache_entries_thread_data *p = &data[i]; | |
2285 | ||
2286 | err = pthread_join(p->pthread, NULL); | |
2287 | if (err) | |
2288 | die(_("unable to join load_cache_entries thread: %s"), strerror(err)); | |
2289 | mem_pool_combine(istate->ce_mem_pool, p->ce_mem_pool); | |
2290 | consumed += p->consumed; | |
2291 | } | |
2292 | ||
2293 | free(data); | |
2294 | ||
2295 | return consumed; | |
2296 | } | |
77ff1127 | 2297 | |
491df5f6 VD |
2298 | static void set_new_index_sparsity(struct index_state *istate) |
2299 | { | |
2300 | /* | |
2301 | * If the index's repo exists, mark it sparse according to | |
2302 | * repo settings. | |
2303 | */ | |
5bdf6d4a ÆAB |
2304 | prepare_repo_settings(istate->repo); |
2305 | if (!istate->repo->settings.command_requires_full_index && | |
2306 | is_sparse_index_allowed(istate, 0)) | |
2307 | istate->sparse_index = 1; | |
491df5f6 VD |
2308 | } |
2309 | ||
8fd2cb40 | 2310 | /* remember to discard_cache() before reading a different cache! */ |
3e52f70b | 2311 | int do_read_index(struct index_state *istate, const char *path, int must_exist) |
e83c5163 | 2312 | { |
77ff1127 | 2313 | int fd; |
e83c5163 | 2314 | struct stat st; |
debed2a6 | 2315 | unsigned long src_offset; |
371ed0de BP |
2316 | const struct cache_header *hdr; |
2317 | const char *mmap; | |
7a51ed66 | 2318 | size_t mmap_size; |
abb4bb83 BP |
2319 | struct load_index_extensions p; |
2320 | size_t extension_offset = 0; | |
77ff1127 BP |
2321 | int nr_threads, cpus; |
2322 | struct index_entry_offset_table *ieot = NULL; | |
e83c5163 | 2323 | |
913e0e99 | 2324 | if (istate->initialized) |
4aab5b46 | 2325 | return istate->cache_nr; |
5d1a5c02 | 2326 | |
fba2f38a KB |
2327 | istate->timestamp.sec = 0; |
2328 | istate->timestamp.nsec = 0; | |
8fd2cb40 | 2329 | fd = open(path, O_RDONLY); |
5d1a5c02 | 2330 | if (fd < 0) { |
491df5f6 VD |
2331 | if (!must_exist && errno == ENOENT) { |
2332 | set_new_index_sparsity(istate); | |
5d1a5c02 | 2333 | return 0; |
491df5f6 | 2334 | } |
9d0a9e90 | 2335 | die_errno(_("%s: index file open failed"), path); |
5d1a5c02 | 2336 | } |
e83c5163 | 2337 | |
3511a377 | 2338 | if (fstat(fd, &st)) |
9d0a9e90 | 2339 | die_errno(_("%s: cannot stat the open index"), path); |
3511a377 | 2340 | |
7a51ed66 | 2341 | mmap_size = xsize_t(st.st_size); |
aab61359 | 2342 | if (mmap_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
9d0a9e90 | 2343 | die(_("%s: index file smaller than expected"), path); |
3511a377 | 2344 | |
02638d1e | 2345 | mmap = xmmap_gently(NULL, mmap_size, PROT_READ, MAP_PRIVATE, fd, 0); |
7a51ed66 | 2346 | if (mmap == MAP_FAILED) |
dc059294 EW |
2347 | die_errno(_("%s: unable to map index file%s"), path, |
2348 | mmap_os_err()); | |
57d84f8d | 2349 | close(fd); |
e83c5163 | 2350 | |
371ed0de | 2351 | hdr = (const struct cache_header *)mmap; |
7a51ed66 | 2352 | if (verify_hdr(hdr, mmap_size) < 0) |
e83c5163 LT |
2353 | goto unmap; |
2354 | ||
92e2cab9 | 2355 | oidread(&istate->oid, (const unsigned char *)hdr + mmap_size - the_hash_algo->rawsz); |
9d227781 | 2356 | istate->version = ntohl(hdr->hdr_version); |
4aab5b46 JH |
2357 | istate->cache_nr = ntohl(hdr->hdr_entries); |
2358 | istate->cache_alloc = alloc_nr(istate->cache_nr); | |
ca56dadb | 2359 | CALLOC_ARRAY(istate->cache, istate->cache_alloc); |
913e0e99 | 2360 | istate->initialized = 1; |
7a51ed66 | 2361 | |
abb4bb83 BP |
2362 | p.istate = istate; |
2363 | p.mmap = mmap; | |
2364 | p.mmap_size = mmap_size; | |
6c9cd161 | 2365 | |
7a51ed66 | 2366 | src_offset = sizeof(*hdr); |
4aab5b46 | 2367 | |
2a9dedef JN |
2368 | if (git_config_get_index_threads(&nr_threads)) |
2369 | nr_threads = 1; | |
7a51ed66 | 2370 | |
77ff1127 BP |
2371 | /* TODO: does creating more threads than cores help? */ |
2372 | if (!nr_threads) { | |
2373 | nr_threads = istate->cache_nr / THREAD_COST; | |
2374 | cpus = online_cpus(); | |
2375 | if (nr_threads > cpus) | |
2376 | nr_threads = cpus; | |
e83c5163 | 2377 | } |
abb4bb83 | 2378 | |
88168b9b NTND |
2379 | if (!HAVE_THREADS) |
2380 | nr_threads = 1; | |
2381 | ||
abb4bb83 BP |
2382 | if (nr_threads > 1) { |
2383 | extension_offset = read_eoie_extension(mmap, mmap_size); | |
2384 | if (extension_offset) { | |
2385 | int err; | |
2386 | ||
2387 | p.src_offset = extension_offset; | |
2388 | err = pthread_create(&p.pthread, NULL, load_index_extensions, &p); | |
2389 | if (err) | |
2390 | die(_("unable to create load_index_extensions thread: %s"), strerror(err)); | |
2391 | ||
2392 | nr_threads--; | |
2393 | } | |
2394 | } | |
abb4bb83 | 2395 | |
77ff1127 BP |
2396 | /* |
2397 | * Locate and read the index entry offset table so that we can use it | |
2398 | * to multi-thread the reading of the cache entries. | |
2399 | */ | |
2400 | if (extension_offset && nr_threads > 1) | |
2401 | ieot = read_ieot_extension(mmap, mmap_size, extension_offset); | |
2402 | ||
2403 | if (ieot) { | |
7bd9631b | 2404 | src_offset += load_cache_entries_threaded(istate, mmap, mmap_size, nr_threads, ieot); |
77ff1127 | 2405 | free(ieot); |
8e72d675 | 2406 | } else { |
77ff1127 | 2407 | src_offset += load_all_cache_entries(istate, mmap, mmap_size, src_offset); |
8e72d675 | 2408 | } |
6c9cd161 | 2409 | |
fba2f38a | 2410 | istate->timestamp.sec = st.st_mtime; |
c06ff490 | 2411 | istate->timestamp.nsec = ST_MTIME_NSEC(st); |
fba2f38a | 2412 | |
abb4bb83 | 2413 | /* if we created a thread, join it otherwise load the extensions on the primary thread */ |
abb4bb83 BP |
2414 | if (extension_offset) { |
2415 | int ret = pthread_join(p.pthread, NULL); | |
2416 | if (ret) | |
2417 | die(_("unable to join load_index_extensions thread: %s"), strerror(ret)); | |
88168b9b | 2418 | } else { |
abb4bb83 BP |
2419 | p.src_offset = src_offset; |
2420 | load_index_extensions(&p); | |
bad68ec9 | 2421 | } |
371ed0de | 2422 | munmap((void *)mmap, mmap_size); |
42fee7a3 JH |
2423 | |
2424 | /* | |
2425 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2426 | * that is associated with the given "istate". | |
2427 | */ | |
2428 | trace2_data_intmax("index", the_repository, "read/version", | |
2429 | istate->version); | |
2430 | trace2_data_intmax("index", the_repository, "read/cache_nr", | |
2431 | istate->cache_nr); | |
2432 | ||
7ca4fc88 VD |
2433 | /* |
2434 | * If the command explicitly requires a full index, force it | |
2435 | * to be full. Otherwise, correct the sparsity based on repository | |
2436 | * settings and other properties of the index (if necessary). | |
2437 | */ | |
4300f844 DS |
2438 | prepare_repo_settings(istate->repo); |
2439 | if (istate->repo->settings.command_requires_full_index) | |
2440 | ensure_full_index(istate); | |
7ca4fc88 VD |
2441 | else |
2442 | ensure_correct_sparsity(istate); | |
4300f844 | 2443 | |
4aab5b46 | 2444 | return istate->cache_nr; |
e83c5163 LT |
2445 | |
2446 | unmap: | |
371ed0de | 2447 | munmap((void *)mmap, mmap_size); |
9d0a9e90 | 2448 | die(_("index file corrupt")); |
e83c5163 LT |
2449 | } |
2450 | ||
0d59ffb4 CC |
2451 | /* |
2452 | * Signal that the shared index is used by updating its mtime. | |
2453 | * | |
2454 | * This way, shared index can be removed if they have not been used | |
2455 | * for some time. | |
2456 | */ | |
a125a223 | 2457 | static void freshen_shared_index(const char *shared_index, int warn) |
0d59ffb4 | 2458 | { |
0d59ffb4 | 2459 | if (!check_and_freshen_file(shared_index, 1) && warn) |
9d0a9e90 | 2460 | warning(_("could not freshen shared index '%s'"), shared_index); |
0d59ffb4 CC |
2461 | } |
2462 | ||
a125a223 TG |
2463 | int read_index_from(struct index_state *istate, const char *path, |
2464 | const char *gitdir) | |
5fc2fc8f NTND |
2465 | { |
2466 | struct split_index *split_index; | |
2467 | int ret; | |
2182abd9 | 2468 | char *base_oid_hex; |
a125a223 | 2469 | char *base_path; |
5fc2fc8f NTND |
2470 | |
2471 | /* istate->initialized covers both .git/index and .git/sharedindex.xxx */ | |
2472 | if (istate->initialized) | |
2473 | return istate->cache_nr; | |
2474 | ||
42fee7a3 JH |
2475 | /* |
2476 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2477 | * that is associated with the given "istate". | |
2478 | */ | |
2479 | trace2_region_enter_printf("index", "do_read_index", the_repository, | |
2480 | "%s", path); | |
c46c406a | 2481 | trace_performance_enter(); |
5fc2fc8f | 2482 | ret = do_read_index(istate, path, 0); |
c46c406a | 2483 | trace_performance_leave("read cache %s", path); |
42fee7a3 JH |
2484 | trace2_region_leave_printf("index", "do_read_index", the_repository, |
2485 | "%s", path); | |
435ec090 | 2486 | |
5fc2fc8f | 2487 | split_index = istate->split_index; |
2182abd9 | 2488 | if (!split_index || is_null_oid(&split_index->base_oid)) { |
435ec090 | 2489 | post_read_index_from(istate); |
5fc2fc8f | 2490 | return ret; |
03f15a79 | 2491 | } |
5fc2fc8f | 2492 | |
c46c406a | 2493 | trace_performance_enter(); |
5fc2fc8f | 2494 | if (split_index->base) |
2f6b1eb7 | 2495 | release_index(split_index->base); |
5fc2fc8f | 2496 | else |
2f6b1eb7 | 2497 | ALLOC_ARRAY(split_index->base, 1); |
6269f8ea | 2498 | index_state_init(split_index->base, istate->repo); |
de6ae5f9 | 2499 | |
2182abd9 | 2500 | base_oid_hex = oid_to_hex(&split_index->base_oid); |
2501 | base_path = xstrfmt("%s/sharedindex.%s", gitdir, base_oid_hex); | |
42fee7a3 JH |
2502 | trace2_region_enter_printf("index", "shared/do_read_index", |
2503 | the_repository, "%s", base_path); | |
998330ac | 2504 | ret = do_read_index(split_index->base, base_path, 0); |
42fee7a3 JH |
2505 | trace2_region_leave_printf("index", "shared/do_read_index", |
2506 | the_repository, "%s", base_path); | |
998330ac SG |
2507 | if (!ret) { |
2508 | char *path_copy = xstrdup(path); | |
652891de JS |
2509 | char *base_path2 = xstrfmt("%s/sharedindex.%s", |
2510 | dirname(path_copy), base_oid_hex); | |
998330ac SG |
2511 | free(path_copy); |
2512 | trace2_region_enter_printf("index", "shared/do_read_index", | |
2513 | the_repository, "%s", base_path2); | |
2514 | ret = do_read_index(split_index->base, base_path2, 1); | |
2515 | trace2_region_leave_printf("index", "shared/do_read_index", | |
2516 | the_repository, "%s", base_path2); | |
652891de | 2517 | free(base_path2); |
998330ac | 2518 | } |
9001dc2a | 2519 | if (!oideq(&split_index->base_oid, &split_index->base->oid)) |
9d0a9e90 | 2520 | die(_("broken index, expect %s in %s, got %s"), |
2182abd9 | 2521 | base_oid_hex, base_path, |
75691ea3 | 2522 | oid_to_hex(&split_index->base->oid)); |
de6ae5f9 | 2523 | |
a125a223 | 2524 | freshen_shared_index(base_path, 0); |
5fc2fc8f | 2525 | merge_base_index(istate); |
435ec090 | 2526 | post_read_index_from(istate); |
c46c406a | 2527 | trace_performance_leave("read cache %s", base_path); |
b42ad7d5 | 2528 | free(base_path); |
5fc2fc8f NTND |
2529 | return ret; |
2530 | } | |
2531 | ||
fa7b3c2f JH |
2532 | int is_index_unborn(struct index_state *istate) |
2533 | { | |
debed2a6 | 2534 | return (!istate->cache_nr && !istate->timestamp.sec); |
fa7b3c2f JH |
2535 | } |
2536 | ||
6269f8ea | 2537 | void index_state_init(struct index_state *istate, struct repository *r) |
2f6b1eb7 | 2538 | { |
6269f8ea | 2539 | struct index_state blank = INDEX_STATE_INIT(r); |
2f6b1eb7 ÆAB |
2540 | memcpy(istate, &blank, sizeof(*istate)); |
2541 | } | |
2542 | ||
2543 | void release_index(struct index_state *istate) | |
6d297f81 | 2544 | { |
8e72d675 JM |
2545 | /* |
2546 | * Cache entries in istate->cache[] should have been allocated | |
2547 | * from the memory pool associated with this index, or from an | |
2548 | * associated split_index. There is no need to free individual | |
8616a2d0 JM |
2549 | * cache entries. validate_cache_entries can detect when this |
2550 | * assertion does not hold. | |
8e72d675 | 2551 | */ |
8616a2d0 | 2552 | validate_cache_entries(istate); |
debed2a6 | 2553 | |
cfc5789a | 2554 | resolve_undo_clear_index(istate); |
2092678c | 2555 | free_name_hash(istate); |
4aab5b46 | 2556 | cache_tree_free(&(istate->cache_tree)); |
2f6b1eb7 ÆAB |
2557 | free(istate->fsmonitor_last_update); |
2558 | free(istate->cache); | |
5fc2fc8f | 2559 | discard_split_index(istate); |
f9e6c649 | 2560 | free_untracked_cache(istate->untracked); |
8e72d675 | 2561 | |
b5fcb1c0 ÆAB |
2562 | if (istate->sparse_checkout_patterns) { |
2563 | clear_pattern_list(istate->sparse_checkout_patterns); | |
2564 | FREE_AND_NULL(istate->sparse_checkout_patterns); | |
2565 | } | |
2566 | ||
8e72d675 | 2567 | if (istate->ce_mem_pool) { |
8616a2d0 | 2568 | mem_pool_discard(istate->ce_mem_pool, should_validate_cache_entries()); |
44c7e1a7 | 2569 | FREE_AND_NULL(istate->ce_mem_pool); |
8e72d675 | 2570 | } |
6d297f81 JS |
2571 | } |
2572 | ||
2f6b1eb7 ÆAB |
2573 | void discard_index(struct index_state *istate) |
2574 | { | |
2575 | release_index(istate); | |
6269f8ea | 2576 | index_state_init(istate, istate->repo); |
2f6b1eb7 ÆAB |
2577 | } |
2578 | ||
8616a2d0 JM |
2579 | /* |
2580 | * Validate the cache entries of this index. | |
2581 | * All cache entries associated with this index | |
2582 | * should have been allocated by the memory pool | |
2583 | * associated with this index, or by a referenced | |
2584 | * split index. | |
2585 | */ | |
2586 | void validate_cache_entries(const struct index_state *istate) | |
2587 | { | |
2588 | int i; | |
2589 | ||
2590 | if (!should_validate_cache_entries() ||!istate || !istate->initialized) | |
2591 | return; | |
2592 | ||
2593 | for (i = 0; i < istate->cache_nr; i++) { | |
2594 | if (!istate) { | |
391408e5 | 2595 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2596 | } else if (!istate->ce_mem_pool || |
2597 | !mem_pool_contains(istate->ce_mem_pool, istate->cache[i])) { | |
2598 | if (!istate->split_index || | |
2599 | !istate->split_index->base || | |
2600 | !istate->split_index->base->ce_mem_pool || | |
2601 | !mem_pool_contains(istate->split_index->base->ce_mem_pool, istate->cache[i])) { | |
391408e5 | 2602 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2603 | } |
2604 | } | |
2605 | } | |
2606 | ||
2607 | if (istate->split_index) | |
2608 | validate_cache_entries(istate->split_index->base); | |
2609 | } | |
2610 | ||
d1f128b0 | 2611 | int unmerged_index(const struct index_state *istate) |
94a5728c DB |
2612 | { |
2613 | int i; | |
2614 | for (i = 0; i < istate->cache_nr; i++) { | |
2615 | if (ce_stage(istate->cache[i])) | |
2616 | return 1; | |
2617 | } | |
2618 | return 0; | |
2619 | } | |
2620 | ||
150fe065 NTND |
2621 | int repo_index_has_changes(struct repository *repo, |
2622 | struct tree *tree, | |
2623 | struct strbuf *sb) | |
cffbfad5 | 2624 | { |
150fe065 | 2625 | struct index_state *istate = repo->index; |
e1f8694f | 2626 | struct object_id cmp; |
cffbfad5 EN |
2627 | int i; |
2628 | ||
e1f8694f EN |
2629 | if (tree) |
2630 | cmp = tree->object.oid; | |
d850b7a5 | 2631 | if (tree || !repo_get_oid_tree(the_repository, "HEAD", &cmp)) { |
cffbfad5 EN |
2632 | struct diff_options opt; |
2633 | ||
150fe065 | 2634 | repo_diff_setup(repo, &opt); |
cffbfad5 EN |
2635 | opt.flags.exit_with_status = 1; |
2636 | if (!sb) | |
2637 | opt.flags.quick = 1; | |
8d833e93 | 2638 | diff_setup_done(&opt); |
e1f8694f | 2639 | do_diff_cache(&cmp, &opt); |
cffbfad5 EN |
2640 | diffcore_std(&opt); |
2641 | for (i = 0; sb && i < diff_queued_diff.nr; i++) { | |
2642 | if (i) | |
2643 | strbuf_addch(sb, ' '); | |
2644 | strbuf_addstr(sb, diff_queued_diff.queue[i]->two->path); | |
2645 | } | |
2646 | diff_flush(&opt); | |
2647 | return opt.flags.has_changes != 0; | |
2648 | } else { | |
0c18c059 DS |
2649 | /* TODO: audit for interaction with sparse-index. */ |
2650 | ensure_full_index(istate); | |
1b9fbefb | 2651 | for (i = 0; sb && i < istate->cache_nr; i++) { |
cffbfad5 EN |
2652 | if (i) |
2653 | strbuf_addch(sb, ' '); | |
1b9fbefb | 2654 | strbuf_addstr(sb, istate->cache[i]->name); |
cffbfad5 | 2655 | } |
1b9fbefb | 2656 | return !!istate->cache_nr; |
cffbfad5 EN |
2657 | } |
2658 | } | |
2659 | ||
410334ed DS |
2660 | static int write_index_ext_header(struct hashfile *f, |
2661 | git_hash_ctx *eoie_f, | |
2662 | unsigned int ext, | |
2663 | unsigned int sz) | |
bad68ec9 | 2664 | { |
410334ed DS |
2665 | hashwrite_be32(f, ext); |
2666 | hashwrite_be32(f, sz); | |
2667 | ||
2668 | if (eoie_f) { | |
2669 | ext = htonl(ext); | |
2670 | sz = htonl(sz); | |
2671 | the_hash_algo->update_fn(eoie_f, &ext, sizeof(ext)); | |
2672 | the_hash_algo->update_fn(eoie_f, &sz, sizeof(sz)); | |
3b1d9e04 | 2673 | } |
410334ed | 2674 | return 0; |
bad68ec9 JH |
2675 | } |
2676 | ||
58bf2a4c NTND |
2677 | static void ce_smudge_racily_clean_entry(struct index_state *istate, |
2678 | struct cache_entry *ce) | |
407c8eb0 JH |
2679 | { |
2680 | /* | |
2681 | * The only thing we care about in this function is to smudge the | |
2682 | * falsely clean entry due to touch-update-touch race, so we leave | |
2683 | * everything else as they are. We are called for entries whose | |
c21d39d7 | 2684 | * ce_stat_data.sd_mtime match the index file mtime. |
c70115b4 JH |
2685 | * |
2686 | * Note that this actually does not do much for gitlinks, for | |
2687 | * which ce_match_stat_basic() always goes to the actual | |
2688 | * contents. The caller checks with is_racy_timestamp() which | |
2689 | * always says "no" for gitlinks, so we are not called for them ;-) | |
407c8eb0 JH |
2690 | */ |
2691 | struct stat st; | |
2692 | ||
2693 | if (lstat(ce->name, &st) < 0) | |
2694 | return; | |
2695 | if (ce_match_stat_basic(ce, &st)) | |
2696 | return; | |
58bf2a4c | 2697 | if (ce_modified_check_fs(istate, ce, &st)) { |
4b3511b0 JH |
2698 | /* This is "racily clean"; smudge it. Note that this |
2699 | * is a tricky code. At first glance, it may appear | |
2700 | * that it can break with this sequence: | |
2701 | * | |
2702 | * $ echo xyzzy >frotz | |
2703 | * $ git-update-index --add frotz | |
2704 | * $ : >frotz | |
2705 | * $ sleep 3 | |
2706 | * $ echo filfre >nitfol | |
2707 | * $ git-update-index --add nitfol | |
2708 | * | |
b7e58b17 | 2709 | * but it does not. When the second update-index runs, |
4b3511b0 JH |
2710 | * it notices that the entry "frotz" has the same timestamp |
2711 | * as index, and if we were to smudge it by resetting its | |
2712 | * size to zero here, then the object name recorded | |
2713 | * in index is the 6-byte file but the cached stat information | |
2714 | * becomes zero --- which would then match what we would | |
a6080a0a | 2715 | * obtain from the filesystem next time we stat("frotz"). |
4b3511b0 JH |
2716 | * |
2717 | * However, the second update-index, before calling | |
2718 | * this function, notices that the cached size is 6 | |
2719 | * bytes and what is on the filesystem is an empty | |
2720 | * file, and never calls us, so the cached size information | |
2721 | * for "frotz" stays 6 which does not match the filesystem. | |
2722 | */ | |
c21d39d7 | 2723 | ce->ce_stat_data.sd_size = 0; |
407c8eb0 JH |
2724 | } |
2725 | } | |
2726 | ||
f136f7bf | 2727 | /* Copy miscellaneous fields but not the name */ |
ce012deb | 2728 | static void copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk, |
f136f7bf | 2729 | struct cache_entry *ce) |
7a51ed66 | 2730 | { |
b60e188c | 2731 | short flags; |
575fa8a3 | 2732 | const unsigned hashsz = the_hash_algo->rawsz; |
2733 | uint16_t *flagsp = (uint16_t *)(ondisk->data + hashsz); | |
b60e188c | 2734 | |
c21d39d7 MH |
2735 | ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec); |
2736 | ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec); | |
2737 | ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec); | |
2738 | ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec); | |
2739 | ondisk->dev = htonl(ce->ce_stat_data.sd_dev); | |
2740 | ondisk->ino = htonl(ce->ce_stat_data.sd_ino); | |
7a51ed66 | 2741 | ondisk->mode = htonl(ce->ce_mode); |
c21d39d7 MH |
2742 | ondisk->uid = htonl(ce->ce_stat_data.sd_uid); |
2743 | ondisk->gid = htonl(ce->ce_stat_data.sd_gid); | |
2744 | ondisk->size = htonl(ce->ce_stat_data.sd_size); | |
575fa8a3 | 2745 | hashcpy(ondisk->data, ce->oid.hash); |
b60e188c | 2746 | |
ce51bf09 | 2747 | flags = ce->ce_flags & ~CE_NAMEMASK; |
b60e188c | 2748 | flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce)); |
575fa8a3 | 2749 | flagsp[0] = htons(flags); |
06aaaa0b | 2750 | if (ce->ce_flags & CE_EXTENDED) { |
575fa8a3 | 2751 | flagsp[1] = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16); |
f136f7bf JH |
2752 | } |
2753 | } | |
2754 | ||
410334ed | 2755 | static int ce_write_entry(struct hashfile *f, struct cache_entry *ce, |
ce012deb | 2756 | struct strbuf *previous_name, struct ondisk_cache_entry *ondisk) |
f136f7bf | 2757 | { |
9d227781 | 2758 | int size; |
00a4b035 RJ |
2759 | unsigned int saved_namelen; |
2760 | int stripped_name = 0; | |
ce012deb | 2761 | static unsigned char padding[8] = { 0x00 }; |
f136f7bf | 2762 | |
b3c96fb1 NTND |
2763 | if (ce->ce_flags & CE_STRIP_NAME) { |
2764 | saved_namelen = ce_namelen(ce); | |
2765 | ce->ce_namelen = 0; | |
00a4b035 | 2766 | stripped_name = 1; |
b3c96fb1 NTND |
2767 | } |
2768 | ||
575fa8a3 | 2769 | size = offsetof(struct ondisk_cache_entry,data) + ondisk_data_size(ce->ce_flags, 0); |
ce012deb | 2770 | |
9d227781 | 2771 | if (!previous_name) { |
ce012deb KW |
2772 | int len = ce_namelen(ce); |
2773 | copy_cache_entry_to_ondisk(ondisk, ce); | |
410334ed DS |
2774 | hashwrite(f, ondisk, size); |
2775 | hashwrite(f, ce->name, len); | |
2776 | hashwrite(f, padding, align_padding_size(size, len)); | |
9d227781 JH |
2777 | } else { |
2778 | int common, to_remove, prefix_size; | |
2779 | unsigned char to_remove_vi[16]; | |
2780 | for (common = 0; | |
2781 | (ce->name[common] && | |
2782 | common < previous_name->len && | |
2783 | ce->name[common] == previous_name->buf[common]); | |
2784 | common++) | |
2785 | ; /* still matching */ | |
2786 | to_remove = previous_name->len - common; | |
2787 | prefix_size = encode_varint(to_remove, to_remove_vi); | |
2788 | ||
ce012deb | 2789 | copy_cache_entry_to_ondisk(ondisk, ce); |
410334ed DS |
2790 | hashwrite(f, ondisk, size); |
2791 | hashwrite(f, to_remove_vi, prefix_size); | |
2792 | hashwrite(f, ce->name + common, ce_namelen(ce) - common); | |
2793 | hashwrite(f, padding, 1); | |
9d227781 JH |
2794 | |
2795 | strbuf_splice(previous_name, common, to_remove, | |
2796 | ce->name + common, ce_namelen(ce) - common); | |
06aaaa0b | 2797 | } |
00a4b035 | 2798 | if (stripped_name) { |
b3c96fb1 NTND |
2799 | ce->ce_namelen = saved_namelen; |
2800 | ce->ce_flags &= ~CE_STRIP_NAME; | |
2801 | } | |
7a51ed66 | 2802 | |
410334ed | 2803 | return 0; |
7a51ed66 LT |
2804 | } |
2805 | ||
426ddeea YM |
2806 | /* |
2807 | * This function verifies if index_state has the correct sha1 of the | |
2808 | * index file. Don't die if we have any other failure, just return 0. | |
2809 | */ | |
2810 | static int verify_index_from(const struct index_state *istate, const char *path) | |
2811 | { | |
2812 | int fd; | |
2813 | ssize_t n; | |
2814 | struct stat st; | |
aab61359 | 2815 | unsigned char hash[GIT_MAX_RAWSZ]; |
426ddeea YM |
2816 | |
2817 | if (!istate->initialized) | |
2818 | return 0; | |
2819 | ||
2820 | fd = open(path, O_RDONLY); | |
2821 | if (fd < 0) | |
2822 | return 0; | |
2823 | ||
2824 | if (fstat(fd, &st)) | |
2825 | goto out; | |
2826 | ||
aab61359 | 2827 | if (st.st_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
426ddeea YM |
2828 | goto out; |
2829 | ||
aab61359 | 2830 | n = pread_in_full(fd, hash, the_hash_algo->rawsz, st.st_size - the_hash_algo->rawsz); |
2831 | if (n != the_hash_algo->rawsz) | |
426ddeea YM |
2832 | goto out; |
2833 | ||
67947c34 | 2834 | if (!hasheq(istate->oid.hash, hash)) |
426ddeea YM |
2835 | goto out; |
2836 | ||
2837 | close(fd); | |
2838 | return 1; | |
2839 | ||
2840 | out: | |
2841 | close(fd); | |
2842 | return 0; | |
2843 | } | |
2844 | ||
1b0d968b | 2845 | static int repo_verify_index(struct repository *repo) |
426ddeea | 2846 | { |
1b0d968b | 2847 | return verify_index_from(repo->index, repo->index_file); |
426ddeea YM |
2848 | } |
2849 | ||
2ede073f | 2850 | int has_racy_timestamp(struct index_state *istate) |
483fbe2b JH |
2851 | { |
2852 | int entries = istate->cache_nr; | |
2853 | int i; | |
2854 | ||
2855 | for (i = 0; i < entries; i++) { | |
2856 | struct cache_entry *ce = istate->cache[i]; | |
2857 | if (is_racy_timestamp(istate, ce)) | |
2858 | return 1; | |
2859 | } | |
2860 | return 0; | |
2861 | } | |
2862 | ||
1b0d968b NTND |
2863 | void repo_update_index_if_able(struct repository *repo, |
2864 | struct lock_file *lockfile) | |
ccdc4ec3 | 2865 | { |
1b0d968b NTND |
2866 | if ((repo->index->cache_changed || |
2867 | has_racy_timestamp(repo->index)) && | |
2868 | repo_verify_index(repo)) | |
2869 | write_locked_index(repo->index, lockfile, COMMIT_LOCK); | |
b74c90fb MÅ |
2870 | else |
2871 | rollback_lock_file(lockfile); | |
ccdc4ec3 JH |
2872 | } |
2873 | ||
d8465500 JN |
2874 | static int record_eoie(void) |
2875 | { | |
2876 | int val; | |
2877 | ||
2878 | if (!git_config_get_bool("index.recordendofindexentries", &val)) | |
2879 | return val; | |
2a9dedef JN |
2880 | |
2881 | /* | |
2882 | * As a convenience, the end of index entries extension | |
2883 | * used for threading is written by default if the user | |
2884 | * explicitly requested threaded index reads. | |
2885 | */ | |
2886 | return !git_config_get_index_threads(&val) && val != 1; | |
d8465500 JN |
2887 | } |
2888 | ||
42916054 JN |
2889 | static int record_ieot(void) |
2890 | { | |
2891 | int val; | |
2892 | ||
2893 | if (!git_config_get_bool("index.recordoffsettable", &val)) | |
2894 | return val; | |
2a9dedef JN |
2895 | |
2896 | /* | |
2897 | * As a convenience, the offset table used for threading is | |
2898 | * written by default if the user explicitly requested | |
2899 | * threaded index reads. | |
2900 | */ | |
2901 | return !git_config_get_index_threads(&val) && val != 1; | |
42916054 JN |
2902 | } |
2903 | ||
812d6b00 MÅ |
2904 | /* |
2905 | * On success, `tempfile` is closed. If it is the temporary file | |
2906 | * of a `struct lock_file`, we will therefore effectively perform | |
2907 | * a 'close_lock_file_gently()`. Since that is an implementation | |
2908 | * detail of lockfiles, callers of `do_write_index()` should not | |
2909 | * rely on it. | |
2910 | */ | |
9f41c7a6 | 2911 | static int do_write_index(struct index_state *istate, struct tempfile *tempfile, |
ba95e96d | 2912 | int strip_extensions, unsigned flags) |
197ee8c9 | 2913 | { |
ca54d9ba | 2914 | uint64_t start = getnanotime(); |
410334ed DS |
2915 | struct hashfile *f; |
2916 | git_hash_ctx *eoie_c = NULL; | |
197ee8c9 | 2917 | struct cache_header hdr; |
b50386c7 | 2918 | int i, err = 0, removed, extended, hdr_version; |
4aab5b46 JH |
2919 | struct cache_entry **cache = istate->cache; |
2920 | int entries = istate->cache_nr; | |
e1afca4f | 2921 | struct stat st; |
575fa8a3 | 2922 | struct ondisk_cache_entry ondisk; |
9d227781 | 2923 | struct strbuf previous_name_buf = STRBUF_INIT, *previous_name; |
4bddd983 | 2924 | int drop_cache_tree = istate->drop_cache_tree; |
3b1d9e04 | 2925 | off_t offset; |
ba95e96d | 2926 | int csum_fsync_flag; |
77ff1127 | 2927 | int ieot_entries = 1; |
3255089a BP |
2928 | struct index_entry_offset_table *ieot = NULL; |
2929 | int nr, nr_threads; | |
6269f8ea | 2930 | struct repository *r = istate->repo; |
025a0709 | 2931 | |
410334ed DS |
2932 | f = hashfd(tempfile->fd, tempfile->filename.buf); |
2933 | ||
17194b19 DS |
2934 | prepare_repo_settings(r); |
2935 | f->skip_hash = r->settings.index_skip_hash; | |
ee1f0c24 | 2936 | |
06aaaa0b | 2937 | for (i = removed = extended = 0; i < entries; i++) { |
7a51ed66 | 2938 | if (cache[i]->ce_flags & CE_REMOVE) |
025a0709 | 2939 | removed++; |
197ee8c9 | 2940 | |
06aaaa0b NTND |
2941 | /* reduce extended entries if possible */ |
2942 | cache[i]->ce_flags &= ~CE_EXTENDED; | |
2943 | if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) { | |
2944 | extended++; | |
2945 | cache[i]->ce_flags |= CE_EXTENDED; | |
2946 | } | |
2947 | } | |
2948 | ||
e8ffd034 | 2949 | if (!istate->version) |
7211b9e7 | 2950 | istate->version = get_index_format_default(the_repository); |
9d227781 JH |
2951 | |
2952 | /* demote version 3 to version 2 when the latter suffices */ | |
2953 | if (istate->version == 3 || istate->version == 2) | |
2954 | istate->version = extended ? 3 : 2; | |
2955 | ||
2956 | hdr_version = istate->version; | |
2957 | ||
ccc4feb5 | 2958 | hdr.hdr_signature = htonl(CACHE_SIGNATURE); |
9d227781 | 2959 | hdr.hdr_version = htonl(hdr_version); |
025a0709 | 2960 | hdr.hdr_entries = htonl(entries - removed); |
197ee8c9 | 2961 | |
410334ed | 2962 | hashwrite(f, &hdr, sizeof(hdr)); |
197ee8c9 | 2963 | |
2a9dedef | 2964 | if (!HAVE_THREADS || git_config_get_index_threads(&nr_threads)) |
88168b9b | 2965 | nr_threads = 1; |
62e5ee81 | 2966 | |
42916054 | 2967 | if (nr_threads != 1 && record_ieot()) { |
3255089a BP |
2968 | int ieot_blocks, cpus; |
2969 | ||
2970 | /* | |
2971 | * ensure default number of ieot blocks maps evenly to the | |
2972 | * default number of threads that will process them leaving | |
2973 | * room for the thread to load the index extensions. | |
2974 | */ | |
2975 | if (!nr_threads) { | |
2976 | ieot_blocks = istate->cache_nr / THREAD_COST; | |
2977 | cpus = online_cpus(); | |
2978 | if (ieot_blocks > cpus - 1) | |
2979 | ieot_blocks = cpus - 1; | |
2980 | } else { | |
2981 | ieot_blocks = nr_threads; | |
77ff1127 BP |
2982 | if (ieot_blocks > istate->cache_nr) |
2983 | ieot_blocks = istate->cache_nr; | |
3255089a BP |
2984 | } |
2985 | ||
2986 | /* | |
2987 | * no reason to write out the IEOT extension if we don't | |
2988 | * have enough blocks to utilize multi-threading | |
2989 | */ | |
2990 | if (ieot_blocks > 1) { | |
2991 | ieot = xcalloc(1, sizeof(struct index_entry_offset_table) | |
2992 | + (ieot_blocks * sizeof(struct index_entry_offset))); | |
77ff1127 | 2993 | ieot_entries = DIV_ROUND_UP(entries, ieot_blocks); |
3255089a BP |
2994 | } |
2995 | } | |
3255089a | 2996 | |
410334ed DS |
2997 | offset = hashfile_total(f); |
2998 | ||
3255089a | 2999 | nr = 0; |
9d227781 | 3000 | previous_name = (hdr_version == 4) ? &previous_name_buf : NULL; |
ce012deb | 3001 | |
197ee8c9 LT |
3002 | for (i = 0; i < entries; i++) { |
3003 | struct cache_entry *ce = cache[i]; | |
7a51ed66 | 3004 | if (ce->ce_flags & CE_REMOVE) |
aa16021e | 3005 | continue; |
e06c43c7 | 3006 | if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce)) |
58bf2a4c | 3007 | ce_smudge_racily_clean_entry(istate, ce); |
99d1a986 | 3008 | if (is_null_oid(&ce->oid)) { |
83bd7437 JK |
3009 | static const char msg[] = "cache entry has null sha1: %s"; |
3010 | static int allow = -1; | |
3011 | ||
3012 | if (allow < 0) | |
3013 | allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0); | |
3014 | if (allow) | |
3015 | warning(msg, ce->name); | |
3016 | else | |
b50386c7 | 3017 | err = error(msg, ce->name); |
a96d3cc3 JK |
3018 | |
3019 | drop_cache_tree = 1; | |
83bd7437 | 3020 | } |
77ff1127 | 3021 | if (ieot && i && (i % ieot_entries == 0)) { |
3255089a BP |
3022 | ieot->entries[ieot->nr].nr = nr; |
3023 | ieot->entries[ieot->nr].offset = offset; | |
3024 | ieot->nr++; | |
3025 | /* | |
3026 | * If we have a V4 index, set the first byte to an invalid | |
3027 | * character to ensure there is nothing common with the previous | |
3028 | * entry | |
3029 | */ | |
3030 | if (previous_name) | |
3031 | previous_name->buf[0] = 0; | |
3032 | nr = 0; | |
410334ed DS |
3033 | |
3034 | offset = hashfile_total(f); | |
3255089a | 3035 | } |
410334ed | 3036 | if (ce_write_entry(f, ce, previous_name, (struct ondisk_cache_entry *)&ondisk) < 0) |
b50386c7 KW |
3037 | err = -1; |
3038 | ||
3039 | if (err) | |
3040 | break; | |
3255089a BP |
3041 | nr++; |
3042 | } | |
3043 | if (ieot && nr) { | |
3044 | ieot->entries[ieot->nr].nr = nr; | |
3045 | ieot->entries[ieot->nr].offset = offset; | |
3046 | ieot->nr++; | |
197ee8c9 | 3047 | } |
9d227781 | 3048 | strbuf_release(&previous_name_buf); |
1af1c2b6 | 3049 | |
3255089a BP |
3050 | if (err) { |
3051 | free(ieot); | |
b50386c7 | 3052 | return err; |
3255089a | 3053 | } |
b50386c7 | 3054 | |
410334ed DS |
3055 | offset = hashfile_total(f); |
3056 | ||
3057 | /* | |
3058 | * The extension headers must be hashed on their own for the | |
3059 | * EOIE extension. Create a hashfile here to compute that hash. | |
3060 | */ | |
3061 | if (offset && record_eoie()) { | |
3062 | CALLOC_ARRAY(eoie_c, 1); | |
3063 | the_hash_algo->init_fn(eoie_c); | |
3255089a | 3064 | } |
3b1d9e04 | 3065 | |
3255089a BP |
3066 | /* |
3067 | * Lets write out CACHE_EXT_INDEXENTRYOFFSETTABLE first so that we | |
3068 | * can minimize the number of extensions we have to scan through to | |
3069 | * find it during load. Write it out regardless of the | |
3070 | * strip_extensions parameter as we need it when loading the shared | |
3071 | * index. | |
3072 | */ | |
3255089a BP |
3073 | if (ieot) { |
3074 | struct strbuf sb = STRBUF_INIT; | |
3075 | ||
3076 | write_ieot_extension(&sb, ieot); | |
410334ed DS |
3077 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_INDEXENTRYOFFSETTABLE, sb.len) < 0; |
3078 | hashwrite(f, sb.buf, sb.len); | |
3255089a BP |
3079 | strbuf_release(&sb); |
3080 | free(ieot); | |
3081 | if (err) | |
3082 | return -1; | |
3083 | } | |
3255089a | 3084 | |
6e37c8ed NTND |
3085 | if (!strip_extensions && istate->split_index && |
3086 | !is_null_oid(&istate->split_index->base_oid)) { | |
5fc2fc8f NTND |
3087 | struct strbuf sb = STRBUF_INIT; |
3088 | ||
451b66c5 JS |
3089 | if (istate->sparse_index) |
3090 | die(_("cannot write split index for a sparse index")); | |
3091 | ||
5fc2fc8f | 3092 | err = write_link_extension(&sb, istate) < 0 || |
410334ed DS |
3093 | write_index_ext_header(f, eoie_c, CACHE_EXT_LINK, |
3094 | sb.len) < 0; | |
3095 | hashwrite(f, sb.buf, sb.len); | |
5fc2fc8f NTND |
3096 | strbuf_release(&sb); |
3097 | if (err) | |
3098 | return -1; | |
3099 | } | |
a96d3cc3 | 3100 | if (!strip_extensions && !drop_cache_tree && istate->cache_tree) { |
f285a2d7 | 3101 | struct strbuf sb = STRBUF_INIT; |
1dffb8fa | 3102 | |
1dffb8fa | 3103 | cache_tree_write(&sb, istate->cache_tree); |
410334ed DS |
3104 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_TREE, sb.len) < 0; |
3105 | hashwrite(f, sb.buf, sb.len); | |
1dffb8fa PH |
3106 | strbuf_release(&sb); |
3107 | if (err) | |
bad68ec9 | 3108 | return -1; |
bad68ec9 | 3109 | } |
c18b80a0 | 3110 | if (!strip_extensions && istate->resolve_undo) { |
cfc5789a JH |
3111 | struct strbuf sb = STRBUF_INIT; |
3112 | ||
3113 | resolve_undo_write(&sb, istate->resolve_undo); | |
410334ed DS |
3114 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_RESOLVE_UNDO, |
3115 | sb.len) < 0; | |
3116 | hashwrite(f, sb.buf, sb.len); | |
cfc5789a JH |
3117 | strbuf_release(&sb); |
3118 | if (err) | |
3119 | return -1; | |
3120 | } | |
83c094ad NTND |
3121 | if (!strip_extensions && istate->untracked) { |
3122 | struct strbuf sb = STRBUF_INIT; | |
3123 | ||
3124 | write_untracked_extension(&sb, istate->untracked); | |
410334ed DS |
3125 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_UNTRACKED, |
3126 | sb.len) < 0; | |
3127 | hashwrite(f, sb.buf, sb.len); | |
83c094ad NTND |
3128 | strbuf_release(&sb); |
3129 | if (err) | |
3130 | return -1; | |
3131 | } | |
883e248b BP |
3132 | if (!strip_extensions && istate->fsmonitor_last_update) { |
3133 | struct strbuf sb = STRBUF_INIT; | |
3134 | ||
3135 | write_fsmonitor_extension(&sb, istate); | |
410334ed DS |
3136 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_FSMONITOR, sb.len) < 0; |
3137 | hashwrite(f, sb.buf, sb.len); | |
3b1d9e04 BP |
3138 | strbuf_release(&sb); |
3139 | if (err) | |
3140 | return -1; | |
3141 | } | |
cd42415f | 3142 | if (istate->sparse_index) { |
410334ed | 3143 | if (write_index_ext_header(f, eoie_c, CACHE_EXT_SPARSE_DIRECTORIES, 0) < 0) |
cd42415f DS |
3144 | return -1; |
3145 | } | |
3b1d9e04 BP |
3146 | |
3147 | /* | |
3148 | * CACHE_EXT_ENDOFINDEXENTRIES must be written as the last entry before the SHA1 | |
3149 | * so that it can be found and processed before all the index entries are | |
3150 | * read. Write it out regardless of the strip_extensions parameter as we need it | |
3151 | * when loading the shared index. | |
3152 | */ | |
410334ed | 3153 | if (eoie_c) { |
3b1d9e04 BP |
3154 | struct strbuf sb = STRBUF_INIT; |
3155 | ||
410334ed DS |
3156 | write_eoie_extension(&sb, eoie_c, offset); |
3157 | err = write_index_ext_header(f, NULL, CACHE_EXT_ENDOFINDEXENTRIES, sb.len) < 0; | |
3158 | hashwrite(f, sb.buf, sb.len); | |
883e248b BP |
3159 | strbuf_release(&sb); |
3160 | if (err) | |
3161 | return -1; | |
3162 | } | |
e1afca4f | 3163 | |
ba95e96d NS |
3164 | csum_fsync_flag = 0; |
3165 | if (!alternate_index_output && (flags & COMMIT_LOCK)) | |
3166 | csum_fsync_flag = CSUM_FSYNC; | |
3167 | ||
3168 | finalize_hashfile(f, istate->oid.hash, FSYNC_COMPONENT_INDEX, | |
3169 | CSUM_HASH_IN_STREAM | csum_fsync_flag); | |
3170 | ||
49bd0fc2 | 3171 | if (close_tempfile_gently(tempfile)) { |
6a8c89d0 | 3172 | error(_("could not close '%s'"), get_tempfile_path(tempfile)); |
49bd0fc2 JK |
3173 | return -1; |
3174 | } | |
6a8c89d0 | 3175 | if (stat(get_tempfile_path(tempfile), &st)) |
e1afca4f | 3176 | return -1; |
5bcf109c KB |
3177 | istate->timestamp.sec = (unsigned int)st.st_mtime; |
3178 | istate->timestamp.nsec = ST_MTIME_NSEC(st); | |
ca54d9ba | 3179 | trace_performance_since(start, "write index, changed mask = %x", istate->cache_changed); |
42fee7a3 JH |
3180 | |
3181 | /* | |
3182 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3183 | * that is associated with the given "istate". | |
3184 | */ | |
3185 | trace2_data_intmax("index", the_repository, "write/version", | |
3186 | istate->version); | |
3187 | trace2_data_intmax("index", the_repository, "write/cache_nr", | |
3188 | istate->cache_nr); | |
3189 | ||
e1afca4f | 3190 | return 0; |
197ee8c9 | 3191 | } |
e46bbcf6 | 3192 | |
626f35c8 NTND |
3193 | void set_alternate_index_output(const char *name) |
3194 | { | |
3195 | alternate_index_output = name; | |
3196 | } | |
3197 | ||
3198 | static int commit_locked_index(struct lock_file *lk) | |
3199 | { | |
751baced MH |
3200 | if (alternate_index_output) |
3201 | return commit_lock_file_to(lk, alternate_index_output); | |
3202 | else | |
626f35c8 | 3203 | return commit_lock_file(lk); |
626f35c8 NTND |
3204 | } |
3205 | ||
03b86647 NTND |
3206 | static int do_write_locked_index(struct index_state *istate, struct lock_file *lock, |
3207 | unsigned flags) | |
3208 | { | |
42fee7a3 | 3209 | int ret; |
9fadb373 | 3210 | int was_full = istate->sparse_index == INDEX_EXPANDED; |
6e773527 | 3211 | |
ce7a9f01 | 3212 | ret = convert_to_sparse(istate, 0); |
6e773527 DS |
3213 | |
3214 | if (ret) { | |
3215 | warning(_("failed to convert to a sparse-index")); | |
3216 | return ret; | |
3217 | } | |
42fee7a3 JH |
3218 | |
3219 | /* | |
3220 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3221 | * that is associated with the given "istate". | |
3222 | */ | |
3223 | trace2_region_enter_printf("index", "do_write_index", the_repository, | |
6a8c89d0 | 3224 | "%s", get_lock_file_path(lock)); |
ba95e96d | 3225 | ret = do_write_index(istate, lock->tempfile, 0, flags); |
42fee7a3 | 3226 | trace2_region_leave_printf("index", "do_write_index", the_repository, |
6a8c89d0 | 3227 | "%s", get_lock_file_path(lock)); |
42fee7a3 | 3228 | |
6e773527 DS |
3229 | if (was_full) |
3230 | ensure_full_index(istate); | |
3231 | ||
03b86647 NTND |
3232 | if (ret) |
3233 | return ret; | |
03b86647 | 3234 | if (flags & COMMIT_LOCK) |
1956ecd0 BP |
3235 | ret = commit_locked_index(lock); |
3236 | else | |
3237 | ret = close_lock_file_gently(lock); | |
3238 | ||
dbb1c613 | 3239 | run_hooks_l("post-index-change", |
1956ecd0 BP |
3240 | istate->updated_workdir ? "1" : "0", |
3241 | istate->updated_skipworktree ? "1" : "0", NULL); | |
3242 | istate->updated_workdir = 0; | |
3243 | istate->updated_skipworktree = 0; | |
3244 | ||
3245 | return ret; | |
03b86647 NTND |
3246 | } |
3247 | ||
5fc2fc8f NTND |
3248 | static int write_split_index(struct index_state *istate, |
3249 | struct lock_file *lock, | |
3250 | unsigned flags) | |
3251 | { | |
3252 | int ret; | |
3253 | prepare_to_write_split_index(istate); | |
3254 | ret = do_write_locked_index(istate, lock, flags); | |
3255 | finish_writing_split_index(istate); | |
3256 | return ret; | |
3257 | } | |
3258 | ||
b9683722 CC |
3259 | static const char *shared_index_expire = "2.weeks.ago"; |
3260 | ||
3261 | static unsigned long get_shared_index_expire_date(void) | |
3262 | { | |
3263 | static unsigned long shared_index_expire_date; | |
3264 | static int shared_index_expire_date_prepared; | |
3265 | ||
3266 | if (!shared_index_expire_date_prepared) { | |
3267 | git_config_get_expiry("splitindex.sharedindexexpire", | |
3268 | &shared_index_expire); | |
3269 | shared_index_expire_date = approxidate(shared_index_expire); | |
3270 | shared_index_expire_date_prepared = 1; | |
3271 | } | |
3272 | ||
3273 | return shared_index_expire_date; | |
3274 | } | |
3275 | ||
3276 | static int should_delete_shared_index(const char *shared_index_path) | |
3277 | { | |
3278 | struct stat st; | |
3279 | unsigned long expiration; | |
3280 | ||
3281 | /* Check timestamp */ | |
3282 | expiration = get_shared_index_expire_date(); | |
3283 | if (!expiration) | |
3284 | return 0; | |
3285 | if (stat(shared_index_path, &st)) | |
78bde923 | 3286 | return error_errno(_("could not stat '%s'"), shared_index_path); |
b9683722 CC |
3287 | if (st.st_mtime > expiration) |
3288 | return 0; | |
3289 | ||
3290 | return 1; | |
3291 | } | |
3292 | ||
3293 | static int clean_shared_index_files(const char *current_hex) | |
3294 | { | |
3295 | struct dirent *de; | |
3296 | DIR *dir = opendir(get_git_dir()); | |
3297 | ||
3298 | if (!dir) | |
3299 | return error_errno(_("unable to open git dir: %s"), get_git_dir()); | |
3300 | ||
3301 | while ((de = readdir(dir)) != NULL) { | |
3302 | const char *sha1_hex; | |
3303 | const char *shared_index_path; | |
3304 | if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex)) | |
3305 | continue; | |
3306 | if (!strcmp(sha1_hex, current_hex)) | |
3307 | continue; | |
3308 | shared_index_path = git_path("%s", de->d_name); | |
3309 | if (should_delete_shared_index(shared_index_path) > 0 && | |
3310 | unlink(shared_index_path)) | |
3311 | warning_errno(_("unable to unlink: %s"), shared_index_path); | |
3312 | } | |
3313 | closedir(dir); | |
3314 | ||
3315 | return 0; | |
3316 | } | |
3317 | ||
a0a96756 | 3318 | static int write_shared_index(struct index_state *istate, |
ba95e96d | 3319 | struct tempfile **temp, unsigned flags) |
c18b80a0 NTND |
3320 | { |
3321 | struct split_index *si = istate->split_index; | |
6e773527 | 3322 | int ret, was_full = !istate->sparse_index; |
c18b80a0 | 3323 | |
c18b80a0 | 3324 | move_cache_to_base_index(istate); |
ce7a9f01 | 3325 | convert_to_sparse(istate, 0); |
42fee7a3 JH |
3326 | |
3327 | trace2_region_enter_printf("index", "shared/do_write_index", | |
6a8c89d0 | 3328 | the_repository, "%s", get_tempfile_path(*temp)); |
ba95e96d | 3329 | ret = do_write_index(si->base, *temp, 1, flags); |
c173542c | 3330 | trace2_region_leave_printf("index", "shared/do_write_index", |
6a8c89d0 | 3331 | the_repository, "%s", get_tempfile_path(*temp)); |
42fee7a3 | 3332 | |
6e773527 DS |
3333 | if (was_full) |
3334 | ensure_full_index(istate); | |
3335 | ||
59f9d2dd | 3336 | if (ret) |
c18b80a0 | 3337 | return ret; |
7db2d08c | 3338 | ret = adjust_shared_perm(get_tempfile_path(*temp)); |
df801f3f | 3339 | if (ret) { |
9d0a9e90 | 3340 | error(_("cannot fix permission bits on '%s'"), get_tempfile_path(*temp)); |
df801f3f CC |
3341 | return ret; |
3342 | } | |
7db2d08c | 3343 | ret = rename_tempfile(temp, |
75691ea3 | 3344 | git_path("sharedindex.%s", oid_to_hex(&si->base->oid))); |
b9683722 | 3345 | if (!ret) { |
75691ea3 | 3346 | oidcpy(&si->base_oid, &si->base->oid); |
3347 | clean_shared_index_files(oid_to_hex(&si->base->oid)); | |
b9683722 CC |
3348 | } |
3349 | ||
c18b80a0 NTND |
3350 | return ret; |
3351 | } | |
3352 | ||
e6a1dd77 CC |
3353 | static const int default_max_percent_split_change = 20; |
3354 | ||
3355 | static int too_many_not_shared_entries(struct index_state *istate) | |
3356 | { | |
3357 | int i, not_shared = 0; | |
3358 | int max_split = git_config_get_max_percent_split_change(); | |
3359 | ||
3360 | switch (max_split) { | |
3361 | case -1: | |
3362 | /* not or badly configured: use the default value */ | |
3363 | max_split = default_max_percent_split_change; | |
3364 | break; | |
3365 | case 0: | |
3366 | return 1; /* 0% means always write a new shared index */ | |
3367 | case 100: | |
3368 | return 0; /* 100% means never write a new shared index */ | |
3369 | default: | |
3370 | break; /* just use the configured value */ | |
3371 | } | |
3372 | ||
3373 | /* Count not shared entries */ | |
3374 | for (i = 0; i < istate->cache_nr; i++) { | |
3375 | struct cache_entry *ce = istate->cache[i]; | |
3376 | if (!ce->index) | |
3377 | not_shared++; | |
3378 | } | |
3379 | ||
3380 | return (int64_t)istate->cache_nr * max_split < (int64_t)not_shared * 100; | |
3381 | } | |
3382 | ||
03b86647 NTND |
3383 | int write_locked_index(struct index_state *istate, struct lock_file *lock, |
3384 | unsigned flags) | |
3385 | { | |
e8ffd034 | 3386 | int new_shared_index, ret, test_split_index_env; |
5fc2fc8f NTND |
3387 | struct split_index *si = istate->split_index; |
3388 | ||
4592e608 | 3389 | if (git_env_bool("GIT_TEST_CHECK_CACHE_TREE", 0)) |
c207e9e1 | 3390 | cache_tree_verify(the_repository, istate); |
4592e608 | 3391 | |
61000814 MÅ |
3392 | if ((flags & SKIP_IF_UNCHANGED) && !istate->cache_changed) { |
3393 | if (flags & COMMIT_LOCK) | |
3394 | rollback_lock_file(lock); | |
3395 | return 0; | |
3396 | } | |
3397 | ||
3bd28eb2 AV |
3398 | if (istate->fsmonitor_last_update) |
3399 | fill_fsmonitor_bitmap(istate); | |
3400 | ||
e8ffd034 SG |
3401 | test_split_index_env = git_env_bool("GIT_TEST_SPLIT_INDEX", 0); |
3402 | ||
3403 | if ((!si && !test_split_index_env) || | |
3404 | alternate_index_output || | |
5165dd59 | 3405 | (istate->cache_changed & ~EXTMASK)) { |
5fc2fc8f | 3406 | if (si) |
2182abd9 | 3407 | oidclr(&si->base_oid); |
df60cf57 MÅ |
3408 | ret = do_write_locked_index(istate, lock, flags); |
3409 | goto out; | |
5fc2fc8f NTND |
3410 | } |
3411 | ||
e8ffd034 SG |
3412 | if (test_split_index_env) { |
3413 | if (!si) { | |
3414 | si = init_split_index(istate); | |
d6e3c181 | 3415 | istate->cache_changed |= SPLIT_INDEX_ORDERED; |
e8ffd034 SG |
3416 | } else { |
3417 | int v = si->base_oid.hash[0]; | |
3418 | if ((v & 15) < 6) | |
3419 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
3420 | } | |
d6e3c181 | 3421 | } |
e6a1dd77 CC |
3422 | if (too_many_not_shared_entries(istate)) |
3423 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
0d59ffb4 CC |
3424 | |
3425 | new_shared_index = istate->cache_changed & SPLIT_INDEX_ORDERED; | |
3426 | ||
3427 | if (new_shared_index) { | |
59f9d2dd NTND |
3428 | struct tempfile *temp; |
3429 | int saved_errno; | |
3430 | ||
c9d6c788 ÆAB |
3431 | /* Same initial permissions as the main .git/index file */ |
3432 | temp = mks_tempfile_sm(git_path("sharedindex_XXXXXX"), 0, 0666); | |
59f9d2dd | 3433 | if (!temp) { |
2182abd9 | 3434 | oidclr(&si->base_oid); |
59f9d2dd | 3435 | ret = do_write_locked_index(istate, lock, flags); |
ef5b3a6c NTND |
3436 | goto out; |
3437 | } | |
ba95e96d | 3438 | ret = write_shared_index(istate, &temp, flags); |
59f9d2dd NTND |
3439 | |
3440 | saved_errno = errno; | |
3441 | if (is_tempfile_active(temp)) | |
3442 | delete_tempfile(&temp); | |
3443 | errno = saved_errno; | |
3444 | ||
c18b80a0 | 3445 | if (ret) |
df60cf57 | 3446 | goto out; |
c18b80a0 NTND |
3447 | } |
3448 | ||
0d59ffb4 CC |
3449 | ret = write_split_index(istate, lock, flags); |
3450 | ||
3451 | /* Freshen the shared index only if the split-index was written */ | |
6e37c8ed | 3452 | if (!ret && !new_shared_index && !is_null_oid(&si->base_oid)) { |
a125a223 | 3453 | const char *shared_index = git_path("sharedindex.%s", |
2182abd9 | 3454 | oid_to_hex(&si->base_oid)); |
a125a223 TG |
3455 | freshen_shared_index(shared_index, 1); |
3456 | } | |
0d59ffb4 | 3457 | |
df60cf57 MÅ |
3458 | out: |
3459 | if (flags & COMMIT_LOCK) | |
3460 | rollback_lock_file(lock); | |
0d59ffb4 | 3461 | return ret; |
03b86647 NTND |
3462 | } |
3463 | ||
e46bbcf6 MV |
3464 | /* |
3465 | * Read the index file that is potentially unmerged into given | |
ad376204 EN |
3466 | * index_state, dropping any unmerged entries to stage #0 (potentially |
3467 | * resulting in a path appearing as both a file and a directory in the | |
3468 | * index; the caller is responsible to clear out the extra entries | |
3469 | * before writing the index to a tree). Returns true if the index is | |
3470 | * unmerged. Callers who want to refuse to work from an unmerged | |
3471 | * state can call this and check its return value, instead of calling | |
3472 | * read_cache(). | |
e46bbcf6 | 3473 | */ |
e1ff0a32 | 3474 | int repo_read_index_unmerged(struct repository *repo) |
e46bbcf6 | 3475 | { |
e1ff0a32 | 3476 | struct index_state *istate; |
e46bbcf6 | 3477 | int i; |
d1a43f2a | 3478 | int unmerged = 0; |
e46bbcf6 | 3479 | |
e1ff0a32 NTND |
3480 | repo_read_index(repo); |
3481 | istate = repo->index; | |
e46bbcf6 MV |
3482 | for (i = 0; i < istate->cache_nr; i++) { |
3483 | struct cache_entry *ce = istate->cache[i]; | |
d1a43f2a | 3484 | struct cache_entry *new_ce; |
a849735b | 3485 | int len; |
d1a43f2a JH |
3486 | |
3487 | if (!ce_stage(ce)) | |
e46bbcf6 | 3488 | continue; |
d1a43f2a | 3489 | unmerged = 1; |
68c4f6a5 | 3490 | len = ce_namelen(ce); |
a849735b | 3491 | new_ce = make_empty_cache_entry(istate, len); |
d1a43f2a | 3492 | memcpy(new_ce->name, ce->name, len); |
b60e188c TG |
3493 | new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED; |
3494 | new_ce->ce_namelen = len; | |
d1a43f2a | 3495 | new_ce->ce_mode = ce->ce_mode; |
ad376204 | 3496 | if (add_index_entry(istate, new_ce, ADD_CACHE_SKIP_DFCHECK)) |
9d0a9e90 | 3497 | return error(_("%s: cannot drop to stage #0"), |
5699d17e | 3498 | new_ce->name); |
e46bbcf6 | 3499 | } |
d1a43f2a | 3500 | return unmerged; |
e46bbcf6 | 3501 | } |
041aee31 | 3502 | |
98fa4738 JK |
3503 | /* |
3504 | * Returns 1 if the path is an "other" path with respect to | |
3505 | * the index; that is, the path is not mentioned in the index at all, | |
3506 | * either as a file, a directory with some files in the index, | |
3507 | * or as an unmerged entry. | |
3508 | * | |
3509 | * We helpfully remove a trailing "/" from directories so that | |
3510 | * the output of read_directory can be used as-is. | |
3511 | */ | |
847a9e5d DS |
3512 | int index_name_is_other(struct index_state *istate, const char *name, |
3513 | int namelen) | |
98fa4738 JK |
3514 | { |
3515 | int pos; | |
3516 | if (namelen && name[namelen - 1] == '/') | |
3517 | namelen--; | |
3518 | pos = index_name_pos(istate, name, namelen); | |
3519 | if (0 <= pos) | |
3520 | return 0; /* exact match */ | |
3521 | pos = -pos - 1; | |
3522 | if (pos < istate->cache_nr) { | |
3523 | struct cache_entry *ce = istate->cache[pos]; | |
3524 | if (ce_namelen(ce) == namelen && | |
3525 | !memcmp(ce->name, name, namelen)) | |
3526 | return 0; /* Yup, this one exists unmerged */ | |
3527 | } | |
3528 | return 1; | |
3529 | } | |
29fb37b2 | 3530 | |
847a9e5d | 3531 | void *read_blob_data_from_index(struct index_state *istate, |
87542508 | 3532 | const char *path, unsigned long *size) |
29fb37b2 LF |
3533 | { |
3534 | int pos, len; | |
3535 | unsigned long sz; | |
3536 | enum object_type type; | |
3537 | void *data; | |
3538 | ||
3539 | len = strlen(path); | |
3540 | pos = index_name_pos(istate, path, len); | |
3541 | if (pos < 0) { | |
3542 | /* | |
3543 | * We might be in the middle of a merge, in which | |
3544 | * case we would read stage #2 (ours). | |
3545 | */ | |
3546 | int i; | |
3547 | for (i = -pos - 1; | |
3548 | (pos < 0 && i < istate->cache_nr && | |
3549 | !strcmp(istate->cache[i]->name, path)); | |
3550 | i++) | |
3551 | if (ce_stage(istate->cache[i]) == 2) | |
3552 | pos = i; | |
3553 | } | |
3554 | if (pos < 0) | |
3555 | return NULL; | |
b4f5aca4 | 3556 | data = read_object_file(&istate->cache[pos]->oid, &type, &sz); |
29fb37b2 LF |
3557 | if (!data || type != OBJ_BLOB) { |
3558 | free(data); | |
3559 | return NULL; | |
3560 | } | |
ff366825 LF |
3561 | if (size) |
3562 | *size = sz; | |
29fb37b2 LF |
3563 | return data; |
3564 | } | |
38612532 MH |
3565 | |
3566 | void stat_validity_clear(struct stat_validity *sv) | |
3567 | { | |
6a83d902 | 3568 | FREE_AND_NULL(sv->sd); |
38612532 MH |
3569 | } |
3570 | ||
3571 | int stat_validity_check(struct stat_validity *sv, const char *path) | |
3572 | { | |
3573 | struct stat st; | |
3574 | ||
3575 | if (stat(path, &st) < 0) | |
3576 | return sv->sd == NULL; | |
3577 | if (!sv->sd) | |
3578 | return 0; | |
3579 | return S_ISREG(st.st_mode) && !match_stat_data(sv->sd, &st); | |
3580 | } | |
3581 | ||
3582 | void stat_validity_update(struct stat_validity *sv, int fd) | |
3583 | { | |
3584 | struct stat st; | |
3585 | ||
3586 | if (fstat(fd, &st) < 0 || !S_ISREG(st.st_mode)) | |
3587 | stat_validity_clear(sv); | |
3588 | else { | |
3589 | if (!sv->sd) | |
ca56dadb | 3590 | CALLOC_ARRAY(sv->sd, 1); |
38612532 MH |
3591 | fill_stat_data(sv->sd, &st); |
3592 | } | |
3593 | } | |
edf3b905 DT |
3594 | |
3595 | void move_index_extensions(struct index_state *dst, struct index_state *src) | |
3596 | { | |
3597 | dst->untracked = src->untracked; | |
3598 | src->untracked = NULL; | |
836ef2b6 NTND |
3599 | dst->cache_tree = src->cache_tree; |
3600 | src->cache_tree = NULL; | |
edf3b905 | 3601 | } |
a849735b | 3602 | |
8e72d675 JM |
3603 | struct cache_entry *dup_cache_entry(const struct cache_entry *ce, |
3604 | struct index_state *istate) | |
3605 | { | |
3606 | unsigned int size = ce_size(ce); | |
3607 | int mem_pool_allocated; | |
3608 | struct cache_entry *new_entry = make_empty_cache_entry(istate, ce_namelen(ce)); | |
3609 | mem_pool_allocated = new_entry->mem_pool_allocated; | |
3610 | ||
3611 | memcpy(new_entry, ce, size); | |
3612 | new_entry->mem_pool_allocated = mem_pool_allocated; | |
3613 | return new_entry; | |
3614 | } | |
3615 | ||
a849735b JM |
3616 | void discard_cache_entry(struct cache_entry *ce) |
3617 | { | |
8616a2d0 JM |
3618 | if (ce && should_validate_cache_entries()) |
3619 | memset(ce, 0xCD, cache_entry_size(ce->ce_namelen)); | |
3620 | ||
8e72d675 JM |
3621 | if (ce && ce->mem_pool_allocated) |
3622 | return; | |
3623 | ||
a849735b JM |
3624 | free(ce); |
3625 | } | |
8616a2d0 JM |
3626 | |
3627 | int should_validate_cache_entries(void) | |
3628 | { | |
3629 | static int validate_index_cache_entries = -1; | |
3630 | ||
3631 | if (validate_index_cache_entries < 0) { | |
3632 | if (getenv("GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES")) | |
3633 | validate_index_cache_entries = 1; | |
3634 | else | |
3635 | validate_index_cache_entries = 0; | |
3636 | } | |
3637 | ||
3638 | return validate_index_cache_entries; | |
3639 | } | |
3b1d9e04 BP |
3640 | |
3641 | #define EOIE_SIZE (4 + GIT_SHA1_RAWSZ) /* <4-byte offset> + <20-byte hash> */ | |
3642 | #define EOIE_SIZE_WITH_HEADER (4 + 4 + EOIE_SIZE) /* <4-byte signature> + <4-byte length> + EOIE_SIZE */ | |
3643 | ||
3644 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size) | |
3645 | { | |
3646 | /* | |
3647 | * The end of index entries (EOIE) extension is guaranteed to be last | |
3648 | * so that it can be found by scanning backwards from the EOF. | |
3649 | * | |
3650 | * "EOIE" | |
3651 | * <4-byte length> | |
3652 | * <4-byte offset> | |
3653 | * <20-byte hash> | |
3654 | */ | |
3655 | const char *index, *eoie; | |
3656 | uint32_t extsize; | |
3657 | size_t offset, src_offset; | |
3658 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3659 | git_hash_ctx c; | |
3660 | ||
3661 | /* ensure we have an index big enough to contain an EOIE extension */ | |
3662 | if (mmap_size < sizeof(struct cache_header) + EOIE_SIZE_WITH_HEADER + the_hash_algo->rawsz) | |
3663 | return 0; | |
3664 | ||
3665 | /* validate the extension signature */ | |
3666 | index = eoie = mmap + mmap_size - EOIE_SIZE_WITH_HEADER - the_hash_algo->rawsz; | |
3667 | if (CACHE_EXT(index) != CACHE_EXT_ENDOFINDEXENTRIES) | |
3668 | return 0; | |
3669 | index += sizeof(uint32_t); | |
3670 | ||
3671 | /* validate the extension size */ | |
3672 | extsize = get_be32(index); | |
3673 | if (extsize != EOIE_SIZE) | |
3674 | return 0; | |
3675 | index += sizeof(uint32_t); | |
3676 | ||
3677 | /* | |
3678 | * Validate the offset we're going to look for the first extension | |
3679 | * signature is after the index header and before the eoie extension. | |
3680 | */ | |
3681 | offset = get_be32(index); | |
3682 | if (mmap + offset < mmap + sizeof(struct cache_header)) | |
3683 | return 0; | |
3684 | if (mmap + offset >= eoie) | |
3685 | return 0; | |
3686 | index += sizeof(uint32_t); | |
3687 | ||
3688 | /* | |
3689 | * The hash is computed over extension types and their sizes (but not | |
3690 | * their contents). E.g. if we have "TREE" extension that is N-bytes | |
3691 | * long, "REUC" extension that is M-bytes long, followed by "EOIE", | |
3692 | * then the hash would be: | |
3693 | * | |
3694 | * SHA-1("TREE" + <binary representation of N> + | |
3695 | * "REUC" + <binary representation of M>) | |
3696 | */ | |
3697 | src_offset = offset; | |
3698 | the_hash_algo->init_fn(&c); | |
3699 | while (src_offset < mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) { | |
3700 | /* After an array of active_nr index entries, | |
3701 | * there can be arbitrary number of extended | |
3702 | * sections, each of which is prefixed with | |
3703 | * extension name (4-byte) and section length | |
3704 | * in 4-byte network byte order. | |
3705 | */ | |
3706 | uint32_t extsize; | |
3707 | memcpy(&extsize, mmap + src_offset + 4, 4); | |
3708 | extsize = ntohl(extsize); | |
3709 | ||
3710 | /* verify the extension size isn't so large it will wrap around */ | |
3711 | if (src_offset + 8 + extsize < src_offset) | |
3712 | return 0; | |
3713 | ||
3714 | the_hash_algo->update_fn(&c, mmap + src_offset, 8); | |
3715 | ||
3716 | src_offset += 8; | |
3717 | src_offset += extsize; | |
3718 | } | |
3719 | the_hash_algo->final_fn(hash, &c); | |
3720 | if (!hasheq(hash, (const unsigned char *)index)) | |
3721 | return 0; | |
3722 | ||
3723 | /* Validate that the extension offsets returned us back to the eoie extension. */ | |
3724 | if (src_offset != mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) | |
3725 | return 0; | |
3726 | ||
3727 | return offset; | |
3728 | } | |
3729 | ||
3730 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset) | |
3731 | { | |
3732 | uint32_t buffer; | |
3733 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3734 | ||
3735 | /* offset */ | |
3736 | put_be32(&buffer, offset); | |
3737 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3738 | ||
3739 | /* hash */ | |
3740 | the_hash_algo->final_fn(hash, eoie_context); | |
3741 | strbuf_add(sb, hash, the_hash_algo->rawsz); | |
3742 | } | |
3255089a | 3743 | |
3255089a BP |
3744 | #define IEOT_VERSION (1) |
3745 | ||
3746 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset) | |
3747 | { | |
ec36c42a NTND |
3748 | const char *index = NULL; |
3749 | uint32_t extsize, ext_version; | |
3750 | struct index_entry_offset_table *ieot; | |
3751 | int i, nr; | |
3752 | ||
3753 | /* find the IEOT extension */ | |
3754 | if (!offset) | |
3755 | return NULL; | |
3756 | while (offset <= mmap_size - the_hash_algo->rawsz - 8) { | |
3757 | extsize = get_be32(mmap + offset + 4); | |
3758 | if (CACHE_EXT((mmap + offset)) == CACHE_EXT_INDEXENTRYOFFSETTABLE) { | |
3759 | index = mmap + offset + 4 + 4; | |
3760 | break; | |
3761 | } | |
3762 | offset += 8; | |
3763 | offset += extsize; | |
3764 | } | |
3765 | if (!index) | |
3766 | return NULL; | |
3767 | ||
3768 | /* validate the version is IEOT_VERSION */ | |
3769 | ext_version = get_be32(index); | |
3770 | if (ext_version != IEOT_VERSION) { | |
3771 | error("invalid IEOT version %d", ext_version); | |
3772 | return NULL; | |
3773 | } | |
3774 | index += sizeof(uint32_t); | |
3775 | ||
3776 | /* extension size - version bytes / bytes per entry */ | |
3777 | nr = (extsize - sizeof(uint32_t)) / (sizeof(uint32_t) + sizeof(uint32_t)); | |
3778 | if (!nr) { | |
3779 | error("invalid number of IEOT entries %d", nr); | |
3780 | return NULL; | |
3781 | } | |
3782 | ieot = xmalloc(sizeof(struct index_entry_offset_table) | |
3783 | + (nr * sizeof(struct index_entry_offset))); | |
3784 | ieot->nr = nr; | |
3785 | for (i = 0; i < nr; i++) { | |
3786 | ieot->entries[i].offset = get_be32(index); | |
3787 | index += sizeof(uint32_t); | |
3788 | ieot->entries[i].nr = get_be32(index); | |
3789 | index += sizeof(uint32_t); | |
3790 | } | |
3791 | ||
3792 | return ieot; | |
3255089a BP |
3793 | } |
3794 | ||
3795 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot) | |
3796 | { | |
ec36c42a NTND |
3797 | uint32_t buffer; |
3798 | int i; | |
3255089a | 3799 | |
ec36c42a NTND |
3800 | /* version */ |
3801 | put_be32(&buffer, IEOT_VERSION); | |
3802 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3803 | |
ec36c42a NTND |
3804 | /* ieot */ |
3805 | for (i = 0; i < ieot->nr; i++) { | |
3255089a | 3806 | |
ec36c42a NTND |
3807 | /* offset */ |
3808 | put_be32(&buffer, ieot->entries[i].offset); | |
3809 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3810 | |
ec36c42a NTND |
3811 | /* count */ |
3812 | put_be32(&buffer, ieot->entries[i].nr); | |
3813 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3814 | } | |
3255089a | 3815 | } |
b2896d27 JT |
3816 | |
3817 | void prefetch_cache_entries(const struct index_state *istate, | |
3818 | must_prefetch_predicate must_prefetch) | |
3819 | { | |
3820 | int i; | |
3821 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
3822 | ||
3823 | for (i = 0; i < istate->cache_nr; i++) { | |
3824 | struct cache_entry *ce = istate->cache[i]; | |
3825 | ||
3826 | if (S_ISGITLINK(ce->ce_mode) || !must_prefetch(ce)) | |
3827 | continue; | |
3828 | if (!oid_object_info_extended(the_repository, &ce->oid, | |
3829 | NULL, | |
3830 | OBJECT_INFO_FOR_PREFETCH)) | |
3831 | continue; | |
3832 | oid_array_append(&to_fetch, &ce->oid); | |
3833 | } | |
3834 | promisor_remote_get_direct(the_repository, | |
3835 | to_fetch.oid, to_fetch.nr); | |
3836 | oid_array_clear(&to_fetch); | |
3837 | } |