]>
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 DS |
114 | if (S_ISSPARSEDIR(ce->ce_mode)) |
115 | istate->sparse_index = 1; | |
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 | { | |
285c2e25 | 137 | struct cache_entry *old_entry = istate->cache[nr], *new_entry; |
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); |
285c2e25 | 150 | add_index_entry(istate, new_entry, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); |
81dc2307 PB |
151 | } |
152 | ||
c21d39d7 MH |
153 | void fill_stat_data(struct stat_data *sd, struct stat *st) |
154 | { | |
155 | sd->sd_ctime.sec = (unsigned int)st->st_ctime; | |
156 | sd->sd_mtime.sec = (unsigned int)st->st_mtime; | |
157 | sd->sd_ctime.nsec = ST_CTIME_NSEC(*st); | |
158 | sd->sd_mtime.nsec = ST_MTIME_NSEC(*st); | |
159 | sd->sd_dev = st->st_dev; | |
160 | sd->sd_ino = st->st_ino; | |
161 | sd->sd_uid = st->st_uid; | |
162 | sd->sd_gid = st->st_gid; | |
163 | sd->sd_size = st->st_size; | |
164 | } | |
165 | ||
166 | int match_stat_data(const struct stat_data *sd, struct stat *st) | |
167 | { | |
168 | int changed = 0; | |
169 | ||
170 | if (sd->sd_mtime.sec != (unsigned int)st->st_mtime) | |
171 | changed |= MTIME_CHANGED; | |
172 | if (trust_ctime && check_stat && | |
173 | sd->sd_ctime.sec != (unsigned int)st->st_ctime) | |
174 | changed |= CTIME_CHANGED; | |
175 | ||
176 | #ifdef USE_NSEC | |
177 | if (check_stat && sd->sd_mtime.nsec != ST_MTIME_NSEC(*st)) | |
178 | changed |= MTIME_CHANGED; | |
179 | if (trust_ctime && check_stat && | |
180 | sd->sd_ctime.nsec != ST_CTIME_NSEC(*st)) | |
181 | changed |= CTIME_CHANGED; | |
182 | #endif | |
183 | ||
184 | if (check_stat) { | |
185 | if (sd->sd_uid != (unsigned int) st->st_uid || | |
186 | sd->sd_gid != (unsigned int) st->st_gid) | |
187 | changed |= OWNER_CHANGED; | |
188 | if (sd->sd_ino != (unsigned int) st->st_ino) | |
189 | changed |= INODE_CHANGED; | |
190 | } | |
191 | ||
192 | #ifdef USE_STDEV | |
193 | /* | |
194 | * st_dev breaks on network filesystems where different | |
195 | * clients will have different views of what "device" | |
196 | * the filesystem is on | |
197 | */ | |
198 | if (check_stat && sd->sd_dev != (unsigned int) st->st_dev) | |
199 | changed |= INODE_CHANGED; | |
200 | #endif | |
201 | ||
202 | if (sd->sd_size != (unsigned int) st->st_size) | |
203 | changed |= DATA_CHANGED; | |
204 | ||
205 | return changed; | |
206 | } | |
207 | ||
415e96c8 JH |
208 | /* |
209 | * This only updates the "non-critical" parts of the directory | |
210 | * cache, ie the parts that aren't tracked by GIT, and only used | |
211 | * to validate the cache. | |
212 | */ | |
d4c0a3ac | 213 | void fill_stat_cache_info(struct index_state *istate, struct cache_entry *ce, struct stat *st) |
415e96c8 | 214 | { |
c21d39d7 | 215 | fill_stat_data(&ce->ce_stat_data, st); |
5f73076c JH |
216 | |
217 | if (assume_unchanged) | |
7a51ed66 | 218 | ce->ce_flags |= CE_VALID; |
eadb5831 | 219 | |
883e248b | 220 | if (S_ISREG(st->st_mode)) { |
eadb5831 | 221 | ce_mark_uptodate(ce); |
b5a81697 | 222 | mark_fsmonitor_valid(istate, ce); |
883e248b | 223 | } |
415e96c8 JH |
224 | } |
225 | ||
58bf2a4c NTND |
226 | static int ce_compare_data(struct index_state *istate, |
227 | const struct cache_entry *ce, | |
228 | struct stat *st) | |
29e4d363 JH |
229 | { |
230 | int match = -1; | |
1b8ac5ea | 231 | int fd = git_open_cloexec(ce->name, O_RDONLY); |
29e4d363 JH |
232 | |
233 | if (fd >= 0) { | |
bebfecb9 | 234 | struct object_id oid; |
58bf2a4c | 235 | if (!index_fd(istate, &oid, fd, st, OBJ_BLOB, ce->name, 0)) |
6a29d7b7 | 236 | match = !oideq(&oid, &ce->oid); |
7f8508e8 | 237 | /* index_fd() closed the file descriptor already */ |
29e4d363 JH |
238 | } |
239 | return match; | |
240 | } | |
241 | ||
21a6b9fa | 242 | static int ce_compare_link(const struct cache_entry *ce, size_t expected_size) |
29e4d363 JH |
243 | { |
244 | int match = -1; | |
29e4d363 JH |
245 | void *buffer; |
246 | unsigned long size; | |
21666f1a | 247 | enum object_type type; |
a60272b3 | 248 | struct strbuf sb = STRBUF_INIT; |
29e4d363 | 249 | |
a60272b3 | 250 | if (strbuf_readlink(&sb, ce->name, expected_size)) |
29e4d363 | 251 | return -1; |
a60272b3 | 252 | |
b4f5aca4 | 253 | buffer = read_object_file(&ce->oid, &type, &size); |
a60272b3 LT |
254 | if (buffer) { |
255 | if (size == sb.len) | |
256 | match = memcmp(buffer, sb.buf, size); | |
257 | free(buffer); | |
29e4d363 | 258 | } |
a60272b3 | 259 | strbuf_release(&sb); |
29e4d363 JH |
260 | return match; |
261 | } | |
262 | ||
21a6b9fa | 263 | static int ce_compare_gitlink(const struct cache_entry *ce) |
f35a6d3b | 264 | { |
1053fe82 | 265 | struct object_id oid; |
f35a6d3b LT |
266 | |
267 | /* | |
268 | * We don't actually require that the .git directory | |
302b9282 | 269 | * under GITLINK directory be a valid git directory. It |
f35a6d3b LT |
270 | * might even be missing (in case nobody populated that |
271 | * sub-project). | |
272 | * | |
273 | * If so, we consider it always to match. | |
274 | */ | |
a98e6101 | 275 | if (resolve_gitlink_ref(ce->name, "HEAD", &oid) < 0) |
f35a6d3b | 276 | return 0; |
6a29d7b7 | 277 | return !oideq(&oid, &ce->oid); |
f35a6d3b LT |
278 | } |
279 | ||
58bf2a4c NTND |
280 | static int ce_modified_check_fs(struct index_state *istate, |
281 | const struct cache_entry *ce, | |
282 | struct stat *st) | |
29e4d363 JH |
283 | { |
284 | switch (st->st_mode & S_IFMT) { | |
285 | case S_IFREG: | |
58bf2a4c | 286 | if (ce_compare_data(istate, ce, st)) |
29e4d363 JH |
287 | return DATA_CHANGED; |
288 | break; | |
289 | case S_IFLNK: | |
dc49cd76 | 290 | if (ce_compare_link(ce, xsize_t(st->st_size))) |
29e4d363 JH |
291 | return DATA_CHANGED; |
292 | break; | |
a8ee75bc | 293 | case S_IFDIR: |
7a51ed66 | 294 | if (S_ISGITLINK(ce->ce_mode)) |
c70115b4 | 295 | return ce_compare_gitlink(ce) ? DATA_CHANGED : 0; |
1cf01a34 | 296 | /* else fallthrough */ |
29e4d363 JH |
297 | default: |
298 | return TYPE_CHANGED; | |
299 | } | |
300 | return 0; | |
301 | } | |
302 | ||
21a6b9fa | 303 | static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st) |
734aab75 LT |
304 | { |
305 | unsigned int changed = 0; | |
306 | ||
7a51ed66 LT |
307 | if (ce->ce_flags & CE_REMOVE) |
308 | return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED; | |
309 | ||
310 | switch (ce->ce_mode & S_IFMT) { | |
8ae0a8c5 KS |
311 | case S_IFREG: |
312 | changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0; | |
3e09cdfd JH |
313 | /* We consider only the owner x bit to be relevant for |
314 | * "mode changes" | |
315 | */ | |
316 | if (trust_executable_bit && | |
7a51ed66 | 317 | (0100 & (ce->ce_mode ^ st->st_mode))) |
ffbe1add | 318 | changed |= MODE_CHANGED; |
8ae0a8c5 KS |
319 | break; |
320 | case S_IFLNK: | |
78a8d641 JS |
321 | if (!S_ISLNK(st->st_mode) && |
322 | (has_symlinks || !S_ISREG(st->st_mode))) | |
323 | changed |= TYPE_CHANGED; | |
8ae0a8c5 | 324 | break; |
302b9282 | 325 | case S_IFGITLINK: |
c70115b4 | 326 | /* We ignore most of the st_xxx fields for gitlinks */ |
f35a6d3b LT |
327 | if (!S_ISDIR(st->st_mode)) |
328 | changed |= TYPE_CHANGED; | |
329 | else if (ce_compare_gitlink(ce)) | |
330 | changed |= DATA_CHANGED; | |
a8ee75bc | 331 | return changed; |
8ae0a8c5 | 332 | default: |
391408e5 | 333 | BUG("unsupported ce_mode: %o", ce->ce_mode); |
8ae0a8c5 | 334 | } |
2cb45e95 | 335 | |
c21d39d7 | 336 | changed |= match_stat_data(&ce->ce_stat_data, st); |
b0391890 | 337 | |
f49c2c22 | 338 | /* Racily smudged entry? */ |
c21d39d7 | 339 | if (!ce->ce_stat_data.sd_size) { |
99d1a986 | 340 | if (!is_empty_blob_sha1(ce->oid.hash)) |
f49c2c22 LT |
341 | changed |= DATA_CHANGED; |
342 | } | |
343 | ||
407c8eb0 JH |
344 | return changed; |
345 | } | |
346 | ||
2bb4cda1 NTND |
347 | static int is_racy_stat(const struct index_state *istate, |
348 | const struct stat_data *sd) | |
6d91da6d | 349 | { |
2bb4cda1 | 350 | return (istate->timestamp.sec && |
fba2f38a KB |
351 | #ifdef USE_NSEC |
352 | /* nanosecond timestamped files can also be racy! */ | |
2bb4cda1 NTND |
353 | (istate->timestamp.sec < sd->sd_mtime.sec || |
354 | (istate->timestamp.sec == sd->sd_mtime.sec && | |
355 | istate->timestamp.nsec <= sd->sd_mtime.nsec)) | |
fba2f38a | 356 | #else |
2bb4cda1 | 357 | istate->timestamp.sec <= sd->sd_mtime.sec |
fba2f38a | 358 | #endif |
2bb4cda1 NTND |
359 | ); |
360 | } | |
361 | ||
5581a019 | 362 | int is_racy_timestamp(const struct index_state *istate, |
2bb4cda1 NTND |
363 | const struct cache_entry *ce) |
364 | { | |
365 | return (!S_ISGITLINK(ce->ce_mode) && | |
366 | is_racy_stat(istate, &ce->ce_stat_data)); | |
6d91da6d JH |
367 | } |
368 | ||
ed4efab1 NTND |
369 | int match_stat_data_racy(const struct index_state *istate, |
370 | const struct stat_data *sd, struct stat *st) | |
371 | { | |
372 | if (is_racy_stat(istate, sd)) | |
373 | return MTIME_CHANGED; | |
374 | return match_stat_data(sd, st); | |
6d91da6d JH |
375 | } |
376 | ||
883e248b | 377 | int ie_match_stat(struct index_state *istate, |
21a6b9fa | 378 | const struct cache_entry *ce, struct stat *st, |
4bd5b7da | 379 | unsigned int options) |
407c8eb0 | 380 | { |
5f73076c | 381 | unsigned int changed; |
4bd5b7da | 382 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 383 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
4bd5b7da | 384 | int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY; |
883e248b | 385 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
5f73076c | 386 | |
883e248b BP |
387 | if (!ignore_fsmonitor) |
388 | refresh_fsmonitor(istate); | |
5f73076c JH |
389 | /* |
390 | * If it's marked as always valid in the index, it's | |
391 | * valid whatever the checked-out copy says. | |
56cac48c NTND |
392 | * |
393 | * skip-worktree has the same effect with higher precedence | |
5f73076c | 394 | */ |
56cac48c NTND |
395 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) |
396 | return 0; | |
7a51ed66 | 397 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) |
5f73076c | 398 | return 0; |
883e248b BP |
399 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) |
400 | return 0; | |
5f73076c | 401 | |
331fcb59 JH |
402 | /* |
403 | * Intent-to-add entries have not been added, so the index entry | |
404 | * by definition never matches what is in the work tree until it | |
405 | * actually gets added. | |
406 | */ | |
895ff3b2 | 407 | if (ce_intent_to_add(ce)) |
331fcb59 JH |
408 | return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED; |
409 | ||
5f73076c | 410 | changed = ce_match_stat_basic(ce, st); |
407c8eb0 | 411 | |
29e4d363 JH |
412 | /* |
413 | * Within 1 second of this sequence: | |
414 | * echo xyzzy >file && git-update-index --add file | |
415 | * running this command: | |
416 | * echo frotz >file | |
417 | * would give a falsely clean cache entry. The mtime and | |
418 | * length match the cache, and other stat fields do not change. | |
419 | * | |
420 | * We could detect this at update-index time (the cache entry | |
421 | * being registered/updated records the same time as "now") | |
422 | * and delay the return from git-update-index, but that would | |
423 | * effectively mean we can make at most one commit per second, | |
424 | * which is not acceptable. Instead, we check cache entries | |
425 | * whose mtime are the same as the index file timestamp more | |
5f73076c | 426 | * carefully than others. |
29e4d363 | 427 | */ |
6d91da6d | 428 | if (!changed && is_racy_timestamp(istate, ce)) { |
42f77406 JH |
429 | if (assume_racy_is_modified) |
430 | changed |= DATA_CHANGED; | |
431 | else | |
58bf2a4c | 432 | changed |= ce_modified_check_fs(istate, ce, st); |
42f77406 | 433 | } |
b0391890 | 434 | |
29e4d363 | 435 | return changed; |
b0391890 JH |
436 | } |
437 | ||
883e248b | 438 | int ie_modified(struct index_state *istate, |
21a6b9fa RS |
439 | const struct cache_entry *ce, |
440 | struct stat *st, unsigned int options) | |
b0391890 | 441 | { |
29e4d363 | 442 | int changed, changed_fs; |
4bd5b7da JH |
443 | |
444 | changed = ie_match_stat(istate, ce, st, options); | |
b0391890 JH |
445 | if (!changed) |
446 | return 0; | |
b0391890 JH |
447 | /* |
448 | * If the mode or type has changed, there's no point in trying | |
449 | * to refresh the entry - it's not going to match | |
450 | */ | |
451 | if (changed & (MODE_CHANGED | TYPE_CHANGED)) | |
452 | return changed; | |
453 | ||
c70115b4 JH |
454 | /* |
455 | * Immediately after read-tree or update-index --cacheinfo, | |
456 | * the length field is zero, as we have never even read the | |
457 | * lstat(2) information once, and we cannot trust DATA_CHANGED | |
458 | * returned by ie_match_stat() which in turn was returned by | |
459 | * ce_match_stat_basic() to signal that the filesize of the | |
460 | * blob changed. We have to actually go to the filesystem to | |
461 | * see if the contents match, and if so, should answer "unchanged". | |
462 | * | |
463 | * The logic does not apply to gitlinks, as ce_match_stat_basic() | |
464 | * already has checked the actual HEAD from the filesystem in the | |
465 | * subproject. If ie_match_stat() already said it is different, | |
466 | * then we know it is. | |
b0391890 | 467 | */ |
c70115b4 | 468 | if ((changed & DATA_CHANGED) && |
c21d39d7 | 469 | (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0)) |
b0391890 JH |
470 | return changed; |
471 | ||
58bf2a4c | 472 | changed_fs = ce_modified_check_fs(istate, ce, st); |
29e4d363 JH |
473 | if (changed_fs) |
474 | return changed | changed_fs; | |
b0391890 JH |
475 | return 0; |
476 | } | |
477 | ||
958ba6c9 LT |
478 | int base_name_compare(const char *name1, int len1, int mode1, |
479 | const char *name2, int len2, int mode2) | |
480 | { | |
481 | unsigned char c1, c2; | |
482 | int len = len1 < len2 ? len1 : len2; | |
483 | int cmp; | |
484 | ||
485 | cmp = memcmp(name1, name2, len); | |
486 | if (cmp) | |
487 | return cmp; | |
488 | c1 = name1[len]; | |
489 | c2 = name2[len]; | |
1833a925 | 490 | if (!c1 && S_ISDIR(mode1)) |
958ba6c9 | 491 | c1 = '/'; |
1833a925 | 492 | if (!c2 && S_ISDIR(mode2)) |
958ba6c9 LT |
493 | c2 = '/'; |
494 | return (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0; | |
495 | } | |
496 | ||
0ab9e1e8 LT |
497 | /* |
498 | * df_name_compare() is identical to base_name_compare(), except it | |
499 | * compares conflicting directory/file entries as equal. Note that | |
500 | * while a directory name compares as equal to a regular file, they | |
501 | * then individually compare _differently_ to a filename that has | |
502 | * a dot after the basename (because '\0' < '.' < '/'). | |
503 | * | |
504 | * This is used by routines that want to traverse the git namespace | |
505 | * but then handle conflicting entries together when possible. | |
506 | */ | |
507 | int df_name_compare(const char *name1, int len1, int mode1, | |
508 | const char *name2, int len2, int mode2) | |
509 | { | |
510 | int len = len1 < len2 ? len1 : len2, cmp; | |
511 | unsigned char c1, c2; | |
512 | ||
513 | cmp = memcmp(name1, name2, len); | |
514 | if (cmp) | |
515 | return cmp; | |
516 | /* Directories and files compare equal (same length, same name) */ | |
517 | if (len1 == len2) | |
518 | return 0; | |
519 | c1 = name1[len]; | |
520 | if (!c1 && S_ISDIR(mode1)) | |
521 | c1 = '/'; | |
522 | c2 = name2[len]; | |
523 | if (!c2 && S_ISDIR(mode2)) | |
524 | c2 = '/'; | |
525 | if (c1 == '/' && !c2) | |
526 | return 0; | |
527 | if (c2 == '/' && !c1) | |
528 | return 0; | |
529 | return c1 - c2; | |
530 | } | |
531 | ||
ccdd4a0f | 532 | int name_compare(const char *name1, size_t len1, const char *name2, size_t len2) |
eb38c22f | 533 | { |
ccdd4a0f JM |
534 | size_t min_len = (len1 < len2) ? len1 : len2; |
535 | int cmp = memcmp(name1, name2, min_len); | |
eb38c22f LT |
536 | if (cmp) |
537 | return cmp; | |
538 | if (len1 < len2) | |
539 | return -1; | |
540 | if (len1 > len2) | |
541 | return 1; | |
ccdd4a0f JM |
542 | return 0; |
543 | } | |
544 | ||
545 | int cache_name_stage_compare(const char *name1, int len1, int stage1, const char *name2, int len2, int stage2) | |
546 | { | |
547 | int cmp; | |
548 | ||
549 | cmp = name_compare(name1, len1, name2, len2); | |
550 | if (cmp) | |
551 | return cmp; | |
5f73076c | 552 | |
b60e188c | 553 | if (stage1 < stage2) |
95fd5bf8 | 554 | return -1; |
b60e188c | 555 | if (stage1 > stage2) |
95fd5bf8 | 556 | return 1; |
eb38c22f LT |
557 | return 0; |
558 | } | |
559 | ||
20ec2d03 VD |
560 | static int index_name_stage_pos(struct index_state *istate, |
561 | const char *name, int namelen, | |
562 | int stage, | |
563 | enum index_search_mode search_mode) | |
eb38c22f LT |
564 | { |
565 | int first, last; | |
566 | ||
567 | first = 0; | |
4aab5b46 | 568 | last = istate->cache_nr; |
eb38c22f | 569 | while (last > first) { |
568a05c5 | 570 | int next = first + ((last - first) >> 1); |
4aab5b46 | 571 | struct cache_entry *ce = istate->cache[next]; |
b60e188c | 572 | int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce)); |
eb38c22f | 573 | if (!cmp) |
76e7f4ec | 574 | return next; |
eb38c22f LT |
575 | if (cmp < 0) { |
576 | last = next; | |
577 | continue; | |
578 | } | |
579 | first = next+1; | |
580 | } | |
95e0321c | 581 | |
20ec2d03 | 582 | if (search_mode == EXPAND_SPARSE && istate->sparse_index && |
95e0321c DS |
583 | first > 0) { |
584 | /* Note: first <= istate->cache_nr */ | |
585 | struct cache_entry *ce = istate->cache[first - 1]; | |
586 | ||
587 | /* | |
588 | * If we are in a sparse-index _and_ the entry before the | |
589 | * insertion position is a sparse-directory entry that is | |
590 | * an ancestor of 'name', then we need to expand the index | |
591 | * and search again. This will only trigger once, because | |
592 | * thereafter the index is fully expanded. | |
593 | */ | |
594 | if (S_ISSPARSEDIR(ce->ce_mode) && | |
595 | ce_namelen(ce) < namelen && | |
596 | !strncmp(name, ce->name, ce_namelen(ce))) { | |
597 | ensure_full_index(istate); | |
20ec2d03 | 598 | return index_name_stage_pos(istate, name, namelen, stage, search_mode); |
95e0321c DS |
599 | } |
600 | } | |
601 | ||
76e7f4ec | 602 | return -first-1; |
eb38c22f LT |
603 | } |
604 | ||
847a9e5d | 605 | int index_name_pos(struct index_state *istate, const char *name, int namelen) |
b60e188c | 606 | { |
20ec2d03 VD |
607 | return index_name_stage_pos(istate, name, namelen, 0, EXPAND_SPARSE); |
608 | } | |
609 | ||
610 | int index_entry_exists(struct index_state *istate, const char *name, int namelen) | |
611 | { | |
612 | return index_name_stage_pos(istate, name, namelen, 0, NO_EXPAND_SPARSE) >= 0; | |
b60e188c TG |
613 | } |
614 | ||
4aab5b46 | 615 | int remove_index_entry_at(struct index_state *istate, int pos) |
7b937ca3 | 616 | { |
cf558704 LT |
617 | struct cache_entry *ce = istate->cache[pos]; |
618 | ||
cfc5789a | 619 | record_resolve_undo(istate, ce); |
2092678c | 620 | remove_name_hash(istate, ce); |
045113a5 | 621 | save_or_free_index_entry(istate, ce); |
e636a7b4 | 622 | istate->cache_changed |= CE_ENTRY_REMOVED; |
4aab5b46 JH |
623 | istate->cache_nr--; |
624 | if (pos >= istate->cache_nr) | |
7b937ca3 | 625 | return 0; |
f331ab9d RS |
626 | MOVE_ARRAY(istate->cache + pos, istate->cache + pos + 1, |
627 | istate->cache_nr - pos); | |
7b937ca3 LT |
628 | return 1; |
629 | } | |
630 | ||
36419c8e | 631 | /* |
98e023de | 632 | * Remove all cache entries marked for removal, that is where |
36419c8e KB |
633 | * CE_REMOVE is set in ce_flags. This is much more effective than |
634 | * calling remove_index_entry_at() for each entry to be removed. | |
635 | */ | |
6fdc2057 | 636 | void remove_marked_cache_entries(struct index_state *istate, int invalidate) |
36419c8e KB |
637 | { |
638 | struct cache_entry **ce_array = istate->cache; | |
639 | unsigned int i, j; | |
640 | ||
641 | for (i = j = 0; i < istate->cache_nr; i++) { | |
5699d17e | 642 | if (ce_array[i]->ce_flags & CE_REMOVE) { |
6fdc2057 TG |
643 | if (invalidate) { |
644 | cache_tree_invalidate_path(istate, | |
645 | ce_array[i]->name); | |
646 | untracked_cache_remove_from_index(istate, | |
647 | ce_array[i]->name); | |
648 | } | |
2092678c | 649 | remove_name_hash(istate, ce_array[i]); |
045113a5 | 650 | save_or_free_index_entry(istate, ce_array[i]); |
5699d17e | 651 | } |
36419c8e KB |
652 | else |
653 | ce_array[j++] = ce_array[i]; | |
654 | } | |
ad837d9e NTND |
655 | if (j == istate->cache_nr) |
656 | return; | |
e636a7b4 | 657 | istate->cache_changed |= CE_ENTRY_REMOVED; |
36419c8e KB |
658 | istate->cache_nr = j; |
659 | } | |
660 | ||
4aab5b46 | 661 | int remove_file_from_index(struct index_state *istate, const char *path) |
197ee8c9 | 662 | { |
4aab5b46 | 663 | int pos = index_name_pos(istate, path, strlen(path)); |
c4e3cca1 JH |
664 | if (pos < 0) |
665 | pos = -pos-1; | |
a5400efe | 666 | cache_tree_invalidate_path(istate, path); |
e931371a | 667 | untracked_cache_remove_from_index(istate, path); |
4aab5b46 JH |
668 | while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path)) |
669 | remove_index_entry_at(istate, pos); | |
197ee8c9 LT |
670 | return 0; |
671 | } | |
672 | ||
20314271 JS |
673 | static int compare_name(struct cache_entry *ce, const char *path, int namelen) |
674 | { | |
675 | return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen); | |
676 | } | |
677 | ||
678 | static int index_name_pos_also_unmerged(struct index_state *istate, | |
679 | const char *path, int namelen) | |
680 | { | |
681 | int pos = index_name_pos(istate, path, namelen); | |
682 | struct cache_entry *ce; | |
683 | ||
684 | if (pos >= 0) | |
685 | return pos; | |
686 | ||
687 | /* maybe unmerged? */ | |
688 | pos = -1 - pos; | |
689 | if (pos >= istate->cache_nr || | |
690 | compare_name((ce = istate->cache[pos]), path, namelen)) | |
691 | return -1; | |
692 | ||
693 | /* order of preference: stage 2, 1, 3 */ | |
694 | if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr && | |
695 | ce_stage((ce = istate->cache[pos + 1])) == 2 && | |
696 | !compare_name(ce, path, namelen)) | |
697 | pos++; | |
698 | return pos; | |
699 | } | |
700 | ||
1102952b LT |
701 | static int different_name(struct cache_entry *ce, struct cache_entry *alias) |
702 | { | |
703 | int len = ce_namelen(ce); | |
704 | return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len); | |
705 | } | |
706 | ||
707 | /* | |
708 | * If we add a filename that aliases in the cache, we will use the | |
709 | * name that we already have - but we don't want to update the same | |
710 | * alias twice, because that implies that there were actually two | |
711 | * different files with aliasing names! | |
712 | * | |
713 | * So we use the CE_ADDED flag to verify that the alias was an old | |
714 | * one before we accept it as | |
715 | */ | |
045113a5 NTND |
716 | static struct cache_entry *create_alias_ce(struct index_state *istate, |
717 | struct cache_entry *ce, | |
718 | struct cache_entry *alias) | |
1102952b LT |
719 | { |
720 | int len; | |
285c2e25 | 721 | struct cache_entry *new_entry; |
1102952b LT |
722 | |
723 | if (alias->ce_flags & CE_ADDED) | |
9d0a9e90 NTND |
724 | die(_("will not add file alias '%s' ('%s' already exists in index)"), |
725 | ce->name, alias->name); | |
1102952b LT |
726 | |
727 | /* Ok, create the new entry using the name of the existing alias */ | |
728 | len = ce_namelen(alias); | |
a849735b | 729 | new_entry = make_empty_cache_entry(istate, len); |
285c2e25 BW |
730 | memcpy(new_entry->name, alias->name, len); |
731 | copy_cache_entry(new_entry, ce); | |
045113a5 | 732 | save_or_free_index_entry(istate, ce); |
285c2e25 | 733 | return new_entry; |
1102952b LT |
734 | } |
735 | ||
b4b313f9 | 736 | void set_object_name_for_intent_to_add_entry(struct cache_entry *ce) |
39425819 | 737 | { |
a09c985e | 738 | struct object_id oid; |
c80d226a | 739 | if (write_object_file("", 0, OBJ_BLOB, &oid)) |
9d0a9e90 | 740 | die(_("cannot create an empty blob in the object database")); |
a09c985e | 741 | oidcpy(&ce->oid, &oid); |
39425819 JH |
742 | } |
743 | ||
610d55af | 744 | int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags) |
11be42a4 | 745 | { |
a849735b | 746 | int namelen, was_same; |
d177cab0 | 747 | mode_t st_mode = st->st_mode; |
9472935d | 748 | struct cache_entry *ce, *alias = NULL; |
56cac48c | 749 | unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY; |
38ed1d89 JH |
750 | int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND); |
751 | int pretend = flags & ADD_CACHE_PRETEND; | |
39425819 JH |
752 | int intent_only = flags & ADD_CACHE_INTENT; |
753 | int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE| | |
754 | (intent_only ? ADD_CACHE_NEW_ONLY : 0)); | |
e578d031 | 755 | unsigned hash_flags = pretend ? 0 : HASH_WRITE_OBJECT; |
f937bc2f | 756 | struct object_id oid; |
9472935d | 757 | |
9e5da3d0 JK |
758 | if (flags & ADD_CACHE_RENORMALIZE) |
759 | hash_flags |= HASH_RENORMALIZE; | |
11be42a4 | 760 | |
d177cab0 | 761 | if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode)) |
9d0a9e90 | 762 | return error(_("%s: can only add regular files, symbolic links or git-directories"), path); |
11be42a4 JS |
763 | |
764 | namelen = strlen(path); | |
d177cab0 | 765 | if (S_ISDIR(st_mode)) { |
f937bc2f KM |
766 | if (resolve_gitlink_ref(path, "HEAD", &oid) < 0) |
767 | return error(_("'%s' does not have a commit checked out"), path); | |
09595258 LT |
768 | while (namelen && path[namelen-1] == '/') |
769 | namelen--; | |
770 | } | |
a849735b | 771 | ce = make_empty_cache_entry(istate, namelen); |
11be42a4 | 772 | memcpy(ce->name, path, namelen); |
b60e188c | 773 | ce->ce_namelen = namelen; |
39425819 | 774 | if (!intent_only) |
d4c0a3ac | 775 | fill_stat_cache_info(istate, ce, st); |
388b2acd JH |
776 | else |
777 | ce->ce_flags |= CE_INTENT_TO_ADD; | |
11be42a4 | 778 | |
610d55af TG |
779 | |
780 | if (trust_executable_bit && has_symlinks) { | |
d177cab0 | 781 | ce->ce_mode = create_ce_mode(st_mode); |
610d55af | 782 | } else { |
78a8d641 JS |
783 | /* If there is an existing entry, pick the mode bits and type |
784 | * from it, otherwise assume unexecutable regular file. | |
11be42a4 | 785 | */ |
185c975f | 786 | struct cache_entry *ent; |
20314271 | 787 | int pos = index_name_pos_also_unmerged(istate, path, namelen); |
185c975f | 788 | |
4aab5b46 | 789 | ent = (0 <= pos) ? istate->cache[pos] : NULL; |
d177cab0 | 790 | ce->ce_mode = ce_mode_from_stat(ent, st_mode); |
11be42a4 JS |
791 | } |
792 | ||
dc1ae704 JJ |
793 | /* When core.ignorecase=true, determine if a directory of the same name but differing |
794 | * case already exists within the Git repository. If it does, ensure the directory | |
795 | * case of the file being added to the repository matches (is folded into) the existing | |
796 | * entry's directory case. | |
797 | */ | |
798 | if (ignore_case) { | |
41284eb0 | 799 | adjust_dirname_case(istate, ce->name); |
dc1ae704 | 800 | } |
e2c2a375 | 801 | if (!(flags & ADD_CACHE_RENORMALIZE)) { |
9472935d TB |
802 | alias = index_file_exists(istate, ce->name, |
803 | ce_namelen(ce), ignore_case); | |
804 | if (alias && | |
805 | !ce_stage(alias) && | |
806 | !ie_match_stat(istate, alias, st, ce_option)) { | |
807 | /* Nothing changed, really */ | |
808 | if (!S_ISGITLINK(alias->ce_mode)) | |
809 | ce_mark_uptodate(alias); | |
810 | alias->ce_flags |= CE_ADDED; | |
dc1ae704 | 811 | |
a849735b | 812 | discard_cache_entry(ce); |
9472935d TB |
813 | return 0; |
814 | } | |
0781b8a9 | 815 | } |
39425819 | 816 | if (!intent_only) { |
1c418243 | 817 | if (index_path(istate, &ce->oid, path, st, hash_flags)) { |
a849735b | 818 | discard_cache_entry(ce); |
9d0a9e90 | 819 | return error(_("unable to index file '%s'"), path); |
2d9426b0 | 820 | } |
39425819 | 821 | } else |
b4b313f9 | 822 | set_object_name_for_intent_to_add_entry(ce); |
39425819 | 823 | |
1102952b | 824 | if (ignore_case && alias && different_name(ce, alias)) |
045113a5 | 825 | ce = create_alias_ce(istate, ce, alias); |
1102952b | 826 | ce->ce_flags |= CE_ADDED; |
38ed1d89 | 827 | |
3bf0dd1f | 828 | /* It was suspected to be racily clean, but it turns out to be Ok */ |
38ed1d89 JH |
829 | was_same = (alias && |
830 | !ce_stage(alias) && | |
4a7e27e9 | 831 | oideq(&alias->oid, &ce->oid) && |
38ed1d89 JH |
832 | ce->ce_mode == alias->ce_mode); |
833 | ||
834 | if (pretend) | |
a849735b | 835 | discard_cache_entry(ce); |
067178ed | 836 | else if (add_index_entry(istate, ce, add_option)) { |
a849735b | 837 | discard_cache_entry(ce); |
9d0a9e90 | 838 | return error(_("unable to add '%s' to index"), path); |
067178ed | 839 | } |
38ed1d89 | 840 | if (verbose && !was_same) |
11be42a4 | 841 | printf("add '%s'\n", path); |
11be42a4 JS |
842 | return 0; |
843 | } | |
844 | ||
610d55af | 845 | int add_file_to_index(struct index_state *istate, const char *path, int flags) |
d177cab0 LT |
846 | { |
847 | struct stat st; | |
848 | if (lstat(path, &st)) | |
9d0a9e90 | 849 | die_errno(_("unable to stat '%s'"), path); |
610d55af | 850 | return add_to_index(istate, path, &st, flags); |
d177cab0 LT |
851 | } |
852 | ||
a849735b JM |
853 | struct cache_entry *make_empty_cache_entry(struct index_state *istate, size_t len) |
854 | { | |
8e72d675 | 855 | return mem_pool__ce_calloc(find_mem_pool(istate), len); |
a849735b JM |
856 | } |
857 | ||
96168827 MT |
858 | struct cache_entry *make_empty_transient_cache_entry(size_t len, |
859 | struct mem_pool *ce_mem_pool) | |
a849735b | 860 | { |
96168827 MT |
861 | if (ce_mem_pool) |
862 | return mem_pool__ce_calloc(ce_mem_pool, len); | |
a849735b JM |
863 | return xcalloc(1, cache_entry_size(len)); |
864 | } | |
865 | ||
2a1ae649 RS |
866 | enum verify_path_result { |
867 | PATH_OK, | |
868 | PATH_INVALID, | |
869 | PATH_DIR_WITH_SEP, | |
870 | }; | |
871 | ||
872 | static enum verify_path_result verify_path_internal(const char *, unsigned); | |
873 | ||
874 | int verify_path(const char *path, unsigned mode) | |
875 | { | |
c8ad9d04 | 876 | return verify_path_internal(path, mode) == PATH_OK; |
2a1ae649 RS |
877 | } |
878 | ||
a849735b JM |
879 | struct cache_entry *make_cache_entry(struct index_state *istate, |
880 | unsigned int mode, | |
825ed4d9 JM |
881 | const struct object_id *oid, |
882 | const char *path, | |
883 | int stage, | |
884 | unsigned int refresh_options) | |
6640f881 | 885 | { |
bc1c2caa | 886 | struct cache_entry *ce, *ret; |
a849735b | 887 | int len; |
6640f881 | 888 | |
c8ad9d04 | 889 | if (verify_path_internal(path, mode) == PATH_INVALID) { |
9d0a9e90 | 890 | error(_("invalid path '%s'"), path); |
6640f881 | 891 | return NULL; |
7e7abea9 | 892 | } |
6640f881 CR |
893 | |
894 | len = strlen(path); | |
a849735b | 895 | ce = make_empty_cache_entry(istate, len); |
6640f881 | 896 | |
825ed4d9 | 897 | oidcpy(&ce->oid, oid); |
6640f881 | 898 | memcpy(ce->name, path, len); |
b60e188c TG |
899 | ce->ce_flags = create_ce_flags(stage); |
900 | ce->ce_namelen = len; | |
6640f881 CR |
901 | ce->ce_mode = create_ce_mode(mode); |
902 | ||
d7b665c3 | 903 | ret = refresh_cache_entry(istate, ce, refresh_options); |
915e44c6 | 904 | if (ret != ce) |
a849735b | 905 | discard_cache_entry(ce); |
915e44c6 | 906 | return ret; |
6640f881 CR |
907 | } |
908 | ||
96168827 MT |
909 | struct cache_entry *make_transient_cache_entry(unsigned int mode, |
910 | const struct object_id *oid, | |
911 | const char *path, | |
912 | int stage, | |
913 | struct mem_pool *ce_mem_pool) | |
a849735b JM |
914 | { |
915 | struct cache_entry *ce; | |
916 | int len; | |
917 | ||
918 | if (!verify_path(path, mode)) { | |
9d0a9e90 | 919 | error(_("invalid path '%s'"), path); |
a849735b JM |
920 | return NULL; |
921 | } | |
922 | ||
923 | len = strlen(path); | |
96168827 | 924 | ce = make_empty_transient_cache_entry(len, ce_mem_pool); |
a849735b JM |
925 | |
926 | oidcpy(&ce->oid, oid); | |
927 | memcpy(ce->name, path, len); | |
928 | ce->ce_flags = create_ce_flags(stage); | |
929 | ce->ce_namelen = len; | |
930 | ce->ce_mode = create_ce_mode(mode); | |
931 | ||
932 | return ce; | |
933 | } | |
934 | ||
d9d70966 TG |
935 | /* |
936 | * Chmod an index entry with either +x or -x. | |
937 | * | |
938 | * Returns -1 if the chmod for the particular cache entry failed (if it's | |
939 | * not a regular file), -2 if an invalid flip argument is passed in, 0 | |
940 | * otherwise. | |
941 | */ | |
942 | int chmod_index_entry(struct index_state *istate, struct cache_entry *ce, | |
943 | char flip) | |
944 | { | |
945 | if (!S_ISREG(ce->ce_mode)) | |
946 | return -1; | |
947 | switch (flip) { | |
948 | case '+': | |
949 | ce->ce_mode |= 0111; | |
950 | break; | |
951 | case '-': | |
952 | ce->ce_mode &= ~0111; | |
953 | break; | |
954 | default: | |
955 | return -2; | |
956 | } | |
957 | cache_tree_invalidate_path(istate, ce->name); | |
958 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
883e248b | 959 | mark_fsmonitor_invalid(istate, ce); |
d9d70966 TG |
960 | istate->cache_changed |= CE_ENTRY_CHANGED; |
961 | ||
962 | return 0; | |
963 | } | |
964 | ||
9c5e6c80 | 965 | int ce_same_name(const struct cache_entry *a, const struct cache_entry *b) |
7b937ca3 LT |
966 | { |
967 | int len = ce_namelen(a); | |
968 | return ce_namelen(b) == len && !memcmp(a->name, b->name, len); | |
969 | } | |
970 | ||
8dcf39c4 LT |
971 | /* |
972 | * We fundamentally don't like some paths: we don't want | |
973 | * dot or dot-dot anywhere, and for obvious reasons don't | |
974 | * want to recurse into ".git" either. | |
975 | * | |
976 | * Also, we don't want double slashes or slashes at the | |
977 | * end that can make pathnames ambiguous. | |
978 | */ | |
10ecfa76 | 979 | static int verify_dotfile(const char *rest, unsigned mode) |
8dcf39c4 LT |
980 | { |
981 | /* | |
982 | * The first character was '.', but that | |
983 | * has already been discarded, we now test | |
984 | * the rest. | |
985 | */ | |
e0f530ff | 986 | |
8dcf39c4 | 987 | /* "." is not allowed */ |
e0f530ff | 988 | if (*rest == '\0' || is_dir_sep(*rest)) |
8dcf39c4 LT |
989 | return 0; |
990 | ||
e0f530ff | 991 | switch (*rest) { |
8dcf39c4 | 992 | /* |
641084b6 JK |
993 | * ".git" followed by NUL or slash is bad. Note that we match |
994 | * case-insensitively here, even if ignore_case is not set. | |
995 | * This outlaws ".GIT" everywhere out of an abundance of caution, | |
996 | * since there's really no good reason to allow it. | |
10ecfa76 JK |
997 | * |
998 | * Once we've seen ".git", we can also find ".gitmodules", etc (also | |
999 | * case-insensitively). | |
8dcf39c4 LT |
1000 | */ |
1001 | case 'g': | |
cc2fc7c2 JK |
1002 | case 'G': |
1003 | if (rest[1] != 'i' && rest[1] != 'I') | |
8dcf39c4 | 1004 | break; |
cc2fc7c2 | 1005 | if (rest[2] != 't' && rest[2] != 'T') |
8dcf39c4 | 1006 | break; |
e19e5e66 JK |
1007 | if (rest[3] == '\0' || is_dir_sep(rest[3])) |
1008 | return 0; | |
10ecfa76 JK |
1009 | if (S_ISLNK(mode)) { |
1010 | rest += 3; | |
1011 | if (skip_iprefix(rest, "modules", &rest) && | |
1012 | (*rest == '\0' || is_dir_sep(*rest))) | |
1013 | return 0; | |
1014 | } | |
e19e5e66 | 1015 | break; |
8dcf39c4 | 1016 | case '.': |
e0f530ff | 1017 | if (rest[1] == '\0' || is_dir_sep(rest[1])) |
8dcf39c4 LT |
1018 | return 0; |
1019 | } | |
1020 | return 1; | |
1021 | } | |
1022 | ||
2a1ae649 RS |
1023 | static enum verify_path_result verify_path_internal(const char *path, |
1024 | unsigned mode) | |
8dcf39c4 | 1025 | { |
49e268e2 | 1026 | char c = 0; |
8dcf39c4 | 1027 | |
56948cb6 | 1028 | if (has_dos_drive_prefix(path)) |
2a1ae649 | 1029 | return PATH_INVALID; |
56948cb6 | 1030 | |
d2c84dad | 1031 | if (!is_valid_path(path)) |
2a1ae649 | 1032 | return PATH_INVALID; |
d2c84dad | 1033 | |
8dcf39c4 LT |
1034 | goto inside; |
1035 | for (;;) { | |
1036 | if (!c) | |
2a1ae649 | 1037 | return PATH_OK; |
56948cb6 | 1038 | if (is_dir_sep(c)) { |
8dcf39c4 | 1039 | inside: |
10ecfa76 | 1040 | if (protect_hfs) { |
49e268e2 | 1041 | |
10ecfa76 | 1042 | if (is_hfs_dotgit(path)) |
2a1ae649 | 1043 | return PATH_INVALID; |
10ecfa76 JK |
1044 | if (S_ISLNK(mode)) { |
1045 | if (is_hfs_dotgitmodules(path)) | |
2a1ae649 | 1046 | return PATH_INVALID; |
10ecfa76 JK |
1047 | } |
1048 | } | |
1049 | if (protect_ntfs) { | |
bccc37fd | 1050 | #if defined GIT_WINDOWS_NATIVE || defined __CYGWIN__ |
49e268e2 | 1051 | if (c == '\\') |
2a1ae649 | 1052 | return PATH_INVALID; |
49e268e2 | 1053 | #endif |
10ecfa76 | 1054 | if (is_ntfs_dotgit(path)) |
2a1ae649 | 1055 | return PATH_INVALID; |
10ecfa76 JK |
1056 | if (S_ISLNK(mode)) { |
1057 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 1058 | return PATH_INVALID; |
10ecfa76 JK |
1059 | } |
1060 | } | |
1061 | ||
8dcf39c4 | 1062 | c = *path++; |
10ecfa76 | 1063 | if ((c == '.' && !verify_dotfile(path, mode)) || |
6e773527 | 1064 | is_dir_sep(c)) |
2a1ae649 | 1065 | return PATH_INVALID; |
6e773527 DS |
1066 | /* |
1067 | * allow terminating directory separators for | |
1068 | * sparse directory entries. | |
1069 | */ | |
1070 | if (c == '\0') | |
2a1ae649 RS |
1071 | return S_ISDIR(mode) ? PATH_DIR_WITH_SEP : |
1072 | PATH_INVALID; | |
288a74bc JS |
1073 | } else if (c == '\\' && protect_ntfs) { |
1074 | if (is_ntfs_dotgit(path)) | |
2a1ae649 | 1075 | return PATH_INVALID; |
288a74bc JS |
1076 | if (S_ISLNK(mode)) { |
1077 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 1078 | return PATH_INVALID; |
288a74bc | 1079 | } |
8dcf39c4 | 1080 | } |
288a74bc | 1081 | |
8dcf39c4 LT |
1082 | c = *path++; |
1083 | } | |
1084 | } | |
1085 | ||
12676608 LT |
1086 | /* |
1087 | * Do we have another file that has the beginning components being a | |
1088 | * proper superset of the name we're trying to add? | |
0f1e4f04 | 1089 | */ |
4aab5b46 JH |
1090 | static int has_file_name(struct index_state *istate, |
1091 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
0f1e4f04 | 1092 | { |
12676608 LT |
1093 | int retval = 0; |
1094 | int len = ce_namelen(ce); | |
b155725d | 1095 | int stage = ce_stage(ce); |
12676608 | 1096 | const char *name = ce->name; |
0f1e4f04 | 1097 | |
4aab5b46 JH |
1098 | while (pos < istate->cache_nr) { |
1099 | struct cache_entry *p = istate->cache[pos++]; | |
0f1e4f04 | 1100 | |
12676608 | 1101 | if (len >= ce_namelen(p)) |
0f1e4f04 | 1102 | break; |
12676608 LT |
1103 | if (memcmp(name, p->name, len)) |
1104 | break; | |
b155725d JH |
1105 | if (ce_stage(p) != stage) |
1106 | continue; | |
12676608 LT |
1107 | if (p->name[len] != '/') |
1108 | continue; | |
7a51ed66 | 1109 | if (p->ce_flags & CE_REMOVE) |
21cd8d00 | 1110 | continue; |
12676608 LT |
1111 | retval = -1; |
1112 | if (!ok_to_replace) | |
1113 | break; | |
4aab5b46 | 1114 | remove_index_entry_at(istate, --pos); |
0f1e4f04 | 1115 | } |
12676608 LT |
1116 | return retval; |
1117 | } | |
0f1e4f04 | 1118 | |
a6db3fbb JH |
1119 | |
1120 | /* | |
1121 | * Like strcmp(), but also return the offset of the first change. | |
1122 | * If strings are equal, return the length. | |
1123 | */ | |
1124 | int strcmp_offset(const char *s1, const char *s2, size_t *first_change) | |
1125 | { | |
1126 | size_t k; | |
1127 | ||
1128 | if (!first_change) | |
1129 | return strcmp(s1, s2); | |
1130 | ||
1131 | for (k = 0; s1[k] == s2[k]; k++) | |
1132 | if (s1[k] == '\0') | |
1133 | break; | |
1134 | ||
1135 | *first_change = k; | |
1136 | return (unsigned char)s1[k] - (unsigned char)s2[k]; | |
1137 | } | |
1138 | ||
12676608 LT |
1139 | /* |
1140 | * Do we have another file with a pathname that is a proper | |
1141 | * subset of the name we're trying to add? | |
06b6d81b JH |
1142 | * |
1143 | * That is, is there another file in the index with a path | |
1144 | * that matches a sub-directory in the given entry? | |
12676608 | 1145 | */ |
4aab5b46 JH |
1146 | static int has_dir_name(struct index_state *istate, |
1147 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
12676608 LT |
1148 | { |
1149 | int retval = 0; | |
b155725d | 1150 | int stage = ce_stage(ce); |
12676608 LT |
1151 | const char *name = ce->name; |
1152 | const char *slash = name + ce_namelen(ce); | |
06b6d81b JH |
1153 | size_t len_eq_last; |
1154 | int cmp_last = 0; | |
1155 | ||
1156 | /* | |
1157 | * We are frequently called during an iteration on a sorted | |
1158 | * list of pathnames and while building a new index. Therefore, | |
1159 | * there is a high probability that this entry will eventually | |
1160 | * be appended to the index, rather than inserted in the middle. | |
1161 | * If we can confirm that, we can avoid binary searches on the | |
1162 | * components of the pathname. | |
1163 | * | |
1164 | * Compare the entry's full path with the last path in the index. | |
1165 | */ | |
1166 | if (istate->cache_nr > 0) { | |
1167 | cmp_last = strcmp_offset(name, | |
1168 | istate->cache[istate->cache_nr - 1]->name, | |
1169 | &len_eq_last); | |
1170 | if (cmp_last > 0) { | |
1171 | if (len_eq_last == 0) { | |
1172 | /* | |
1173 | * The entry sorts AFTER the last one in the | |
1174 | * index and their paths have no common prefix, | |
1175 | * so there cannot be a F/D conflict. | |
1176 | */ | |
1177 | return retval; | |
1178 | } else { | |
1179 | /* | |
1180 | * The entry sorts AFTER the last one in the | |
1181 | * index, but has a common prefix. Fall through | |
1182 | * to the loop below to disect the entry's path | |
1183 | * and see where the difference is. | |
1184 | */ | |
1185 | } | |
1186 | } else if (cmp_last == 0) { | |
1187 | /* | |
1188 | * The entry exactly matches the last one in the | |
1189 | * index, but because of multiple stage and CE_REMOVE | |
1190 | * items, we fall through and let the regular search | |
1191 | * code handle it. | |
1192 | */ | |
1193 | } | |
1194 | } | |
0f1e4f04 | 1195 | |
12676608 | 1196 | for (;;) { |
b986df5c | 1197 | size_t len; |
0f1e4f04 | 1198 | |
12676608 LT |
1199 | for (;;) { |
1200 | if (*--slash == '/') | |
1201 | break; | |
1202 | if (slash <= ce->name) | |
1203 | return retval; | |
1204 | } | |
1205 | len = slash - name; | |
0f1e4f04 | 1206 | |
b986df5c JH |
1207 | if (cmp_last > 0) { |
1208 | /* | |
1209 | * (len + 1) is a directory boundary (including | |
1210 | * the trailing slash). And since the loop is | |
1211 | * decrementing "slash", the first iteration is | |
1212 | * the longest directory prefix; subsequent | |
1213 | * iterations consider parent directories. | |
1214 | */ | |
1215 | ||
1216 | if (len + 1 <= len_eq_last) { | |
1217 | /* | |
1218 | * The directory prefix (including the trailing | |
1219 | * slash) also appears as a prefix in the last | |
1220 | * entry, so the remainder cannot collide (because | |
1221 | * strcmp said the whole path was greater). | |
1222 | * | |
1223 | * EQ: last: xxx/A | |
1224 | * this: xxx/B | |
1225 | * | |
1226 | * LT: last: xxx/file_A | |
1227 | * this: xxx/file_B | |
1228 | */ | |
1229 | return retval; | |
1230 | } | |
1231 | ||
1232 | if (len > len_eq_last) { | |
1233 | /* | |
1234 | * This part of the directory prefix (excluding | |
1235 | * the trailing slash) is longer than the known | |
1236 | * equal portions, so this sub-directory cannot | |
1237 | * collide with a file. | |
1238 | * | |
1239 | * GT: last: xxxA | |
1240 | * this: xxxB/file | |
1241 | */ | |
1242 | return retval; | |
1243 | } | |
1244 | ||
b986df5c JH |
1245 | /* |
1246 | * This is a possible collision. Fall through and | |
1247 | * let the regular search code handle it. | |
1248 | * | |
1249 | * last: xxx | |
1250 | * this: xxx/file | |
1251 | */ | |
1252 | } | |
1253 | ||
20ec2d03 | 1254 | pos = index_name_stage_pos(istate, name, len, stage, EXPAND_SPARSE); |
12676608 | 1255 | if (pos >= 0) { |
21cd8d00 JH |
1256 | /* |
1257 | * Found one, but not so fast. This could | |
1258 | * be a marker that says "I was here, but | |
1259 | * I am being removed". Such an entry is | |
1260 | * not a part of the resulting tree, and | |
1261 | * it is Ok to have a directory at the same | |
1262 | * path. | |
1263 | */ | |
077c48df | 1264 | if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) { |
21cd8d00 JH |
1265 | retval = -1; |
1266 | if (!ok_to_replace) | |
1267 | break; | |
4aab5b46 | 1268 | remove_index_entry_at(istate, pos); |
21cd8d00 JH |
1269 | continue; |
1270 | } | |
12676608 | 1271 | } |
21cd8d00 JH |
1272 | else |
1273 | pos = -pos-1; | |
12676608 LT |
1274 | |
1275 | /* | |
1276 | * Trivial optimization: if we find an entry that | |
1277 | * already matches the sub-directory, then we know | |
b155725d | 1278 | * we're ok, and we can exit. |
12676608 | 1279 | */ |
4aab5b46 JH |
1280 | while (pos < istate->cache_nr) { |
1281 | struct cache_entry *p = istate->cache[pos]; | |
b155725d JH |
1282 | if ((ce_namelen(p) <= len) || |
1283 | (p->name[len] != '/') || | |
1284 | memcmp(p->name, name, len)) | |
1285 | break; /* not our subdirectory */ | |
077c48df JH |
1286 | if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE)) |
1287 | /* | |
1288 | * p is at the same stage as our entry, and | |
b155725d JH |
1289 | * is a subdirectory of what we are looking |
1290 | * at, so we cannot have conflicts at our | |
1291 | * level or anything shorter. | |
1292 | */ | |
1293 | return retval; | |
1294 | pos++; | |
192268c1 | 1295 | } |
0f1e4f04 | 1296 | } |
12676608 LT |
1297 | return retval; |
1298 | } | |
1299 | ||
1300 | /* We may be in a situation where we already have path/file and path | |
1301 | * is being added, or we already have path and path/file is being | |
1302 | * added. Either one would result in a nonsense tree that has path | |
1303 | * twice when git-write-tree tries to write it out. Prevent it. | |
a6080a0a | 1304 | * |
12676608 LT |
1305 | * If ok-to-replace is specified, we remove the conflicting entries |
1306 | * from the cache so the caller should recompute the insert position. | |
1307 | * When this happens, we return non-zero. | |
1308 | */ | |
4aab5b46 JH |
1309 | static int check_file_directory_conflict(struct index_state *istate, |
1310 | const struct cache_entry *ce, | |
1311 | int pos, int ok_to_replace) | |
12676608 | 1312 | { |
21cd8d00 JH |
1313 | int retval; |
1314 | ||
1315 | /* | |
1316 | * When ce is an "I am going away" entry, we allow it to be added | |
1317 | */ | |
7a51ed66 | 1318 | if (ce->ce_flags & CE_REMOVE) |
21cd8d00 JH |
1319 | return 0; |
1320 | ||
12676608 LT |
1321 | /* |
1322 | * We check if the path is a sub-path of a subsequent pathname | |
1323 | * first, since removing those will not change the position | |
21cd8d00 | 1324 | * in the array. |
12676608 | 1325 | */ |
4aab5b46 | 1326 | retval = has_file_name(istate, ce, pos, ok_to_replace); |
21cd8d00 | 1327 | |
12676608 LT |
1328 | /* |
1329 | * Then check if the path might have a clashing sub-directory | |
1330 | * before it. | |
1331 | */ | |
4aab5b46 | 1332 | return retval + has_dir_name(istate, ce, pos, ok_to_replace); |
0f1e4f04 JH |
1333 | } |
1334 | ||
af3785dc | 1335 | static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option) |
197ee8c9 LT |
1336 | { |
1337 | int pos; | |
192268c1 JH |
1338 | int ok_to_add = option & ADD_CACHE_OK_TO_ADD; |
1339 | int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE; | |
b155725d | 1340 | int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK; |
39425819 | 1341 | int new_only = option & ADD_CACHE_NEW_ONLY; |
5f73076c | 1342 | |
e5494631 JH |
1343 | /* |
1344 | * If this entry's path sorts after the last entry in the index, | |
1345 | * we can avoid searching for it. | |
1346 | */ | |
1347 | if (istate->cache_nr > 0 && | |
1348 | strcmp(ce->name, istate->cache[istate->cache_nr - 1]->name) > 0) | |
c097b95a | 1349 | pos = index_pos_to_insert_pos(istate->cache_nr); |
e5494631 | 1350 | else |
20ec2d03 | 1351 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce), EXPAND_SPARSE); |
197ee8c9 | 1352 | |
c35e9f5e VD |
1353 | /* |
1354 | * Cache tree path should be invalidated only after index_name_stage_pos, | |
1355 | * in case it expands a sparse index. | |
1356 | */ | |
1357 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) | |
1358 | cache_tree_invalidate_path(istate, ce->name); | |
1359 | ||
3e09cdfd | 1360 | /* existing match? Just replace it. */ |
76e7f4ec | 1361 | if (pos >= 0) { |
39425819 JH |
1362 | if (!new_only) |
1363 | replace_index_entry(istate, pos, ce); | |
197ee8c9 LT |
1364 | return 0; |
1365 | } | |
76e7f4ec | 1366 | pos = -pos-1; |
197ee8c9 | 1367 | |
ffcc9ba7 NTND |
1368 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1369 | untracked_cache_add_to_index(istate, ce->name); | |
e931371a | 1370 | |
7b937ca3 LT |
1371 | /* |
1372 | * Inserting a merged entry ("stage 0") into the index | |
1373 | * will always replace all non-merged entries.. | |
1374 | */ | |
4aab5b46 JH |
1375 | if (pos < istate->cache_nr && ce_stage(ce) == 0) { |
1376 | while (ce_same_name(istate->cache[pos], ce)) { | |
7b937ca3 | 1377 | ok_to_add = 1; |
4aab5b46 | 1378 | if (!remove_index_entry_at(istate, pos)) |
7b937ca3 LT |
1379 | break; |
1380 | } | |
1381 | } | |
1382 | ||
121481ab LT |
1383 | if (!ok_to_add) |
1384 | return -1; | |
c8ad9d04 | 1385 | if (verify_path_internal(ce->name, ce->ce_mode) == PATH_INVALID) |
9d0a9e90 | 1386 | return error(_("invalid path '%s'"), ce->name); |
121481ab | 1387 | |
3e09cdfd | 1388 | if (!skip_df_check && |
4aab5b46 | 1389 | check_file_directory_conflict(istate, ce, pos, ok_to_replace)) { |
192268c1 | 1390 | if (!ok_to_replace) |
9d0a9e90 | 1391 | return error(_("'%s' appears as both a file and as a directory"), |
4aab5b46 | 1392 | ce->name); |
20ec2d03 | 1393 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce), EXPAND_SPARSE); |
192268c1 JH |
1394 | pos = -pos-1; |
1395 | } | |
af3785dc JH |
1396 | return pos + 1; |
1397 | } | |
1398 | ||
1399 | int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option) | |
1400 | { | |
1401 | int pos; | |
1402 | ||
1403 | if (option & ADD_CACHE_JUST_APPEND) | |
1404 | pos = istate->cache_nr; | |
1405 | else { | |
1406 | int ret; | |
1407 | ret = add_index_entry_with_check(istate, ce, option); | |
1408 | if (ret <= 0) | |
1409 | return ret; | |
1410 | pos = ret - 1; | |
1411 | } | |
0f1e4f04 | 1412 | |
197ee8c9 | 1413 | /* Make sure the array is big enough .. */ |
999f5660 | 1414 | ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc); |
197ee8c9 LT |
1415 | |
1416 | /* Add it in.. */ | |
4aab5b46 | 1417 | istate->cache_nr++; |
af3785dc | 1418 | if (istate->cache_nr > pos + 1) |
f919ffeb SG |
1419 | MOVE_ARRAY(istate->cache + pos + 1, istate->cache + pos, |
1420 | istate->cache_nr - pos - 1); | |
cf558704 | 1421 | set_index_entry(istate, pos, ce); |
e636a7b4 | 1422 | istate->cache_changed |= CE_ENTRY_ADDED; |
197ee8c9 LT |
1423 | return 0; |
1424 | } | |
1425 | ||
405e5b2f LT |
1426 | /* |
1427 | * "refresh" does not calculate a new sha1 file or bring the | |
1428 | * cache up-to-date for mode/content changes. But what it | |
1429 | * _does_ do is to "re-match" the stat information of a file | |
1430 | * with the cache, so that you can refresh the cache for a | |
1431 | * file that hasn't been changed but where the stat entry is | |
1432 | * out of date. | |
1433 | * | |
1434 | * For example, you'd want to do this after doing a "git-read-tree", | |
1435 | * to link up the stat cache details with the proper files. | |
1436 | */ | |
4aab5b46 | 1437 | static struct cache_entry *refresh_cache_ent(struct index_state *istate, |
4bd5b7da | 1438 | struct cache_entry *ce, |
d05e6970 | 1439 | unsigned int options, int *err, |
a98e0f2d | 1440 | int *changed_ret, |
15268d12 JH |
1441 | int *t2_did_lstat, |
1442 | int *t2_did_scan) | |
405e5b2f LT |
1443 | { |
1444 | struct stat st; | |
1445 | struct cache_entry *updated; | |
a849735b | 1446 | int changed; |
25762726 | 1447 | int refresh = options & CE_MATCH_REFRESH; |
4bd5b7da | 1448 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 1449 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
2e2e7ec1 | 1450 | int ignore_missing = options & CE_MATCH_IGNORE_MISSING; |
883e248b | 1451 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
405e5b2f | 1452 | |
25762726 | 1453 | if (!refresh || ce_uptodate(ce)) |
eadb5831 JH |
1454 | return ce; |
1455 | ||
883e248b BP |
1456 | if (!ignore_fsmonitor) |
1457 | refresh_fsmonitor(istate); | |
aa9349d4 | 1458 | /* |
56cac48c NTND |
1459 | * CE_VALID or CE_SKIP_WORKTREE means the user promised us |
1460 | * that the change to the work tree does not matter and told | |
1461 | * us not to worry. | |
aa9349d4 | 1462 | */ |
56cac48c NTND |
1463 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) { |
1464 | ce_mark_uptodate(ce); | |
1465 | return ce; | |
1466 | } | |
aa9349d4 MSO |
1467 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) { |
1468 | ce_mark_uptodate(ce); | |
1469 | return ce; | |
1470 | } | |
883e248b BP |
1471 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) { |
1472 | ce_mark_uptodate(ce); | |
1473 | return ce; | |
1474 | } | |
aa9349d4 | 1475 | |
ccad42d4 RS |
1476 | if (has_symlink_leading_path(ce->name, ce_namelen(ce))) { |
1477 | if (ignore_missing) | |
1478 | return ce; | |
1479 | if (err) | |
1480 | *err = ENOENT; | |
1481 | return NULL; | |
1482 | } | |
1483 | ||
a98e0f2d JH |
1484 | if (t2_did_lstat) |
1485 | *t2_did_lstat = 1; | |
8fd2cb40 | 1486 | if (lstat(ce->name, &st) < 0) { |
2e2e7ec1 BK |
1487 | if (ignore_missing && errno == ENOENT) |
1488 | return ce; | |
ec0cc704 JH |
1489 | if (err) |
1490 | *err = errno; | |
8fd2cb40 JS |
1491 | return NULL; |
1492 | } | |
405e5b2f | 1493 | |
4bd5b7da | 1494 | changed = ie_match_stat(istate, ce, &st, options); |
d05e6970 JK |
1495 | if (changed_ret) |
1496 | *changed_ret = changed; | |
405e5b2f | 1497 | if (!changed) { |
4bd5b7da JH |
1498 | /* |
1499 | * The path is unchanged. If we were told to ignore | |
1500 | * valid bit, then we did the actual stat check and | |
1501 | * found that the entry is unmodified. If the entry | |
1502 | * is not marked VALID, this is the place to mark it | |
1503 | * valid again, under "assume unchanged" mode. | |
1504 | */ | |
1505 | if (ignore_valid && assume_unchanged && | |
7a51ed66 | 1506 | !(ce->ce_flags & CE_VALID)) |
405e5b2f | 1507 | ; /* mark this one VALID again */ |
eadb5831 JH |
1508 | else { |
1509 | /* | |
1510 | * We do not mark the index itself "modified" | |
1511 | * because CE_UPTODATE flag is in-core only; | |
1512 | * we are not going to write this change out. | |
1513 | */ | |
883e248b | 1514 | if (!S_ISGITLINK(ce->ce_mode)) { |
125fd984 | 1515 | ce_mark_uptodate(ce); |
b5a81697 | 1516 | mark_fsmonitor_valid(istate, ce); |
883e248b | 1517 | } |
8fd2cb40 | 1518 | return ce; |
eadb5831 | 1519 | } |
405e5b2f LT |
1520 | } |
1521 | ||
15268d12 JH |
1522 | if (t2_did_scan) |
1523 | *t2_did_scan = 1; | |
4bd5b7da | 1524 | if (ie_modified(istate, ce, &st, options)) { |
ec0cc704 JH |
1525 | if (err) |
1526 | *err = EINVAL; | |
8fd2cb40 JS |
1527 | return NULL; |
1528 | } | |
405e5b2f | 1529 | |
a849735b | 1530 | updated = make_empty_cache_entry(istate, ce_namelen(ce)); |
0e267b7a BP |
1531 | copy_cache_entry(updated, ce); |
1532 | memcpy(updated->name, ce->name, ce->ce_namelen + 1); | |
d4c0a3ac | 1533 | fill_stat_cache_info(istate, updated, &st); |
4bd5b7da JH |
1534 | /* |
1535 | * If ignore_valid is not set, we should leave CE_VALID bit | |
1536 | * alone. Otherwise, paths marked with --no-assume-unchanged | |
1537 | * (i.e. things to be edited) will reacquire CE_VALID bit | |
1538 | * automatically, which is not really what we want. | |
405e5b2f | 1539 | */ |
4bd5b7da | 1540 | if (!ignore_valid && assume_unchanged && |
7a51ed66 LT |
1541 | !(ce->ce_flags & CE_VALID)) |
1542 | updated->ce_flags &= ~CE_VALID; | |
405e5b2f | 1543 | |
e636a7b4 | 1544 | /* istate->cache_changed is updated in the caller */ |
405e5b2f LT |
1545 | return updated; |
1546 | } | |
1547 | ||
3deffc52 | 1548 | static void show_file(const char * fmt, const char * name, int in_porcelain, |
046613c5 | 1549 | int * first, const char *header_msg) |
3deffc52 MM |
1550 | { |
1551 | if (in_porcelain && *first && header_msg) { | |
1552 | printf("%s\n", header_msg); | |
cd2b8ae9 | 1553 | *first = 0; |
3deffc52 MM |
1554 | } |
1555 | printf(fmt, name); | |
1556 | } | |
1557 | ||
22184497 TG |
1558 | int repo_refresh_and_write_index(struct repository *repo, |
1559 | unsigned int refresh_flags, | |
1560 | unsigned int write_flags, | |
1561 | int gentle, | |
1562 | const struct pathspec *pathspec, | |
1563 | char *seen, const char *header_msg) | |
1564 | { | |
1565 | struct lock_file lock_file = LOCK_INIT; | |
1566 | int fd, ret = 0; | |
1567 | ||
1568 | fd = repo_hold_locked_index(repo, &lock_file, 0); | |
1569 | if (!gentle && fd < 0) | |
1570 | return -1; | |
1571 | if (refresh_index(repo->index, refresh_flags, pathspec, seen, header_msg)) | |
1572 | ret = 1; | |
1573 | if (0 <= fd && write_locked_index(repo->index, &lock_file, COMMIT_LOCK | write_flags)) | |
1574 | ret = -1; | |
1575 | return ret; | |
1576 | } | |
1577 | ||
1578 | ||
9b2d6149 NTND |
1579 | int refresh_index(struct index_state *istate, unsigned int flags, |
1580 | const struct pathspec *pathspec, | |
046613c5 | 1581 | char *seen, const char *header_msg) |
405e5b2f LT |
1582 | { |
1583 | int i; | |
1584 | int has_errors = 0; | |
1585 | int really = (flags & REFRESH_REALLY) != 0; | |
1586 | int allow_unmerged = (flags & REFRESH_UNMERGED) != 0; | |
1587 | int quiet = (flags & REFRESH_QUIET) != 0; | |
1588 | int not_new = (flags & REFRESH_IGNORE_MISSING) != 0; | |
5fdeacb0 | 1589 | int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0; |
b243012c | 1590 | int ignore_skip_worktree = (flags & REFRESH_IGNORE_SKIP_WORKTREE) != 0; |
3deffc52 MM |
1591 | int first = 1; |
1592 | int in_porcelain = (flags & REFRESH_IN_PORCELAIN); | |
25762726 BK |
1593 | unsigned int options = (CE_MATCH_REFRESH | |
1594 | (really ? CE_MATCH_IGNORE_VALID : 0) | | |
2e2e7ec1 | 1595 | (not_new ? CE_MATCH_IGNORE_MISSING : 0)); |
4bd4e730 | 1596 | const char *modified_fmt; |
73b7eae6 JK |
1597 | const char *deleted_fmt; |
1598 | const char *typechange_fmt; | |
1599 | const char *added_fmt; | |
4bd4e730 | 1600 | const char *unmerged_fmt; |
ae9af122 | 1601 | struct progress *progress = NULL; |
a98e0f2d | 1602 | int t2_sum_lstat = 0; |
15268d12 | 1603 | int t2_sum_scan = 0; |
ae9af122 NTND |
1604 | |
1605 | if (flags & REFRESH_PROGRESS && isatty(2)) | |
1606 | progress = start_delayed_progress(_("Refresh index"), | |
1607 | istate->cache_nr); | |
405e5b2f | 1608 | |
c46c406a | 1609 | trace_performance_enter(); |
a71806a7 NTND |
1610 | modified_fmt = in_porcelain ? "M\t%s\n" : "%s: needs update\n"; |
1611 | deleted_fmt = in_porcelain ? "D\t%s\n" : "%s: needs update\n"; | |
1612 | typechange_fmt = in_porcelain ? "T\t%s\n" : "%s: needs update\n"; | |
1613 | added_fmt = in_porcelain ? "A\t%s\n" : "%s: needs update\n"; | |
1614 | unmerged_fmt = in_porcelain ? "U\t%s\n" : "%s: needs merge\n"; | |
99ce720c BP |
1615 | /* |
1616 | * Use the multi-threaded preload_index() to refresh most of the | |
1617 | * cache entries quickly then in the single threaded loop below, | |
1618 | * we only have to do the special cases that are left. | |
1619 | */ | |
1620 | preload_index(istate, pathspec, 0); | |
a98e0f2d | 1621 | trace2_region_enter("index", "refresh", NULL); |
d76723ee | 1622 | |
4aab5b46 | 1623 | for (i = 0; i < istate->cache_nr; i++) { |
285c2e25 | 1624 | struct cache_entry *ce, *new_entry; |
ec0cc704 | 1625 | int cache_errno = 0; |
73b7eae6 | 1626 | int changed = 0; |
3d1f148c | 1627 | int filtered = 0; |
a98e0f2d | 1628 | int t2_did_lstat = 0; |
15268d12 | 1629 | int t2_did_scan = 0; |
ec0cc704 | 1630 | |
4aab5b46 | 1631 | ce = istate->cache[i]; |
5fdeacb0 JS |
1632 | if (ignore_submodules && S_ISGITLINK(ce->ce_mode)) |
1633 | continue; | |
b243012c MT |
1634 | if (ignore_skip_worktree && ce_skip_worktree(ce)) |
1635 | continue; | |
5fdeacb0 | 1636 | |
d76723ee DS |
1637 | /* |
1638 | * If this entry is a sparse directory, then there isn't | |
1639 | * any stat() information to update. Ignore the entry. | |
1640 | */ | |
1641 | if (S_ISSPARSEDIR(ce->ce_mode)) | |
1642 | continue; | |
1643 | ||
d7b665c3 | 1644 | if (pathspec && !ce_path_match(istate, ce, pathspec, seen)) |
3d1f148c JH |
1645 | filtered = 1; |
1646 | ||
405e5b2f | 1647 | if (ce_stage(ce)) { |
4aab5b46 JH |
1648 | while ((i < istate->cache_nr) && |
1649 | ! strcmp(istate->cache[i]->name, ce->name)) | |
405e5b2f LT |
1650 | i++; |
1651 | i--; | |
1652 | if (allow_unmerged) | |
1653 | continue; | |
3d1f148c JH |
1654 | if (!filtered) |
1655 | show_file(unmerged_fmt, ce->name, in_porcelain, | |
1656 | &first, header_msg); | |
405e5b2f LT |
1657 | has_errors = 1; |
1658 | continue; | |
1659 | } | |
1660 | ||
3d1f148c | 1661 | if (filtered) |
d616813d AJ |
1662 | continue; |
1663 | ||
a98e0f2d JH |
1664 | new_entry = refresh_cache_ent(istate, ce, options, |
1665 | &cache_errno, &changed, | |
15268d12 | 1666 | &t2_did_lstat, &t2_did_scan); |
a98e0f2d | 1667 | t2_sum_lstat += t2_did_lstat; |
15268d12 | 1668 | t2_sum_scan += t2_did_scan; |
285c2e25 | 1669 | if (new_entry == ce) |
405e5b2f | 1670 | continue; |
b7b793d1 | 1671 | display_progress(progress, i); |
285c2e25 | 1672 | if (!new_entry) { |
73b7eae6 JK |
1673 | const char *fmt; |
1674 | ||
8fd2cb40 | 1675 | if (really && cache_errno == EINVAL) { |
405e5b2f LT |
1676 | /* If we are doing --really-refresh that |
1677 | * means the index is not valid anymore. | |
1678 | */ | |
7a51ed66 | 1679 | ce->ce_flags &= ~CE_VALID; |
078a58e8 | 1680 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 1681 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 1682 | istate->cache_changed |= CE_ENTRY_CHANGED; |
405e5b2f LT |
1683 | } |
1684 | if (quiet) | |
1685 | continue; | |
73b7eae6 JK |
1686 | |
1687 | if (cache_errno == ENOENT) | |
1688 | fmt = deleted_fmt; | |
895ff3b2 | 1689 | else if (ce_intent_to_add(ce)) |
73b7eae6 JK |
1690 | fmt = added_fmt; /* must be before other checks */ |
1691 | else if (changed & TYPE_CHANGED) | |
1692 | fmt = typechange_fmt; | |
1693 | else | |
1694 | fmt = modified_fmt; | |
1695 | show_file(fmt, | |
1696 | ce->name, in_porcelain, &first, header_msg); | |
405e5b2f LT |
1697 | has_errors = 1; |
1698 | continue; | |
1699 | } | |
cf558704 | 1700 | |
285c2e25 | 1701 | replace_index_entry(istate, i, new_entry); |
405e5b2f | 1702 | } |
a98e0f2d | 1703 | trace2_data_intmax("index", NULL, "refresh/sum_lstat", t2_sum_lstat); |
15268d12 | 1704 | trace2_data_intmax("index", NULL, "refresh/sum_scan", t2_sum_scan); |
a98e0f2d | 1705 | trace2_region_leave("index", "refresh", NULL); |
b7b793d1 ÆAB |
1706 | display_progress(progress, istate->cache_nr); |
1707 | stop_progress(&progress); | |
c46c406a | 1708 | trace_performance_leave("refresh index"); |
405e5b2f LT |
1709 | return has_errors; |
1710 | } | |
1711 | ||
768d7965 JM |
1712 | struct cache_entry *refresh_cache_entry(struct index_state *istate, |
1713 | struct cache_entry *ce, | |
1714 | unsigned int options) | |
ec0cc704 | 1715 | { |
15268d12 | 1716 | return refresh_cache_ent(istate, ce, options, NULL, NULL, NULL, NULL); |
ec0cc704 JH |
1717 | } |
1718 | ||
db3b313c JH |
1719 | |
1720 | /***************************************************************** | |
1721 | * Index File I/O | |
1722 | *****************************************************************/ | |
1723 | ||
9d227781 JH |
1724 | #define INDEX_FORMAT_DEFAULT 3 |
1725 | ||
7211b9e7 | 1726 | static unsigned int get_index_format_default(struct repository *r) |
136347d7 TG |
1727 | { |
1728 | char *envversion = getenv("GIT_INDEX_VERSION"); | |
3c09d684 TG |
1729 | char *endp; |
1730 | unsigned int version = INDEX_FORMAT_DEFAULT; | |
1731 | ||
136347d7 | 1732 | if (!envversion) { |
7211b9e7 DS |
1733 | prepare_repo_settings(r); |
1734 | ||
1735 | if (r->settings.index_version >= 0) | |
1736 | version = r->settings.index_version; | |
3c09d684 TG |
1737 | if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { |
1738 | warning(_("index.version set, but the value is invalid.\n" | |
136347d7 | 1739 | "Using version %i"), INDEX_FORMAT_DEFAULT); |
3c09d684 | 1740 | return INDEX_FORMAT_DEFAULT; |
136347d7 TG |
1741 | } |
1742 | return version; | |
1743 | } | |
3c09d684 TG |
1744 | |
1745 | version = strtoul(envversion, &endp, 10); | |
1746 | if (*endp || | |
1747 | version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { | |
1748 | warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n" | |
1749 | "Using version %i"), INDEX_FORMAT_DEFAULT); | |
1750 | version = INDEX_FORMAT_DEFAULT; | |
1751 | } | |
1752 | return version; | |
136347d7 TG |
1753 | } |
1754 | ||
db3b313c JH |
1755 | /* |
1756 | * dev/ino/uid/gid/size are also just tracked to the low 32 bits | |
1757 | * Again - this is just a (very strong in practice) heuristic that | |
1758 | * the inode hasn't changed. | |
1759 | * | |
1760 | * We save the fields in big-endian order to allow using the | |
1761 | * index file over NFS transparently. | |
1762 | */ | |
1763 | struct ondisk_cache_entry { | |
1764 | struct cache_time ctime; | |
1765 | struct cache_time mtime; | |
7800c1eb TG |
1766 | uint32_t dev; |
1767 | uint32_t ino; | |
1768 | uint32_t mode; | |
1769 | uint32_t uid; | |
1770 | uint32_t gid; | |
1771 | uint32_t size; | |
575fa8a3 | 1772 | /* |
1773 | * unsigned char hash[hashsz]; | |
1774 | * uint16_t flags; | |
1775 | * if (flags & CE_EXTENDED) | |
1776 | * uint16_t flags2; | |
1777 | */ | |
1778 | unsigned char data[GIT_MAX_RAWSZ + 2 * sizeof(uint16_t)]; | |
1779 | char name[FLEX_ARRAY]; | |
db3b313c JH |
1780 | }; |
1781 | ||
6c9cd161 | 1782 | /* These are only used for v3 or lower */ |
ce012deb | 1783 | #define align_padding_size(size, len) ((size + (len) + 8) & ~7) - (size + len) |
575fa8a3 | 1784 | #define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,data) + (len) + 8) & ~7) |
db3b313c | 1785 | #define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len) |
575fa8a3 | 1786 | #define ondisk_data_size(flags, len) (the_hash_algo->rawsz + \ |
1787 | ((flags & CE_EXTENDED) ? 2 : 1) * sizeof(uint16_t) + len) | |
1788 | #define ondisk_data_size_max(len) (ondisk_data_size(CE_EXTENDED, len)) | |
1789 | #define ondisk_ce_size(ce) (ondisk_cache_entry_size(ondisk_data_size((ce)->ce_flags, ce_namelen(ce)))) | |
db3b313c | 1790 | |
a33fc72f JH |
1791 | /* Allow fsck to force verification of the index checksum. */ |
1792 | int verify_index_checksum; | |
1793 | ||
00ec50e5 BP |
1794 | /* Allow fsck to force verification of the cache entry order. */ |
1795 | int verify_ce_order; | |
1796 | ||
371ed0de | 1797 | static int verify_hdr(const struct cache_header *hdr, unsigned long size) |
e83c5163 | 1798 | { |
aab61359 | 1799 | git_hash_ctx c; |
1800 | unsigned char hash[GIT_MAX_RAWSZ]; | |
0136bac9 | 1801 | int hdr_version; |
e83c5163 | 1802 | |
ccc4feb5 | 1803 | if (hdr->hdr_signature != htonl(CACHE_SIGNATURE)) |
9d0a9e90 | 1804 | return error(_("bad signature 0x%08x"), hdr->hdr_signature); |
0136bac9 | 1805 | hdr_version = ntohl(hdr->hdr_version); |
b82a7b5b | 1806 | if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version) |
9d0a9e90 | 1807 | return error(_("bad index version %d"), hdr_version); |
a33fc72f JH |
1808 | |
1809 | if (!verify_index_checksum) | |
1810 | return 0; | |
1811 | ||
aab61359 | 1812 | the_hash_algo->init_fn(&c); |
1813 | the_hash_algo->update_fn(&c, hdr, size - the_hash_algo->rawsz); | |
1814 | the_hash_algo->final_fn(hash, &c); | |
67947c34 | 1815 | if (!hasheq(hash, (unsigned char *)hdr + size - the_hash_algo->rawsz)) |
9d0a9e90 | 1816 | return error(_("bad index file sha1 signature")); |
e83c5163 LT |
1817 | return 0; |
1818 | } | |
1819 | ||
4aab5b46 | 1820 | static int read_index_extension(struct index_state *istate, |
371ed0de | 1821 | const char *ext, const char *data, unsigned long sz) |
bad68ec9 JH |
1822 | { |
1823 | switch (CACHE_EXT(ext)) { | |
1824 | case CACHE_EXT_TREE: | |
4aab5b46 | 1825 | istate->cache_tree = cache_tree_read(data, sz); |
bad68ec9 | 1826 | break; |
cfc5789a JH |
1827 | case CACHE_EXT_RESOLVE_UNDO: |
1828 | istate->resolve_undo = resolve_undo_read(data, sz); | |
1829 | break; | |
5fc2fc8f NTND |
1830 | case CACHE_EXT_LINK: |
1831 | if (read_link_extension(istate, data, sz)) | |
1832 | return -1; | |
1833 | break; | |
f9e6c649 NTND |
1834 | case CACHE_EXT_UNTRACKED: |
1835 | istate->untracked = read_untracked_extension(data, sz); | |
1836 | break; | |
883e248b BP |
1837 | case CACHE_EXT_FSMONITOR: |
1838 | read_fsmonitor_extension(istate, data, sz); | |
1839 | break; | |
3b1d9e04 | 1840 | case CACHE_EXT_ENDOFINDEXENTRIES: |
3255089a | 1841 | case CACHE_EXT_INDEXENTRYOFFSETTABLE: |
3b1d9e04 BP |
1842 | /* already handled in do_read_index() */ |
1843 | break; | |
cd42415f DS |
1844 | case CACHE_EXT_SPARSE_DIRECTORIES: |
1845 | /* no content, only an indicator */ | |
1846 | istate->sparse_index = 1; | |
1847 | break; | |
bad68ec9 JH |
1848 | default: |
1849 | if (*ext < 'A' || 'Z' < *ext) | |
9d0a9e90 | 1850 | return error(_("index uses %.4s extension, which we do not understand"), |
bad68ec9 | 1851 | ext); |
9d0a9e90 | 1852 | fprintf_ln(stderr, _("ignoring %.4s extension"), ext); |
bad68ec9 JH |
1853 | break; |
1854 | } | |
1855 | return 0; | |
1856 | } | |
1857 | ||
77ff1127 BP |
1858 | static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool, |
1859 | unsigned int version, | |
a849735b | 1860 | struct ondisk_cache_entry *ondisk, |
6c9cd161 | 1861 | unsigned long *ent_size, |
252d079c | 1862 | const struct cache_entry *previous_ce) |
7a51ed66 | 1863 | { |
debed2a6 | 1864 | struct cache_entry *ce; |
7fec10b7 | 1865 | size_t len; |
06aaaa0b | 1866 | const char *name; |
575fa8a3 | 1867 | const unsigned hashsz = the_hash_algo->rawsz; |
1868 | const uint16_t *flagsp = (const uint16_t *)(ondisk->data + hashsz); | |
debed2a6 | 1869 | unsigned int flags; |
f5c4a9af | 1870 | size_t copy_len = 0; |
252d079c NTND |
1871 | /* |
1872 | * Adjacent cache entries tend to share the leading paths, so it makes | |
1873 | * sense to only store the differences in later entries. In the v4 | |
1874 | * on-disk format of the index, each on-disk cache entry stores the | |
1875 | * number of bytes to be stripped from the end of the previous name, | |
1876 | * and the bytes to append to the result, to come up with its name. | |
1877 | */ | |
77ff1127 | 1878 | int expand_name_field = version == 4; |
7fec10b7 | 1879 | |
7a51ed66 | 1880 | /* On-disk flags are just 16 bits */ |
575fa8a3 | 1881 | flags = get_be16(flagsp); |
debed2a6 | 1882 | len = flags & CE_NAMEMASK; |
7fec10b7 | 1883 | |
debed2a6 | 1884 | if (flags & CE_EXTENDED) { |
06aaaa0b | 1885 | int extended_flags; |
575fa8a3 | 1886 | extended_flags = get_be16(flagsp + 1) << 16; |
06aaaa0b NTND |
1887 | /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */ |
1888 | if (extended_flags & ~CE_EXTENDED_FLAGS) | |
9d0a9e90 | 1889 | die(_("unknown index entry format 0x%08x"), extended_flags); |
debed2a6 | 1890 | flags |= extended_flags; |
575fa8a3 | 1891 | name = (const char *)(flagsp + 2); |
06aaaa0b NTND |
1892 | } |
1893 | else | |
575fa8a3 | 1894 | name = (const char *)(flagsp + 1); |
06aaaa0b | 1895 | |
252d079c NTND |
1896 | if (expand_name_field) { |
1897 | const unsigned char *cp = (const unsigned char *)name; | |
1898 | size_t strip_len, previous_len; | |
6c9cd161 | 1899 | |
15beaaa3 | 1900 | /* If we're at the beginning of a block, ignore the previous name */ |
252d079c | 1901 | strip_len = decode_varint(&cp); |
77ff1127 BP |
1902 | if (previous_ce) { |
1903 | previous_len = previous_ce->ce_namelen; | |
1904 | if (previous_len < strip_len) | |
252d079c | 1905 | die(_("malformed name field in the index, near path '%s'"), |
77ff1127 BP |
1906 | previous_ce->name); |
1907 | copy_len = previous_len - strip_len; | |
252d079c | 1908 | } |
252d079c NTND |
1909 | name = (const char *)cp; |
1910 | } | |
6c9cd161 | 1911 | |
252d079c NTND |
1912 | if (len == CE_NAMEMASK) { |
1913 | len = strlen(name); | |
1914 | if (expand_name_field) | |
1915 | len += copy_len; | |
1916 | } | |
6c9cd161 | 1917 | |
77ff1127 | 1918 | ce = mem_pool__ce_alloc(ce_mem_pool, len); |
252d079c NTND |
1919 | |
1920 | ce->ce_stat_data.sd_ctime.sec = get_be32(&ondisk->ctime.sec); | |
1921 | ce->ce_stat_data.sd_mtime.sec = get_be32(&ondisk->mtime.sec); | |
1922 | ce->ce_stat_data.sd_ctime.nsec = get_be32(&ondisk->ctime.nsec); | |
1923 | ce->ce_stat_data.sd_mtime.nsec = get_be32(&ondisk->mtime.nsec); | |
1924 | ce->ce_stat_data.sd_dev = get_be32(&ondisk->dev); | |
1925 | ce->ce_stat_data.sd_ino = get_be32(&ondisk->ino); | |
1926 | ce->ce_mode = get_be32(&ondisk->mode); | |
1927 | ce->ce_stat_data.sd_uid = get_be32(&ondisk->uid); | |
1928 | ce->ce_stat_data.sd_gid = get_be32(&ondisk->gid); | |
1929 | ce->ce_stat_data.sd_size = get_be32(&ondisk->size); | |
1930 | ce->ce_flags = flags & ~CE_NAMEMASK; | |
1931 | ce->ce_namelen = len; | |
1932 | ce->index = 0; | |
92e2cab9 | 1933 | oidread(&ce->oid, ondisk->data); |
575fa8a3 | 1934 | memcpy(ce->name, name, len); |
1935 | ce->name[len] = '\0'; | |
6c9cd161 | 1936 | |
252d079c NTND |
1937 | if (expand_name_field) { |
1938 | if (copy_len) | |
1939 | memcpy(ce->name, previous_ce->name, copy_len); | |
1940 | memcpy(ce->name + copy_len, name, len + 1 - copy_len); | |
1941 | *ent_size = (name - ((char *)ondisk)) + len + 1 - copy_len; | |
1942 | } else { | |
1943 | memcpy(ce->name, name, len + 1); | |
1944 | *ent_size = ondisk_ce_size(ce); | |
6c9cd161 | 1945 | } |
debed2a6 | 1946 | return ce; |
cf558704 LT |
1947 | } |
1948 | ||
03f15a79 | 1949 | static void check_ce_order(struct index_state *istate) |
15999d0b | 1950 | { |
03f15a79 TG |
1951 | unsigned int i; |
1952 | ||
00ec50e5 BP |
1953 | if (!verify_ce_order) |
1954 | return; | |
1955 | ||
03f15a79 TG |
1956 | for (i = 1; i < istate->cache_nr; i++) { |
1957 | struct cache_entry *ce = istate->cache[i - 1]; | |
1958 | struct cache_entry *next_ce = istate->cache[i]; | |
1959 | int name_compare = strcmp(ce->name, next_ce->name); | |
1960 | ||
1961 | if (0 < name_compare) | |
9d0a9e90 | 1962 | die(_("unordered stage entries in index")); |
03f15a79 TG |
1963 | if (!name_compare) { |
1964 | if (!ce_stage(ce)) | |
9d0a9e90 | 1965 | die(_("multiple stage entries for merged file '%s'"), |
03f15a79 TG |
1966 | ce->name); |
1967 | if (ce_stage(ce) > ce_stage(next_ce)) | |
9d0a9e90 | 1968 | die(_("unordered stage entries for '%s'"), |
03f15a79 TG |
1969 | ce->name); |
1970 | } | |
15999d0b JSP |
1971 | } |
1972 | } | |
1973 | ||
435ec090 CC |
1974 | static void tweak_untracked_cache(struct index_state *istate) |
1975 | { | |
ad0fb659 DS |
1976 | struct repository *r = the_repository; |
1977 | ||
1978 | prepare_repo_settings(r); | |
1979 | ||
f1bee828 ÆAB |
1980 | switch (r->settings.core_untracked_cache) { |
1981 | case UNTRACKED_CACHE_REMOVE: | |
435ec090 | 1982 | remove_untracked_cache(istate); |
f1bee828 ÆAB |
1983 | break; |
1984 | case UNTRACKED_CACHE_WRITE: | |
ad0fb659 | 1985 | add_untracked_cache(istate); |
f1bee828 ÆAB |
1986 | break; |
1987 | case UNTRACKED_CACHE_KEEP: | |
3050b6df ÆAB |
1988 | /* |
1989 | * Either an explicit "core.untrackedCache=keep", the | |
1990 | * default if "core.untrackedCache" isn't configured, | |
1991 | * or a fallback on an unknown "core.untrackedCache" | |
1992 | * value. | |
1993 | */ | |
f1bee828 | 1994 | break; |
f1bee828 | 1995 | } |
435ec090 CC |
1996 | } |
1997 | ||
43925312 CC |
1998 | static void tweak_split_index(struct index_state *istate) |
1999 | { | |
2000 | switch (git_config_get_split_index()) { | |
2001 | case -1: /* unset: do nothing */ | |
2002 | break; | |
2003 | case 0: /* false */ | |
2004 | remove_split_index(istate); | |
2005 | break; | |
2006 | case 1: /* true */ | |
2007 | add_split_index(istate); | |
2008 | break; | |
2009 | default: /* unknown value: do nothing */ | |
2010 | break; | |
2011 | } | |
2012 | } | |
2013 | ||
435ec090 CC |
2014 | static void post_read_index_from(struct index_state *istate) |
2015 | { | |
2016 | check_ce_order(istate); | |
2017 | tweak_untracked_cache(istate); | |
43925312 | 2018 | tweak_split_index(istate); |
883e248b | 2019 | tweak_fsmonitor(istate); |
435ec090 CC |
2020 | } |
2021 | ||
8e72d675 JM |
2022 | static size_t estimate_cache_size_from_compressed(unsigned int entries) |
2023 | { | |
2024 | return entries * (sizeof(struct cache_entry) + CACHE_ENTRY_PATH_LENGTH); | |
2025 | } | |
2026 | ||
2027 | static size_t estimate_cache_size(size_t ondisk_size, unsigned int entries) | |
2028 | { | |
2029 | long per_entry = sizeof(struct cache_entry) - sizeof(struct ondisk_cache_entry); | |
2030 | ||
2031 | /* | |
2032 | * Account for potential alignment differences. | |
2033 | */ | |
c097b95a | 2034 | per_entry += align_padding_size(per_entry, 0); |
8e72d675 JM |
2035 | return ondisk_size + entries * per_entry; |
2036 | } | |
2037 | ||
3255089a BP |
2038 | struct index_entry_offset |
2039 | { | |
2040 | /* starting byte offset into index file, count of index entries in this block */ | |
2041 | int offset, nr; | |
2042 | }; | |
2043 | ||
2044 | struct index_entry_offset_table | |
2045 | { | |
2046 | int nr; | |
2047 | struct index_entry_offset entries[FLEX_ARRAY]; | |
2048 | }; | |
2049 | ||
3255089a BP |
2050 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset); |
2051 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot); | |
3255089a | 2052 | |
3b1d9e04 BP |
2053 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size); |
2054 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset); | |
2055 | ||
abb4bb83 BP |
2056 | struct load_index_extensions |
2057 | { | |
abb4bb83 | 2058 | pthread_t pthread; |
abb4bb83 BP |
2059 | struct index_state *istate; |
2060 | const char *mmap; | |
2061 | size_t mmap_size; | |
2062 | unsigned long src_offset; | |
2063 | }; | |
2064 | ||
2065 | static void *load_index_extensions(void *_data) | |
2066 | { | |
2067 | struct load_index_extensions *p = _data; | |
2068 | unsigned long src_offset = p->src_offset; | |
2069 | ||
2070 | while (src_offset <= p->mmap_size - the_hash_algo->rawsz - 8) { | |
2071 | /* After an array of active_nr index entries, | |
2072 | * there can be arbitrary number of extended | |
2073 | * sections, each of which is prefixed with | |
2074 | * extension name (4-byte) and section length | |
2075 | * in 4-byte network byte order. | |
2076 | */ | |
2077 | uint32_t extsize = get_be32(p->mmap + src_offset + 4); | |
2078 | if (read_index_extension(p->istate, | |
2079 | p->mmap + src_offset, | |
2080 | p->mmap + src_offset + 8, | |
2081 | extsize) < 0) { | |
2082 | munmap((void *)p->mmap, p->mmap_size); | |
2083 | die(_("index file corrupt")); | |
2084 | } | |
2085 | src_offset += 8; | |
2086 | src_offset += extsize; | |
2087 | } | |
2088 | ||
2089 | return NULL; | |
2090 | } | |
2091 | ||
77ff1127 BP |
2092 | /* |
2093 | * A helper function that will load the specified range of cache entries | |
2094 | * from the memory mapped file and add them to the given index. | |
2095 | */ | |
2096 | static unsigned long load_cache_entry_block(struct index_state *istate, | |
2097 | struct mem_pool *ce_mem_pool, int offset, int nr, const char *mmap, | |
2098 | unsigned long start_offset, const struct cache_entry *previous_ce) | |
2099 | { | |
2100 | int i; | |
2101 | unsigned long src_offset = start_offset; | |
2102 | ||
2103 | for (i = offset; i < offset + nr; i++) { | |
2104 | struct ondisk_cache_entry *disk_ce; | |
2105 | struct cache_entry *ce; | |
2106 | unsigned long consumed; | |
2107 | ||
2108 | disk_ce = (struct ondisk_cache_entry *)(mmap + src_offset); | |
2109 | ce = create_from_disk(ce_mem_pool, istate->version, disk_ce, &consumed, previous_ce); | |
2110 | set_index_entry(istate, i, ce); | |
2111 | ||
2112 | src_offset += consumed; | |
2113 | previous_ce = ce; | |
2114 | } | |
2115 | return src_offset - start_offset; | |
2116 | } | |
2117 | ||
2118 | static unsigned long load_all_cache_entries(struct index_state *istate, | |
2119 | const char *mmap, size_t mmap_size, unsigned long src_offset) | |
2120 | { | |
2121 | unsigned long consumed; | |
2122 | ||
44c7e1a7 | 2123 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2124 | if (istate->version == 4) { |
44c7e1a7 | 2125 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2126 | estimate_cache_size_from_compressed(istate->cache_nr)); |
2127 | } else { | |
44c7e1a7 | 2128 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2129 | estimate_cache_size(mmap_size, istate->cache_nr)); |
2130 | } | |
2131 | ||
2132 | consumed = load_cache_entry_block(istate, istate->ce_mem_pool, | |
2133 | 0, istate->cache_nr, mmap, src_offset, NULL); | |
2134 | return consumed; | |
2135 | } | |
2136 | ||
3255089a BP |
2137 | /* |
2138 | * Mostly randomly chosen maximum thread counts: we | |
2139 | * cap the parallelism to online_cpus() threads, and we want | |
2140 | * to have at least 10000 cache entries per thread for it to | |
2141 | * be worth starting a thread. | |
2142 | */ | |
2143 | ||
2144 | #define THREAD_COST (10000) | |
2145 | ||
77ff1127 BP |
2146 | struct load_cache_entries_thread_data |
2147 | { | |
2148 | pthread_t pthread; | |
2149 | struct index_state *istate; | |
2150 | struct mem_pool *ce_mem_pool; | |
2151 | int offset; | |
2152 | const char *mmap; | |
2153 | struct index_entry_offset_table *ieot; | |
2154 | int ieot_start; /* starting index into the ieot array */ | |
2155 | int ieot_blocks; /* count of ieot entries to process */ | |
2156 | unsigned long consumed; /* return # of bytes in index file processed */ | |
2157 | }; | |
2158 | ||
2159 | /* | |
2160 | * A thread proc to run the load_cache_entries() computation | |
2161 | * across multiple background threads. | |
2162 | */ | |
2163 | static void *load_cache_entries_thread(void *_data) | |
2164 | { | |
2165 | struct load_cache_entries_thread_data *p = _data; | |
2166 | int i; | |
2167 | ||
2168 | /* iterate across all ieot blocks assigned to this thread */ | |
2169 | for (i = p->ieot_start; i < p->ieot_start + p->ieot_blocks; i++) { | |
2170 | p->consumed += load_cache_entry_block(p->istate, p->ce_mem_pool, | |
2171 | p->offset, p->ieot->entries[i].nr, p->mmap, p->ieot->entries[i].offset, NULL); | |
2172 | p->offset += p->ieot->entries[i].nr; | |
2173 | } | |
2174 | return NULL; | |
2175 | } | |
2176 | ||
2177 | static unsigned long load_cache_entries_threaded(struct index_state *istate, const char *mmap, size_t mmap_size, | |
7bd9631b | 2178 | int nr_threads, struct index_entry_offset_table *ieot) |
77ff1127 BP |
2179 | { |
2180 | int i, offset, ieot_blocks, ieot_start, err; | |
2181 | struct load_cache_entries_thread_data *data; | |
2182 | unsigned long consumed = 0; | |
2183 | ||
2184 | /* a little sanity checking */ | |
2185 | if (istate->name_hash_initialized) | |
2186 | BUG("the name hash isn't thread safe"); | |
2187 | ||
44c7e1a7 EN |
2188 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
2189 | mem_pool_init(istate->ce_mem_pool, 0); | |
77ff1127 BP |
2190 | |
2191 | /* ensure we have no more threads than we have blocks to process */ | |
2192 | if (nr_threads > ieot->nr) | |
2193 | nr_threads = ieot->nr; | |
ca56dadb | 2194 | CALLOC_ARRAY(data, nr_threads); |
77ff1127 BP |
2195 | |
2196 | offset = ieot_start = 0; | |
2197 | ieot_blocks = DIV_ROUND_UP(ieot->nr, nr_threads); | |
2198 | for (i = 0; i < nr_threads; i++) { | |
2199 | struct load_cache_entries_thread_data *p = &data[i]; | |
2200 | int nr, j; | |
2201 | ||
2202 | if (ieot_start + ieot_blocks > ieot->nr) | |
2203 | ieot_blocks = ieot->nr - ieot_start; | |
2204 | ||
2205 | p->istate = istate; | |
2206 | p->offset = offset; | |
2207 | p->mmap = mmap; | |
2208 | p->ieot = ieot; | |
2209 | p->ieot_start = ieot_start; | |
2210 | p->ieot_blocks = ieot_blocks; | |
2211 | ||
2212 | /* create a mem_pool for each thread */ | |
2213 | nr = 0; | |
2214 | for (j = p->ieot_start; j < p->ieot_start + p->ieot_blocks; j++) | |
2215 | nr += p->ieot->entries[j].nr; | |
bcd2c5ee | 2216 | p->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2217 | if (istate->version == 4) { |
44c7e1a7 | 2218 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2219 | estimate_cache_size_from_compressed(nr)); |
2220 | } else { | |
44c7e1a7 | 2221 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2222 | estimate_cache_size(mmap_size, nr)); |
2223 | } | |
2224 | ||
2225 | err = pthread_create(&p->pthread, NULL, load_cache_entries_thread, p); | |
2226 | if (err) | |
2227 | die(_("unable to create load_cache_entries thread: %s"), strerror(err)); | |
2228 | ||
2229 | /* increment by the number of cache entries in the ieot block being processed */ | |
2230 | for (j = 0; j < ieot_blocks; j++) | |
2231 | offset += ieot->entries[ieot_start + j].nr; | |
2232 | ieot_start += ieot_blocks; | |
2233 | } | |
2234 | ||
2235 | for (i = 0; i < nr_threads; i++) { | |
2236 | struct load_cache_entries_thread_data *p = &data[i]; | |
2237 | ||
2238 | err = pthread_join(p->pthread, NULL); | |
2239 | if (err) | |
2240 | die(_("unable to join load_cache_entries thread: %s"), strerror(err)); | |
2241 | mem_pool_combine(istate->ce_mem_pool, p->ce_mem_pool); | |
2242 | consumed += p->consumed; | |
2243 | } | |
2244 | ||
2245 | free(data); | |
2246 | ||
2247 | return consumed; | |
2248 | } | |
77ff1127 | 2249 | |
8fd2cb40 | 2250 | /* remember to discard_cache() before reading a different cache! */ |
3e52f70b | 2251 | int do_read_index(struct index_state *istate, const char *path, int must_exist) |
e83c5163 | 2252 | { |
77ff1127 | 2253 | int fd; |
e83c5163 | 2254 | struct stat st; |
debed2a6 | 2255 | unsigned long src_offset; |
371ed0de BP |
2256 | const struct cache_header *hdr; |
2257 | const char *mmap; | |
7a51ed66 | 2258 | size_t mmap_size; |
abb4bb83 BP |
2259 | struct load_index_extensions p; |
2260 | size_t extension_offset = 0; | |
77ff1127 BP |
2261 | int nr_threads, cpus; |
2262 | struct index_entry_offset_table *ieot = NULL; | |
e83c5163 | 2263 | |
913e0e99 | 2264 | if (istate->initialized) |
4aab5b46 | 2265 | return istate->cache_nr; |
5d1a5c02 | 2266 | |
fba2f38a KB |
2267 | istate->timestamp.sec = 0; |
2268 | istate->timestamp.nsec = 0; | |
8fd2cb40 | 2269 | fd = open(path, O_RDONLY); |
5d1a5c02 | 2270 | if (fd < 0) { |
5fc2fc8f | 2271 | if (!must_exist && errno == ENOENT) |
5d1a5c02 | 2272 | return 0; |
9d0a9e90 | 2273 | die_errno(_("%s: index file open failed"), path); |
5d1a5c02 | 2274 | } |
e83c5163 | 2275 | |
3511a377 | 2276 | if (fstat(fd, &st)) |
9d0a9e90 | 2277 | die_errno(_("%s: cannot stat the open index"), path); |
3511a377 | 2278 | |
7a51ed66 | 2279 | mmap_size = xsize_t(st.st_size); |
aab61359 | 2280 | if (mmap_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
9d0a9e90 | 2281 | die(_("%s: index file smaller than expected"), path); |
3511a377 | 2282 | |
02638d1e | 2283 | mmap = xmmap_gently(NULL, mmap_size, PROT_READ, MAP_PRIVATE, fd, 0); |
7a51ed66 | 2284 | if (mmap == MAP_FAILED) |
dc059294 EW |
2285 | die_errno(_("%s: unable to map index file%s"), path, |
2286 | mmap_os_err()); | |
57d84f8d | 2287 | close(fd); |
e83c5163 | 2288 | |
371ed0de | 2289 | hdr = (const struct cache_header *)mmap; |
7a51ed66 | 2290 | if (verify_hdr(hdr, mmap_size) < 0) |
e83c5163 LT |
2291 | goto unmap; |
2292 | ||
92e2cab9 | 2293 | oidread(&istate->oid, (const unsigned char *)hdr + mmap_size - the_hash_algo->rawsz); |
9d227781 | 2294 | istate->version = ntohl(hdr->hdr_version); |
4aab5b46 JH |
2295 | istate->cache_nr = ntohl(hdr->hdr_entries); |
2296 | istate->cache_alloc = alloc_nr(istate->cache_nr); | |
ca56dadb | 2297 | CALLOC_ARRAY(istate->cache, istate->cache_alloc); |
913e0e99 | 2298 | istate->initialized = 1; |
7a51ed66 | 2299 | |
abb4bb83 BP |
2300 | p.istate = istate; |
2301 | p.mmap = mmap; | |
2302 | p.mmap_size = mmap_size; | |
6c9cd161 | 2303 | |
7a51ed66 | 2304 | src_offset = sizeof(*hdr); |
4aab5b46 | 2305 | |
2a9dedef JN |
2306 | if (git_config_get_index_threads(&nr_threads)) |
2307 | nr_threads = 1; | |
7a51ed66 | 2308 | |
77ff1127 BP |
2309 | /* TODO: does creating more threads than cores help? */ |
2310 | if (!nr_threads) { | |
2311 | nr_threads = istate->cache_nr / THREAD_COST; | |
2312 | cpus = online_cpus(); | |
2313 | if (nr_threads > cpus) | |
2314 | nr_threads = cpus; | |
e83c5163 | 2315 | } |
abb4bb83 | 2316 | |
88168b9b NTND |
2317 | if (!HAVE_THREADS) |
2318 | nr_threads = 1; | |
2319 | ||
abb4bb83 BP |
2320 | if (nr_threads > 1) { |
2321 | extension_offset = read_eoie_extension(mmap, mmap_size); | |
2322 | if (extension_offset) { | |
2323 | int err; | |
2324 | ||
2325 | p.src_offset = extension_offset; | |
2326 | err = pthread_create(&p.pthread, NULL, load_index_extensions, &p); | |
2327 | if (err) | |
2328 | die(_("unable to create load_index_extensions thread: %s"), strerror(err)); | |
2329 | ||
2330 | nr_threads--; | |
2331 | } | |
2332 | } | |
abb4bb83 | 2333 | |
77ff1127 BP |
2334 | /* |
2335 | * Locate and read the index entry offset table so that we can use it | |
2336 | * to multi-thread the reading of the cache entries. | |
2337 | */ | |
2338 | if (extension_offset && nr_threads > 1) | |
2339 | ieot = read_ieot_extension(mmap, mmap_size, extension_offset); | |
2340 | ||
2341 | if (ieot) { | |
7bd9631b | 2342 | src_offset += load_cache_entries_threaded(istate, mmap, mmap_size, nr_threads, ieot); |
77ff1127 | 2343 | free(ieot); |
8e72d675 | 2344 | } else { |
77ff1127 | 2345 | src_offset += load_all_cache_entries(istate, mmap, mmap_size, src_offset); |
8e72d675 | 2346 | } |
6c9cd161 | 2347 | |
fba2f38a | 2348 | istate->timestamp.sec = st.st_mtime; |
c06ff490 | 2349 | istate->timestamp.nsec = ST_MTIME_NSEC(st); |
fba2f38a | 2350 | |
abb4bb83 | 2351 | /* if we created a thread, join it otherwise load the extensions on the primary thread */ |
abb4bb83 BP |
2352 | if (extension_offset) { |
2353 | int ret = pthread_join(p.pthread, NULL); | |
2354 | if (ret) | |
2355 | die(_("unable to join load_index_extensions thread: %s"), strerror(ret)); | |
88168b9b | 2356 | } else { |
abb4bb83 BP |
2357 | p.src_offset = src_offset; |
2358 | load_index_extensions(&p); | |
bad68ec9 | 2359 | } |
371ed0de | 2360 | munmap((void *)mmap, mmap_size); |
42fee7a3 JH |
2361 | |
2362 | /* | |
2363 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2364 | * that is associated with the given "istate". | |
2365 | */ | |
2366 | trace2_data_intmax("index", the_repository, "read/version", | |
2367 | istate->version); | |
2368 | trace2_data_intmax("index", the_repository, "read/cache_nr", | |
2369 | istate->cache_nr); | |
2370 | ||
4300f844 DS |
2371 | if (!istate->repo) |
2372 | istate->repo = the_repository; | |
7ca4fc88 VD |
2373 | |
2374 | /* | |
2375 | * If the command explicitly requires a full index, force it | |
2376 | * to be full. Otherwise, correct the sparsity based on repository | |
2377 | * settings and other properties of the index (if necessary). | |
2378 | */ | |
4300f844 DS |
2379 | prepare_repo_settings(istate->repo); |
2380 | if (istate->repo->settings.command_requires_full_index) | |
2381 | ensure_full_index(istate); | |
7ca4fc88 VD |
2382 | else |
2383 | ensure_correct_sparsity(istate); | |
4300f844 | 2384 | |
4aab5b46 | 2385 | return istate->cache_nr; |
e83c5163 LT |
2386 | |
2387 | unmap: | |
371ed0de | 2388 | munmap((void *)mmap, mmap_size); |
9d0a9e90 | 2389 | die(_("index file corrupt")); |
e83c5163 LT |
2390 | } |
2391 | ||
0d59ffb4 CC |
2392 | /* |
2393 | * Signal that the shared index is used by updating its mtime. | |
2394 | * | |
2395 | * This way, shared index can be removed if they have not been used | |
2396 | * for some time. | |
2397 | */ | |
a125a223 | 2398 | static void freshen_shared_index(const char *shared_index, int warn) |
0d59ffb4 | 2399 | { |
0d59ffb4 | 2400 | if (!check_and_freshen_file(shared_index, 1) && warn) |
9d0a9e90 | 2401 | warning(_("could not freshen shared index '%s'"), shared_index); |
0d59ffb4 CC |
2402 | } |
2403 | ||
a125a223 TG |
2404 | int read_index_from(struct index_state *istate, const char *path, |
2405 | const char *gitdir) | |
5fc2fc8f NTND |
2406 | { |
2407 | struct split_index *split_index; | |
2408 | int ret; | |
2182abd9 | 2409 | char *base_oid_hex; |
a125a223 | 2410 | char *base_path; |
5fc2fc8f NTND |
2411 | |
2412 | /* istate->initialized covers both .git/index and .git/sharedindex.xxx */ | |
2413 | if (istate->initialized) | |
2414 | return istate->cache_nr; | |
2415 | ||
42fee7a3 JH |
2416 | /* |
2417 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2418 | * that is associated with the given "istate". | |
2419 | */ | |
2420 | trace2_region_enter_printf("index", "do_read_index", the_repository, | |
2421 | "%s", path); | |
c46c406a | 2422 | trace_performance_enter(); |
5fc2fc8f | 2423 | ret = do_read_index(istate, path, 0); |
c46c406a | 2424 | trace_performance_leave("read cache %s", path); |
42fee7a3 JH |
2425 | trace2_region_leave_printf("index", "do_read_index", the_repository, |
2426 | "%s", path); | |
435ec090 | 2427 | |
5fc2fc8f | 2428 | split_index = istate->split_index; |
2182abd9 | 2429 | if (!split_index || is_null_oid(&split_index->base_oid)) { |
435ec090 | 2430 | post_read_index_from(istate); |
5fc2fc8f | 2431 | return ret; |
03f15a79 | 2432 | } |
5fc2fc8f | 2433 | |
c46c406a | 2434 | trace_performance_enter(); |
5fc2fc8f NTND |
2435 | if (split_index->base) |
2436 | discard_index(split_index->base); | |
2437 | else | |
ca56dadb | 2438 | CALLOC_ARRAY(split_index->base, 1); |
de6ae5f9 | 2439 | |
2182abd9 | 2440 | base_oid_hex = oid_to_hex(&split_index->base_oid); |
2441 | base_path = xstrfmt("%s/sharedindex.%s", gitdir, base_oid_hex); | |
42fee7a3 JH |
2442 | trace2_region_enter_printf("index", "shared/do_read_index", |
2443 | the_repository, "%s", base_path); | |
998330ac | 2444 | ret = do_read_index(split_index->base, base_path, 0); |
42fee7a3 JH |
2445 | trace2_region_leave_printf("index", "shared/do_read_index", |
2446 | the_repository, "%s", base_path); | |
998330ac SG |
2447 | if (!ret) { |
2448 | char *path_copy = xstrdup(path); | |
2449 | const char *base_path2 = xstrfmt("%s/sharedindex.%s", | |
2450 | dirname(path_copy), | |
2451 | base_oid_hex); | |
2452 | free(path_copy); | |
2453 | trace2_region_enter_printf("index", "shared/do_read_index", | |
2454 | the_repository, "%s", base_path2); | |
2455 | ret = do_read_index(split_index->base, base_path2, 1); | |
2456 | trace2_region_leave_printf("index", "shared/do_read_index", | |
2457 | the_repository, "%s", base_path2); | |
2458 | } | |
9001dc2a | 2459 | if (!oideq(&split_index->base_oid, &split_index->base->oid)) |
9d0a9e90 | 2460 | die(_("broken index, expect %s in %s, got %s"), |
2182abd9 | 2461 | base_oid_hex, base_path, |
75691ea3 | 2462 | oid_to_hex(&split_index->base->oid)); |
de6ae5f9 | 2463 | |
a125a223 | 2464 | freshen_shared_index(base_path, 0); |
5fc2fc8f | 2465 | merge_base_index(istate); |
435ec090 | 2466 | post_read_index_from(istate); |
c46c406a | 2467 | trace_performance_leave("read cache %s", base_path); |
b42ad7d5 | 2468 | free(base_path); |
5fc2fc8f NTND |
2469 | return ret; |
2470 | } | |
2471 | ||
fa7b3c2f JH |
2472 | int is_index_unborn(struct index_state *istate) |
2473 | { | |
debed2a6 | 2474 | return (!istate->cache_nr && !istate->timestamp.sec); |
fa7b3c2f JH |
2475 | } |
2476 | ||
4aab5b46 | 2477 | int discard_index(struct index_state *istate) |
6d297f81 | 2478 | { |
8e72d675 JM |
2479 | /* |
2480 | * Cache entries in istate->cache[] should have been allocated | |
2481 | * from the memory pool associated with this index, or from an | |
2482 | * associated split_index. There is no need to free individual | |
8616a2d0 JM |
2483 | * cache entries. validate_cache_entries can detect when this |
2484 | * assertion does not hold. | |
8e72d675 | 2485 | */ |
8616a2d0 | 2486 | validate_cache_entries(istate); |
debed2a6 | 2487 | |
cfc5789a | 2488 | resolve_undo_clear_index(istate); |
4aab5b46 JH |
2489 | istate->cache_nr = 0; |
2490 | istate->cache_changed = 0; | |
fba2f38a KB |
2491 | istate->timestamp.sec = 0; |
2492 | istate->timestamp.nsec = 0; | |
2092678c | 2493 | free_name_hash(istate); |
4aab5b46 | 2494 | cache_tree_free(&(istate->cache_tree)); |
913e0e99 | 2495 | istate->initialized = 0; |
398a3b08 | 2496 | istate->fsmonitor_has_run_once = 0; |
4abc5784 | 2497 | FREE_AND_NULL(istate->fsmonitor_last_update); |
6a83d902 | 2498 | FREE_AND_NULL(istate->cache); |
a0fc4db0 | 2499 | istate->cache_alloc = 0; |
5fc2fc8f | 2500 | discard_split_index(istate); |
f9e6c649 NTND |
2501 | free_untracked_cache(istate->untracked); |
2502 | istate->untracked = NULL; | |
8e72d675 JM |
2503 | |
2504 | if (istate->ce_mem_pool) { | |
8616a2d0 | 2505 | mem_pool_discard(istate->ce_mem_pool, should_validate_cache_entries()); |
44c7e1a7 | 2506 | FREE_AND_NULL(istate->ce_mem_pool); |
8e72d675 JM |
2507 | } |
2508 | ||
7a51ed66 | 2509 | return 0; |
6d297f81 JS |
2510 | } |
2511 | ||
8616a2d0 JM |
2512 | /* |
2513 | * Validate the cache entries of this index. | |
2514 | * All cache entries associated with this index | |
2515 | * should have been allocated by the memory pool | |
2516 | * associated with this index, or by a referenced | |
2517 | * split index. | |
2518 | */ | |
2519 | void validate_cache_entries(const struct index_state *istate) | |
2520 | { | |
2521 | int i; | |
2522 | ||
2523 | if (!should_validate_cache_entries() ||!istate || !istate->initialized) | |
2524 | return; | |
2525 | ||
2526 | for (i = 0; i < istate->cache_nr; i++) { | |
2527 | if (!istate) { | |
391408e5 | 2528 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2529 | } else if (!istate->ce_mem_pool || |
2530 | !mem_pool_contains(istate->ce_mem_pool, istate->cache[i])) { | |
2531 | if (!istate->split_index || | |
2532 | !istate->split_index->base || | |
2533 | !istate->split_index->base->ce_mem_pool || | |
2534 | !mem_pool_contains(istate->split_index->base->ce_mem_pool, istate->cache[i])) { | |
391408e5 | 2535 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2536 | } |
2537 | } | |
2538 | } | |
2539 | ||
2540 | if (istate->split_index) | |
2541 | validate_cache_entries(istate->split_index->base); | |
2542 | } | |
2543 | ||
d1f128b0 | 2544 | int unmerged_index(const struct index_state *istate) |
94a5728c DB |
2545 | { |
2546 | int i; | |
2547 | for (i = 0; i < istate->cache_nr; i++) { | |
2548 | if (ce_stage(istate->cache[i])) | |
2549 | return 1; | |
2550 | } | |
2551 | return 0; | |
2552 | } | |
2553 | ||
150fe065 NTND |
2554 | int repo_index_has_changes(struct repository *repo, |
2555 | struct tree *tree, | |
2556 | struct strbuf *sb) | |
cffbfad5 | 2557 | { |
150fe065 | 2558 | struct index_state *istate = repo->index; |
e1f8694f | 2559 | struct object_id cmp; |
cffbfad5 EN |
2560 | int i; |
2561 | ||
e1f8694f EN |
2562 | if (tree) |
2563 | cmp = tree->object.oid; | |
2564 | if (tree || !get_oid_tree("HEAD", &cmp)) { | |
cffbfad5 EN |
2565 | struct diff_options opt; |
2566 | ||
150fe065 | 2567 | repo_diff_setup(repo, &opt); |
cffbfad5 EN |
2568 | opt.flags.exit_with_status = 1; |
2569 | if (!sb) | |
2570 | opt.flags.quick = 1; | |
8d833e93 | 2571 | diff_setup_done(&opt); |
e1f8694f | 2572 | do_diff_cache(&cmp, &opt); |
cffbfad5 EN |
2573 | diffcore_std(&opt); |
2574 | for (i = 0; sb && i < diff_queued_diff.nr; i++) { | |
2575 | if (i) | |
2576 | strbuf_addch(sb, ' '); | |
2577 | strbuf_addstr(sb, diff_queued_diff.queue[i]->two->path); | |
2578 | } | |
2579 | diff_flush(&opt); | |
2580 | return opt.flags.has_changes != 0; | |
2581 | } else { | |
0c18c059 DS |
2582 | /* TODO: audit for interaction with sparse-index. */ |
2583 | ensure_full_index(istate); | |
1b9fbefb | 2584 | for (i = 0; sb && i < istate->cache_nr; i++) { |
cffbfad5 EN |
2585 | if (i) |
2586 | strbuf_addch(sb, ' '); | |
1b9fbefb | 2587 | strbuf_addstr(sb, istate->cache[i]->name); |
cffbfad5 | 2588 | } |
1b9fbefb | 2589 | return !!istate->cache_nr; |
cffbfad5 EN |
2590 | } |
2591 | } | |
2592 | ||
410334ed DS |
2593 | static int write_index_ext_header(struct hashfile *f, |
2594 | git_hash_ctx *eoie_f, | |
2595 | unsigned int ext, | |
2596 | unsigned int sz) | |
bad68ec9 | 2597 | { |
410334ed DS |
2598 | hashwrite_be32(f, ext); |
2599 | hashwrite_be32(f, sz); | |
2600 | ||
2601 | if (eoie_f) { | |
2602 | ext = htonl(ext); | |
2603 | sz = htonl(sz); | |
2604 | the_hash_algo->update_fn(eoie_f, &ext, sizeof(ext)); | |
2605 | the_hash_algo->update_fn(eoie_f, &sz, sizeof(sz)); | |
3b1d9e04 | 2606 | } |
410334ed | 2607 | return 0; |
bad68ec9 JH |
2608 | } |
2609 | ||
58bf2a4c NTND |
2610 | static void ce_smudge_racily_clean_entry(struct index_state *istate, |
2611 | struct cache_entry *ce) | |
407c8eb0 JH |
2612 | { |
2613 | /* | |
2614 | * The only thing we care about in this function is to smudge the | |
2615 | * falsely clean entry due to touch-update-touch race, so we leave | |
2616 | * everything else as they are. We are called for entries whose | |
c21d39d7 | 2617 | * ce_stat_data.sd_mtime match the index file mtime. |
c70115b4 JH |
2618 | * |
2619 | * Note that this actually does not do much for gitlinks, for | |
2620 | * which ce_match_stat_basic() always goes to the actual | |
2621 | * contents. The caller checks with is_racy_timestamp() which | |
2622 | * always says "no" for gitlinks, so we are not called for them ;-) | |
407c8eb0 JH |
2623 | */ |
2624 | struct stat st; | |
2625 | ||
2626 | if (lstat(ce->name, &st) < 0) | |
2627 | return; | |
2628 | if (ce_match_stat_basic(ce, &st)) | |
2629 | return; | |
58bf2a4c | 2630 | if (ce_modified_check_fs(istate, ce, &st)) { |
4b3511b0 JH |
2631 | /* This is "racily clean"; smudge it. Note that this |
2632 | * is a tricky code. At first glance, it may appear | |
2633 | * that it can break with this sequence: | |
2634 | * | |
2635 | * $ echo xyzzy >frotz | |
2636 | * $ git-update-index --add frotz | |
2637 | * $ : >frotz | |
2638 | * $ sleep 3 | |
2639 | * $ echo filfre >nitfol | |
2640 | * $ git-update-index --add nitfol | |
2641 | * | |
b7e58b17 | 2642 | * but it does not. When the second update-index runs, |
4b3511b0 JH |
2643 | * it notices that the entry "frotz" has the same timestamp |
2644 | * as index, and if we were to smudge it by resetting its | |
2645 | * size to zero here, then the object name recorded | |
2646 | * in index is the 6-byte file but the cached stat information | |
2647 | * becomes zero --- which would then match what we would | |
a6080a0a | 2648 | * obtain from the filesystem next time we stat("frotz"). |
4b3511b0 JH |
2649 | * |
2650 | * However, the second update-index, before calling | |
2651 | * this function, notices that the cached size is 6 | |
2652 | * bytes and what is on the filesystem is an empty | |
2653 | * file, and never calls us, so the cached size information | |
2654 | * for "frotz" stays 6 which does not match the filesystem. | |
2655 | */ | |
c21d39d7 | 2656 | ce->ce_stat_data.sd_size = 0; |
407c8eb0 JH |
2657 | } |
2658 | } | |
2659 | ||
f136f7bf | 2660 | /* Copy miscellaneous fields but not the name */ |
ce012deb | 2661 | static void copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk, |
f136f7bf | 2662 | struct cache_entry *ce) |
7a51ed66 | 2663 | { |
b60e188c | 2664 | short flags; |
575fa8a3 | 2665 | const unsigned hashsz = the_hash_algo->rawsz; |
2666 | uint16_t *flagsp = (uint16_t *)(ondisk->data + hashsz); | |
b60e188c | 2667 | |
c21d39d7 MH |
2668 | ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec); |
2669 | ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec); | |
2670 | ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec); | |
2671 | ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec); | |
2672 | ondisk->dev = htonl(ce->ce_stat_data.sd_dev); | |
2673 | ondisk->ino = htonl(ce->ce_stat_data.sd_ino); | |
7a51ed66 | 2674 | ondisk->mode = htonl(ce->ce_mode); |
c21d39d7 MH |
2675 | ondisk->uid = htonl(ce->ce_stat_data.sd_uid); |
2676 | ondisk->gid = htonl(ce->ce_stat_data.sd_gid); | |
2677 | ondisk->size = htonl(ce->ce_stat_data.sd_size); | |
575fa8a3 | 2678 | hashcpy(ondisk->data, ce->oid.hash); |
b60e188c | 2679 | |
ce51bf09 | 2680 | flags = ce->ce_flags & ~CE_NAMEMASK; |
b60e188c | 2681 | flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce)); |
575fa8a3 | 2682 | flagsp[0] = htons(flags); |
06aaaa0b | 2683 | if (ce->ce_flags & CE_EXTENDED) { |
575fa8a3 | 2684 | flagsp[1] = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16); |
f136f7bf JH |
2685 | } |
2686 | } | |
2687 | ||
410334ed | 2688 | static int ce_write_entry(struct hashfile *f, struct cache_entry *ce, |
ce012deb | 2689 | struct strbuf *previous_name, struct ondisk_cache_entry *ondisk) |
f136f7bf | 2690 | { |
9d227781 | 2691 | int size; |
00a4b035 RJ |
2692 | unsigned int saved_namelen; |
2693 | int stripped_name = 0; | |
ce012deb | 2694 | static unsigned char padding[8] = { 0x00 }; |
f136f7bf | 2695 | |
b3c96fb1 NTND |
2696 | if (ce->ce_flags & CE_STRIP_NAME) { |
2697 | saved_namelen = ce_namelen(ce); | |
2698 | ce->ce_namelen = 0; | |
00a4b035 | 2699 | stripped_name = 1; |
b3c96fb1 NTND |
2700 | } |
2701 | ||
575fa8a3 | 2702 | size = offsetof(struct ondisk_cache_entry,data) + ondisk_data_size(ce->ce_flags, 0); |
ce012deb | 2703 | |
9d227781 | 2704 | if (!previous_name) { |
ce012deb KW |
2705 | int len = ce_namelen(ce); |
2706 | copy_cache_entry_to_ondisk(ondisk, ce); | |
410334ed DS |
2707 | hashwrite(f, ondisk, size); |
2708 | hashwrite(f, ce->name, len); | |
2709 | hashwrite(f, padding, align_padding_size(size, len)); | |
9d227781 JH |
2710 | } else { |
2711 | int common, to_remove, prefix_size; | |
2712 | unsigned char to_remove_vi[16]; | |
2713 | for (common = 0; | |
2714 | (ce->name[common] && | |
2715 | common < previous_name->len && | |
2716 | ce->name[common] == previous_name->buf[common]); | |
2717 | common++) | |
2718 | ; /* still matching */ | |
2719 | to_remove = previous_name->len - common; | |
2720 | prefix_size = encode_varint(to_remove, to_remove_vi); | |
2721 | ||
ce012deb | 2722 | copy_cache_entry_to_ondisk(ondisk, ce); |
410334ed DS |
2723 | hashwrite(f, ondisk, size); |
2724 | hashwrite(f, to_remove_vi, prefix_size); | |
2725 | hashwrite(f, ce->name + common, ce_namelen(ce) - common); | |
2726 | hashwrite(f, padding, 1); | |
9d227781 JH |
2727 | |
2728 | strbuf_splice(previous_name, common, to_remove, | |
2729 | ce->name + common, ce_namelen(ce) - common); | |
06aaaa0b | 2730 | } |
00a4b035 | 2731 | if (stripped_name) { |
b3c96fb1 NTND |
2732 | ce->ce_namelen = saved_namelen; |
2733 | ce->ce_flags &= ~CE_STRIP_NAME; | |
2734 | } | |
7a51ed66 | 2735 | |
410334ed | 2736 | return 0; |
7a51ed66 LT |
2737 | } |
2738 | ||
426ddeea YM |
2739 | /* |
2740 | * This function verifies if index_state has the correct sha1 of the | |
2741 | * index file. Don't die if we have any other failure, just return 0. | |
2742 | */ | |
2743 | static int verify_index_from(const struct index_state *istate, const char *path) | |
2744 | { | |
2745 | int fd; | |
2746 | ssize_t n; | |
2747 | struct stat st; | |
aab61359 | 2748 | unsigned char hash[GIT_MAX_RAWSZ]; |
426ddeea YM |
2749 | |
2750 | if (!istate->initialized) | |
2751 | return 0; | |
2752 | ||
2753 | fd = open(path, O_RDONLY); | |
2754 | if (fd < 0) | |
2755 | return 0; | |
2756 | ||
2757 | if (fstat(fd, &st)) | |
2758 | goto out; | |
2759 | ||
aab61359 | 2760 | if (st.st_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
426ddeea YM |
2761 | goto out; |
2762 | ||
aab61359 | 2763 | n = pread_in_full(fd, hash, the_hash_algo->rawsz, st.st_size - the_hash_algo->rawsz); |
2764 | if (n != the_hash_algo->rawsz) | |
426ddeea YM |
2765 | goto out; |
2766 | ||
67947c34 | 2767 | if (!hasheq(istate->oid.hash, hash)) |
426ddeea YM |
2768 | goto out; |
2769 | ||
2770 | close(fd); | |
2771 | return 1; | |
2772 | ||
2773 | out: | |
2774 | close(fd); | |
2775 | return 0; | |
2776 | } | |
2777 | ||
1b0d968b | 2778 | static int repo_verify_index(struct repository *repo) |
426ddeea | 2779 | { |
1b0d968b | 2780 | return verify_index_from(repo->index, repo->index_file); |
426ddeea YM |
2781 | } |
2782 | ||
2ede073f | 2783 | int has_racy_timestamp(struct index_state *istate) |
483fbe2b JH |
2784 | { |
2785 | int entries = istate->cache_nr; | |
2786 | int i; | |
2787 | ||
2788 | for (i = 0; i < entries; i++) { | |
2789 | struct cache_entry *ce = istate->cache[i]; | |
2790 | if (is_racy_timestamp(istate, ce)) | |
2791 | return 1; | |
2792 | } | |
2793 | return 0; | |
2794 | } | |
2795 | ||
1b0d968b NTND |
2796 | void repo_update_index_if_able(struct repository *repo, |
2797 | struct lock_file *lockfile) | |
ccdc4ec3 | 2798 | { |
1b0d968b NTND |
2799 | if ((repo->index->cache_changed || |
2800 | has_racy_timestamp(repo->index)) && | |
2801 | repo_verify_index(repo)) | |
2802 | write_locked_index(repo->index, lockfile, COMMIT_LOCK); | |
b74c90fb MÅ |
2803 | else |
2804 | rollback_lock_file(lockfile); | |
ccdc4ec3 JH |
2805 | } |
2806 | ||
d8465500 JN |
2807 | static int record_eoie(void) |
2808 | { | |
2809 | int val; | |
2810 | ||
2811 | if (!git_config_get_bool("index.recordendofindexentries", &val)) | |
2812 | return val; | |
2a9dedef JN |
2813 | |
2814 | /* | |
2815 | * As a convenience, the end of index entries extension | |
2816 | * used for threading is written by default if the user | |
2817 | * explicitly requested threaded index reads. | |
2818 | */ | |
2819 | return !git_config_get_index_threads(&val) && val != 1; | |
d8465500 JN |
2820 | } |
2821 | ||
42916054 JN |
2822 | static int record_ieot(void) |
2823 | { | |
2824 | int val; | |
2825 | ||
2826 | if (!git_config_get_bool("index.recordoffsettable", &val)) | |
2827 | return val; | |
2a9dedef JN |
2828 | |
2829 | /* | |
2830 | * As a convenience, the offset table used for threading is | |
2831 | * written by default if the user explicitly requested | |
2832 | * threaded index reads. | |
2833 | */ | |
2834 | return !git_config_get_index_threads(&val) && val != 1; | |
42916054 JN |
2835 | } |
2836 | ||
812d6b00 MÅ |
2837 | /* |
2838 | * On success, `tempfile` is closed. If it is the temporary file | |
2839 | * of a `struct lock_file`, we will therefore effectively perform | |
2840 | * a 'close_lock_file_gently()`. Since that is an implementation | |
2841 | * detail of lockfiles, callers of `do_write_index()` should not | |
2842 | * rely on it. | |
2843 | */ | |
9f41c7a6 | 2844 | static int do_write_index(struct index_state *istate, struct tempfile *tempfile, |
c18b80a0 | 2845 | int strip_extensions) |
197ee8c9 | 2846 | { |
ca54d9ba | 2847 | uint64_t start = getnanotime(); |
410334ed DS |
2848 | struct hashfile *f; |
2849 | git_hash_ctx *eoie_c = NULL; | |
197ee8c9 | 2850 | struct cache_header hdr; |
b50386c7 | 2851 | int i, err = 0, removed, extended, hdr_version; |
4aab5b46 JH |
2852 | struct cache_entry **cache = istate->cache; |
2853 | int entries = istate->cache_nr; | |
e1afca4f | 2854 | struct stat st; |
575fa8a3 | 2855 | struct ondisk_cache_entry ondisk; |
9d227781 | 2856 | struct strbuf previous_name_buf = STRBUF_INIT, *previous_name; |
4bddd983 | 2857 | int drop_cache_tree = istate->drop_cache_tree; |
3b1d9e04 | 2858 | off_t offset; |
77ff1127 | 2859 | int ieot_entries = 1; |
3255089a BP |
2860 | struct index_entry_offset_table *ieot = NULL; |
2861 | int nr, nr_threads; | |
025a0709 | 2862 | |
410334ed DS |
2863 | f = hashfd(tempfile->fd, tempfile->filename.buf); |
2864 | ||
06aaaa0b | 2865 | for (i = removed = extended = 0; i < entries; i++) { |
7a51ed66 | 2866 | if (cache[i]->ce_flags & CE_REMOVE) |
025a0709 | 2867 | removed++; |
197ee8c9 | 2868 | |
06aaaa0b NTND |
2869 | /* reduce extended entries if possible */ |
2870 | cache[i]->ce_flags &= ~CE_EXTENDED; | |
2871 | if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) { | |
2872 | extended++; | |
2873 | cache[i]->ce_flags |= CE_EXTENDED; | |
2874 | } | |
2875 | } | |
2876 | ||
e8ffd034 | 2877 | if (!istate->version) |
7211b9e7 | 2878 | istate->version = get_index_format_default(the_repository); |
9d227781 JH |
2879 | |
2880 | /* demote version 3 to version 2 when the latter suffices */ | |
2881 | if (istate->version == 3 || istate->version == 2) | |
2882 | istate->version = extended ? 3 : 2; | |
2883 | ||
2884 | hdr_version = istate->version; | |
2885 | ||
ccc4feb5 | 2886 | hdr.hdr_signature = htonl(CACHE_SIGNATURE); |
9d227781 | 2887 | hdr.hdr_version = htonl(hdr_version); |
025a0709 | 2888 | hdr.hdr_entries = htonl(entries - removed); |
197ee8c9 | 2889 | |
410334ed | 2890 | hashwrite(f, &hdr, sizeof(hdr)); |
197ee8c9 | 2891 | |
2a9dedef | 2892 | if (!HAVE_THREADS || git_config_get_index_threads(&nr_threads)) |
88168b9b | 2893 | nr_threads = 1; |
62e5ee81 | 2894 | |
42916054 | 2895 | if (nr_threads != 1 && record_ieot()) { |
3255089a BP |
2896 | int ieot_blocks, cpus; |
2897 | ||
2898 | /* | |
2899 | * ensure default number of ieot blocks maps evenly to the | |
2900 | * default number of threads that will process them leaving | |
2901 | * room for the thread to load the index extensions. | |
2902 | */ | |
2903 | if (!nr_threads) { | |
2904 | ieot_blocks = istate->cache_nr / THREAD_COST; | |
2905 | cpus = online_cpus(); | |
2906 | if (ieot_blocks > cpus - 1) | |
2907 | ieot_blocks = cpus - 1; | |
2908 | } else { | |
2909 | ieot_blocks = nr_threads; | |
77ff1127 BP |
2910 | if (ieot_blocks > istate->cache_nr) |
2911 | ieot_blocks = istate->cache_nr; | |
3255089a BP |
2912 | } |
2913 | ||
2914 | /* | |
2915 | * no reason to write out the IEOT extension if we don't | |
2916 | * have enough blocks to utilize multi-threading | |
2917 | */ | |
2918 | if (ieot_blocks > 1) { | |
2919 | ieot = xcalloc(1, sizeof(struct index_entry_offset_table) | |
2920 | + (ieot_blocks * sizeof(struct index_entry_offset))); | |
77ff1127 | 2921 | ieot_entries = DIV_ROUND_UP(entries, ieot_blocks); |
3255089a BP |
2922 | } |
2923 | } | |
3255089a | 2924 | |
410334ed DS |
2925 | offset = hashfile_total(f); |
2926 | ||
3255089a | 2927 | nr = 0; |
9d227781 | 2928 | previous_name = (hdr_version == 4) ? &previous_name_buf : NULL; |
ce012deb | 2929 | |
197ee8c9 LT |
2930 | for (i = 0; i < entries; i++) { |
2931 | struct cache_entry *ce = cache[i]; | |
7a51ed66 | 2932 | if (ce->ce_flags & CE_REMOVE) |
aa16021e | 2933 | continue; |
e06c43c7 | 2934 | if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce)) |
58bf2a4c | 2935 | ce_smudge_racily_clean_entry(istate, ce); |
99d1a986 | 2936 | if (is_null_oid(&ce->oid)) { |
83bd7437 JK |
2937 | static const char msg[] = "cache entry has null sha1: %s"; |
2938 | static int allow = -1; | |
2939 | ||
2940 | if (allow < 0) | |
2941 | allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0); | |
2942 | if (allow) | |
2943 | warning(msg, ce->name); | |
2944 | else | |
b50386c7 | 2945 | err = error(msg, ce->name); |
a96d3cc3 JK |
2946 | |
2947 | drop_cache_tree = 1; | |
83bd7437 | 2948 | } |
77ff1127 | 2949 | if (ieot && i && (i % ieot_entries == 0)) { |
3255089a BP |
2950 | ieot->entries[ieot->nr].nr = nr; |
2951 | ieot->entries[ieot->nr].offset = offset; | |
2952 | ieot->nr++; | |
2953 | /* | |
2954 | * If we have a V4 index, set the first byte to an invalid | |
2955 | * character to ensure there is nothing common with the previous | |
2956 | * entry | |
2957 | */ | |
2958 | if (previous_name) | |
2959 | previous_name->buf[0] = 0; | |
2960 | nr = 0; | |
410334ed DS |
2961 | |
2962 | offset = hashfile_total(f); | |
3255089a | 2963 | } |
410334ed | 2964 | if (ce_write_entry(f, ce, previous_name, (struct ondisk_cache_entry *)&ondisk) < 0) |
b50386c7 KW |
2965 | err = -1; |
2966 | ||
2967 | if (err) | |
2968 | break; | |
3255089a BP |
2969 | nr++; |
2970 | } | |
2971 | if (ieot && nr) { | |
2972 | ieot->entries[ieot->nr].nr = nr; | |
2973 | ieot->entries[ieot->nr].offset = offset; | |
2974 | ieot->nr++; | |
197ee8c9 | 2975 | } |
9d227781 | 2976 | strbuf_release(&previous_name_buf); |
1af1c2b6 | 2977 | |
3255089a BP |
2978 | if (err) { |
2979 | free(ieot); | |
b50386c7 | 2980 | return err; |
3255089a | 2981 | } |
b50386c7 | 2982 | |
410334ed DS |
2983 | offset = hashfile_total(f); |
2984 | ||
2985 | /* | |
2986 | * The extension headers must be hashed on their own for the | |
2987 | * EOIE extension. Create a hashfile here to compute that hash. | |
2988 | */ | |
2989 | if (offset && record_eoie()) { | |
2990 | CALLOC_ARRAY(eoie_c, 1); | |
2991 | the_hash_algo->init_fn(eoie_c); | |
3255089a | 2992 | } |
3b1d9e04 | 2993 | |
3255089a BP |
2994 | /* |
2995 | * Lets write out CACHE_EXT_INDEXENTRYOFFSETTABLE first so that we | |
2996 | * can minimize the number of extensions we have to scan through to | |
2997 | * find it during load. Write it out regardless of the | |
2998 | * strip_extensions parameter as we need it when loading the shared | |
2999 | * index. | |
3000 | */ | |
3255089a BP |
3001 | if (ieot) { |
3002 | struct strbuf sb = STRBUF_INIT; | |
3003 | ||
3004 | write_ieot_extension(&sb, ieot); | |
410334ed DS |
3005 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_INDEXENTRYOFFSETTABLE, sb.len) < 0; |
3006 | hashwrite(f, sb.buf, sb.len); | |
3255089a BP |
3007 | strbuf_release(&sb); |
3008 | free(ieot); | |
3009 | if (err) | |
3010 | return -1; | |
3011 | } | |
3255089a | 3012 | |
6e37c8ed NTND |
3013 | if (!strip_extensions && istate->split_index && |
3014 | !is_null_oid(&istate->split_index->base_oid)) { | |
5fc2fc8f NTND |
3015 | struct strbuf sb = STRBUF_INIT; |
3016 | ||
451b66c5 JS |
3017 | if (istate->sparse_index) |
3018 | die(_("cannot write split index for a sparse index")); | |
3019 | ||
5fc2fc8f | 3020 | err = write_link_extension(&sb, istate) < 0 || |
410334ed DS |
3021 | write_index_ext_header(f, eoie_c, CACHE_EXT_LINK, |
3022 | sb.len) < 0; | |
3023 | hashwrite(f, sb.buf, sb.len); | |
5fc2fc8f NTND |
3024 | strbuf_release(&sb); |
3025 | if (err) | |
3026 | return -1; | |
3027 | } | |
a96d3cc3 | 3028 | if (!strip_extensions && !drop_cache_tree && istate->cache_tree) { |
f285a2d7 | 3029 | struct strbuf sb = STRBUF_INIT; |
1dffb8fa | 3030 | |
1dffb8fa | 3031 | cache_tree_write(&sb, istate->cache_tree); |
410334ed DS |
3032 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_TREE, sb.len) < 0; |
3033 | hashwrite(f, sb.buf, sb.len); | |
1dffb8fa PH |
3034 | strbuf_release(&sb); |
3035 | if (err) | |
bad68ec9 | 3036 | return -1; |
bad68ec9 | 3037 | } |
c18b80a0 | 3038 | if (!strip_extensions && istate->resolve_undo) { |
cfc5789a JH |
3039 | struct strbuf sb = STRBUF_INIT; |
3040 | ||
3041 | resolve_undo_write(&sb, istate->resolve_undo); | |
410334ed DS |
3042 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_RESOLVE_UNDO, |
3043 | sb.len) < 0; | |
3044 | hashwrite(f, sb.buf, sb.len); | |
cfc5789a JH |
3045 | strbuf_release(&sb); |
3046 | if (err) | |
3047 | return -1; | |
3048 | } | |
83c094ad NTND |
3049 | if (!strip_extensions && istate->untracked) { |
3050 | struct strbuf sb = STRBUF_INIT; | |
3051 | ||
3052 | write_untracked_extension(&sb, istate->untracked); | |
410334ed DS |
3053 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_UNTRACKED, |
3054 | sb.len) < 0; | |
3055 | hashwrite(f, sb.buf, sb.len); | |
83c094ad NTND |
3056 | strbuf_release(&sb); |
3057 | if (err) | |
3058 | return -1; | |
3059 | } | |
883e248b BP |
3060 | if (!strip_extensions && istate->fsmonitor_last_update) { |
3061 | struct strbuf sb = STRBUF_INIT; | |
3062 | ||
3063 | write_fsmonitor_extension(&sb, istate); | |
410334ed DS |
3064 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_FSMONITOR, sb.len) < 0; |
3065 | hashwrite(f, sb.buf, sb.len); | |
3b1d9e04 BP |
3066 | strbuf_release(&sb); |
3067 | if (err) | |
3068 | return -1; | |
3069 | } | |
cd42415f | 3070 | if (istate->sparse_index) { |
410334ed | 3071 | if (write_index_ext_header(f, eoie_c, CACHE_EXT_SPARSE_DIRECTORIES, 0) < 0) |
cd42415f DS |
3072 | return -1; |
3073 | } | |
3b1d9e04 BP |
3074 | |
3075 | /* | |
3076 | * CACHE_EXT_ENDOFINDEXENTRIES must be written as the last entry before the SHA1 | |
3077 | * so that it can be found and processed before all the index entries are | |
3078 | * read. Write it out regardless of the strip_extensions parameter as we need it | |
3079 | * when loading the shared index. | |
3080 | */ | |
410334ed | 3081 | if (eoie_c) { |
3b1d9e04 BP |
3082 | struct strbuf sb = STRBUF_INIT; |
3083 | ||
410334ed DS |
3084 | write_eoie_extension(&sb, eoie_c, offset); |
3085 | err = write_index_ext_header(f, NULL, CACHE_EXT_ENDOFINDEXENTRIES, sb.len) < 0; | |
3086 | hashwrite(f, sb.buf, sb.len); | |
883e248b BP |
3087 | strbuf_release(&sb); |
3088 | if (err) | |
3089 | return -1; | |
3090 | } | |
e1afca4f | 3091 | |
410334ed | 3092 | finalize_hashfile(f, istate->oid.hash, CSUM_HASH_IN_STREAM); |
49bd0fc2 | 3093 | if (close_tempfile_gently(tempfile)) { |
6a8c89d0 | 3094 | error(_("could not close '%s'"), get_tempfile_path(tempfile)); |
49bd0fc2 JK |
3095 | return -1; |
3096 | } | |
6a8c89d0 | 3097 | if (stat(get_tempfile_path(tempfile), &st)) |
e1afca4f | 3098 | return -1; |
5bcf109c KB |
3099 | istate->timestamp.sec = (unsigned int)st.st_mtime; |
3100 | istate->timestamp.nsec = ST_MTIME_NSEC(st); | |
ca54d9ba | 3101 | trace_performance_since(start, "write index, changed mask = %x", istate->cache_changed); |
42fee7a3 JH |
3102 | |
3103 | /* | |
3104 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3105 | * that is associated with the given "istate". | |
3106 | */ | |
3107 | trace2_data_intmax("index", the_repository, "write/version", | |
3108 | istate->version); | |
3109 | trace2_data_intmax("index", the_repository, "write/cache_nr", | |
3110 | istate->cache_nr); | |
3111 | ||
e1afca4f | 3112 | return 0; |
197ee8c9 | 3113 | } |
e46bbcf6 | 3114 | |
626f35c8 NTND |
3115 | void set_alternate_index_output(const char *name) |
3116 | { | |
3117 | alternate_index_output = name; | |
3118 | } | |
3119 | ||
3120 | static int commit_locked_index(struct lock_file *lk) | |
3121 | { | |
751baced MH |
3122 | if (alternate_index_output) |
3123 | return commit_lock_file_to(lk, alternate_index_output); | |
3124 | else | |
626f35c8 | 3125 | return commit_lock_file(lk); |
626f35c8 NTND |
3126 | } |
3127 | ||
03b86647 NTND |
3128 | static int do_write_locked_index(struct index_state *istate, struct lock_file *lock, |
3129 | unsigned flags) | |
3130 | { | |
42fee7a3 | 3131 | int ret; |
6e773527 DS |
3132 | int was_full = !istate->sparse_index; |
3133 | ||
ce7a9f01 | 3134 | ret = convert_to_sparse(istate, 0); |
6e773527 DS |
3135 | |
3136 | if (ret) { | |
3137 | warning(_("failed to convert to a sparse-index")); | |
3138 | return ret; | |
3139 | } | |
42fee7a3 JH |
3140 | |
3141 | /* | |
3142 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3143 | * that is associated with the given "istate". | |
3144 | */ | |
3145 | trace2_region_enter_printf("index", "do_write_index", the_repository, | |
6a8c89d0 | 3146 | "%s", get_lock_file_path(lock)); |
42fee7a3 JH |
3147 | ret = do_write_index(istate, lock->tempfile, 0); |
3148 | trace2_region_leave_printf("index", "do_write_index", the_repository, | |
6a8c89d0 | 3149 | "%s", get_lock_file_path(lock)); |
42fee7a3 | 3150 | |
6e773527 DS |
3151 | if (was_full) |
3152 | ensure_full_index(istate); | |
3153 | ||
03b86647 NTND |
3154 | if (ret) |
3155 | return ret; | |
03b86647 | 3156 | if (flags & COMMIT_LOCK) |
1956ecd0 BP |
3157 | ret = commit_locked_index(lock); |
3158 | else | |
3159 | ret = close_lock_file_gently(lock); | |
3160 | ||
dbb1c613 | 3161 | run_hooks_l("post-index-change", |
1956ecd0 BP |
3162 | istate->updated_workdir ? "1" : "0", |
3163 | istate->updated_skipworktree ? "1" : "0", NULL); | |
3164 | istate->updated_workdir = 0; | |
3165 | istate->updated_skipworktree = 0; | |
3166 | ||
3167 | return ret; | |
03b86647 NTND |
3168 | } |
3169 | ||
5fc2fc8f NTND |
3170 | static int write_split_index(struct index_state *istate, |
3171 | struct lock_file *lock, | |
3172 | unsigned flags) | |
3173 | { | |
3174 | int ret; | |
3175 | prepare_to_write_split_index(istate); | |
3176 | ret = do_write_locked_index(istate, lock, flags); | |
3177 | finish_writing_split_index(istate); | |
3178 | return ret; | |
3179 | } | |
3180 | ||
b9683722 CC |
3181 | static const char *shared_index_expire = "2.weeks.ago"; |
3182 | ||
3183 | static unsigned long get_shared_index_expire_date(void) | |
3184 | { | |
3185 | static unsigned long shared_index_expire_date; | |
3186 | static int shared_index_expire_date_prepared; | |
3187 | ||
3188 | if (!shared_index_expire_date_prepared) { | |
3189 | git_config_get_expiry("splitindex.sharedindexexpire", | |
3190 | &shared_index_expire); | |
3191 | shared_index_expire_date = approxidate(shared_index_expire); | |
3192 | shared_index_expire_date_prepared = 1; | |
3193 | } | |
3194 | ||
3195 | return shared_index_expire_date; | |
3196 | } | |
3197 | ||
3198 | static int should_delete_shared_index(const char *shared_index_path) | |
3199 | { | |
3200 | struct stat st; | |
3201 | unsigned long expiration; | |
3202 | ||
3203 | /* Check timestamp */ | |
3204 | expiration = get_shared_index_expire_date(); | |
3205 | if (!expiration) | |
3206 | return 0; | |
3207 | if (stat(shared_index_path, &st)) | |
78bde923 | 3208 | return error_errno(_("could not stat '%s'"), shared_index_path); |
b9683722 CC |
3209 | if (st.st_mtime > expiration) |
3210 | return 0; | |
3211 | ||
3212 | return 1; | |
3213 | } | |
3214 | ||
3215 | static int clean_shared_index_files(const char *current_hex) | |
3216 | { | |
3217 | struct dirent *de; | |
3218 | DIR *dir = opendir(get_git_dir()); | |
3219 | ||
3220 | if (!dir) | |
3221 | return error_errno(_("unable to open git dir: %s"), get_git_dir()); | |
3222 | ||
3223 | while ((de = readdir(dir)) != NULL) { | |
3224 | const char *sha1_hex; | |
3225 | const char *shared_index_path; | |
3226 | if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex)) | |
3227 | continue; | |
3228 | if (!strcmp(sha1_hex, current_hex)) | |
3229 | continue; | |
3230 | shared_index_path = git_path("%s", de->d_name); | |
3231 | if (should_delete_shared_index(shared_index_path) > 0 && | |
3232 | unlink(shared_index_path)) | |
3233 | warning_errno(_("unable to unlink: %s"), shared_index_path); | |
3234 | } | |
3235 | closedir(dir); | |
3236 | ||
3237 | return 0; | |
3238 | } | |
3239 | ||
a0a96756 | 3240 | static int write_shared_index(struct index_state *istate, |
59f9d2dd | 3241 | struct tempfile **temp) |
c18b80a0 NTND |
3242 | { |
3243 | struct split_index *si = istate->split_index; | |
6e773527 | 3244 | int ret, was_full = !istate->sparse_index; |
c18b80a0 | 3245 | |
c18b80a0 | 3246 | move_cache_to_base_index(istate); |
ce7a9f01 | 3247 | convert_to_sparse(istate, 0); |
42fee7a3 JH |
3248 | |
3249 | trace2_region_enter_printf("index", "shared/do_write_index", | |
6a8c89d0 | 3250 | the_repository, "%s", get_tempfile_path(*temp)); |
7db2d08c | 3251 | ret = do_write_index(si->base, *temp, 1); |
c173542c | 3252 | trace2_region_leave_printf("index", "shared/do_write_index", |
6a8c89d0 | 3253 | the_repository, "%s", get_tempfile_path(*temp)); |
42fee7a3 | 3254 | |
6e773527 DS |
3255 | if (was_full) |
3256 | ensure_full_index(istate); | |
3257 | ||
59f9d2dd | 3258 | if (ret) |
c18b80a0 | 3259 | return ret; |
7db2d08c | 3260 | ret = adjust_shared_perm(get_tempfile_path(*temp)); |
df801f3f | 3261 | if (ret) { |
9d0a9e90 | 3262 | error(_("cannot fix permission bits on '%s'"), get_tempfile_path(*temp)); |
df801f3f CC |
3263 | return ret; |
3264 | } | |
7db2d08c | 3265 | ret = rename_tempfile(temp, |
75691ea3 | 3266 | git_path("sharedindex.%s", oid_to_hex(&si->base->oid))); |
b9683722 | 3267 | if (!ret) { |
75691ea3 | 3268 | oidcpy(&si->base_oid, &si->base->oid); |
3269 | clean_shared_index_files(oid_to_hex(&si->base->oid)); | |
b9683722 CC |
3270 | } |
3271 | ||
c18b80a0 NTND |
3272 | return ret; |
3273 | } | |
3274 | ||
e6a1dd77 CC |
3275 | static const int default_max_percent_split_change = 20; |
3276 | ||
3277 | static int too_many_not_shared_entries(struct index_state *istate) | |
3278 | { | |
3279 | int i, not_shared = 0; | |
3280 | int max_split = git_config_get_max_percent_split_change(); | |
3281 | ||
3282 | switch (max_split) { | |
3283 | case -1: | |
3284 | /* not or badly configured: use the default value */ | |
3285 | max_split = default_max_percent_split_change; | |
3286 | break; | |
3287 | case 0: | |
3288 | return 1; /* 0% means always write a new shared index */ | |
3289 | case 100: | |
3290 | return 0; /* 100% means never write a new shared index */ | |
3291 | default: | |
3292 | break; /* just use the configured value */ | |
3293 | } | |
3294 | ||
3295 | /* Count not shared entries */ | |
3296 | for (i = 0; i < istate->cache_nr; i++) { | |
3297 | struct cache_entry *ce = istate->cache[i]; | |
3298 | if (!ce->index) | |
3299 | not_shared++; | |
3300 | } | |
3301 | ||
3302 | return (int64_t)istate->cache_nr * max_split < (int64_t)not_shared * 100; | |
3303 | } | |
3304 | ||
03b86647 NTND |
3305 | int write_locked_index(struct index_state *istate, struct lock_file *lock, |
3306 | unsigned flags) | |
3307 | { | |
e8ffd034 | 3308 | int new_shared_index, ret, test_split_index_env; |
5fc2fc8f NTND |
3309 | struct split_index *si = istate->split_index; |
3310 | ||
4592e608 | 3311 | if (git_env_bool("GIT_TEST_CHECK_CACHE_TREE", 0)) |
c207e9e1 | 3312 | cache_tree_verify(the_repository, istate); |
4592e608 | 3313 | |
61000814 MÅ |
3314 | if ((flags & SKIP_IF_UNCHANGED) && !istate->cache_changed) { |
3315 | if (flags & COMMIT_LOCK) | |
3316 | rollback_lock_file(lock); | |
3317 | return 0; | |
3318 | } | |
3319 | ||
3bd28eb2 AV |
3320 | if (istate->fsmonitor_last_update) |
3321 | fill_fsmonitor_bitmap(istate); | |
3322 | ||
e8ffd034 SG |
3323 | test_split_index_env = git_env_bool("GIT_TEST_SPLIT_INDEX", 0); |
3324 | ||
3325 | if ((!si && !test_split_index_env) || | |
3326 | alternate_index_output || | |
5165dd59 | 3327 | (istate->cache_changed & ~EXTMASK)) { |
5fc2fc8f | 3328 | if (si) |
2182abd9 | 3329 | oidclr(&si->base_oid); |
df60cf57 MÅ |
3330 | ret = do_write_locked_index(istate, lock, flags); |
3331 | goto out; | |
5fc2fc8f NTND |
3332 | } |
3333 | ||
e8ffd034 SG |
3334 | if (test_split_index_env) { |
3335 | if (!si) { | |
3336 | si = init_split_index(istate); | |
d6e3c181 | 3337 | istate->cache_changed |= SPLIT_INDEX_ORDERED; |
e8ffd034 SG |
3338 | } else { |
3339 | int v = si->base_oid.hash[0]; | |
3340 | if ((v & 15) < 6) | |
3341 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
3342 | } | |
d6e3c181 | 3343 | } |
e6a1dd77 CC |
3344 | if (too_many_not_shared_entries(istate)) |
3345 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
0d59ffb4 CC |
3346 | |
3347 | new_shared_index = istate->cache_changed & SPLIT_INDEX_ORDERED; | |
3348 | ||
3349 | if (new_shared_index) { | |
59f9d2dd NTND |
3350 | struct tempfile *temp; |
3351 | int saved_errno; | |
3352 | ||
c9d6c788 ÆAB |
3353 | /* Same initial permissions as the main .git/index file */ |
3354 | temp = mks_tempfile_sm(git_path("sharedindex_XXXXXX"), 0, 0666); | |
59f9d2dd | 3355 | if (!temp) { |
2182abd9 | 3356 | oidclr(&si->base_oid); |
59f9d2dd | 3357 | ret = do_write_locked_index(istate, lock, flags); |
ef5b3a6c NTND |
3358 | goto out; |
3359 | } | |
3360 | ret = write_shared_index(istate, &temp); | |
59f9d2dd NTND |
3361 | |
3362 | saved_errno = errno; | |
3363 | if (is_tempfile_active(temp)) | |
3364 | delete_tempfile(&temp); | |
3365 | errno = saved_errno; | |
3366 | ||
c18b80a0 | 3367 | if (ret) |
df60cf57 | 3368 | goto out; |
c18b80a0 NTND |
3369 | } |
3370 | ||
0d59ffb4 CC |
3371 | ret = write_split_index(istate, lock, flags); |
3372 | ||
3373 | /* Freshen the shared index only if the split-index was written */ | |
6e37c8ed | 3374 | if (!ret && !new_shared_index && !is_null_oid(&si->base_oid)) { |
a125a223 | 3375 | const char *shared_index = git_path("sharedindex.%s", |
2182abd9 | 3376 | oid_to_hex(&si->base_oid)); |
a125a223 TG |
3377 | freshen_shared_index(shared_index, 1); |
3378 | } | |
0d59ffb4 | 3379 | |
df60cf57 MÅ |
3380 | out: |
3381 | if (flags & COMMIT_LOCK) | |
3382 | rollback_lock_file(lock); | |
0d59ffb4 | 3383 | return ret; |
03b86647 NTND |
3384 | } |
3385 | ||
e46bbcf6 MV |
3386 | /* |
3387 | * Read the index file that is potentially unmerged into given | |
ad376204 EN |
3388 | * index_state, dropping any unmerged entries to stage #0 (potentially |
3389 | * resulting in a path appearing as both a file and a directory in the | |
3390 | * index; the caller is responsible to clear out the extra entries | |
3391 | * before writing the index to a tree). Returns true if the index is | |
3392 | * unmerged. Callers who want to refuse to work from an unmerged | |
3393 | * state can call this and check its return value, instead of calling | |
3394 | * read_cache(). | |
e46bbcf6 | 3395 | */ |
e1ff0a32 | 3396 | int repo_read_index_unmerged(struct repository *repo) |
e46bbcf6 | 3397 | { |
e1ff0a32 | 3398 | struct index_state *istate; |
e46bbcf6 | 3399 | int i; |
d1a43f2a | 3400 | int unmerged = 0; |
e46bbcf6 | 3401 | |
e1ff0a32 NTND |
3402 | repo_read_index(repo); |
3403 | istate = repo->index; | |
e46bbcf6 MV |
3404 | for (i = 0; i < istate->cache_nr; i++) { |
3405 | struct cache_entry *ce = istate->cache[i]; | |
d1a43f2a | 3406 | struct cache_entry *new_ce; |
a849735b | 3407 | int len; |
d1a43f2a JH |
3408 | |
3409 | if (!ce_stage(ce)) | |
e46bbcf6 | 3410 | continue; |
d1a43f2a | 3411 | unmerged = 1; |
68c4f6a5 | 3412 | len = ce_namelen(ce); |
a849735b | 3413 | new_ce = make_empty_cache_entry(istate, len); |
d1a43f2a | 3414 | memcpy(new_ce->name, ce->name, len); |
b60e188c TG |
3415 | new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED; |
3416 | new_ce->ce_namelen = len; | |
d1a43f2a | 3417 | new_ce->ce_mode = ce->ce_mode; |
ad376204 | 3418 | if (add_index_entry(istate, new_ce, ADD_CACHE_SKIP_DFCHECK)) |
9d0a9e90 | 3419 | return error(_("%s: cannot drop to stage #0"), |
5699d17e | 3420 | new_ce->name); |
e46bbcf6 | 3421 | } |
d1a43f2a | 3422 | return unmerged; |
e46bbcf6 | 3423 | } |
041aee31 | 3424 | |
98fa4738 JK |
3425 | /* |
3426 | * Returns 1 if the path is an "other" path with respect to | |
3427 | * the index; that is, the path is not mentioned in the index at all, | |
3428 | * either as a file, a directory with some files in the index, | |
3429 | * or as an unmerged entry. | |
3430 | * | |
3431 | * We helpfully remove a trailing "/" from directories so that | |
3432 | * the output of read_directory can be used as-is. | |
3433 | */ | |
847a9e5d DS |
3434 | int index_name_is_other(struct index_state *istate, const char *name, |
3435 | int namelen) | |
98fa4738 JK |
3436 | { |
3437 | int pos; | |
3438 | if (namelen && name[namelen - 1] == '/') | |
3439 | namelen--; | |
3440 | pos = index_name_pos(istate, name, namelen); | |
3441 | if (0 <= pos) | |
3442 | return 0; /* exact match */ | |
3443 | pos = -pos - 1; | |
3444 | if (pos < istate->cache_nr) { | |
3445 | struct cache_entry *ce = istate->cache[pos]; | |
3446 | if (ce_namelen(ce) == namelen && | |
3447 | !memcmp(ce->name, name, namelen)) | |
3448 | return 0; /* Yup, this one exists unmerged */ | |
3449 | } | |
3450 | return 1; | |
3451 | } | |
29fb37b2 | 3452 | |
847a9e5d | 3453 | void *read_blob_data_from_index(struct index_state *istate, |
87542508 | 3454 | const char *path, unsigned long *size) |
29fb37b2 LF |
3455 | { |
3456 | int pos, len; | |
3457 | unsigned long sz; | |
3458 | enum object_type type; | |
3459 | void *data; | |
3460 | ||
3461 | len = strlen(path); | |
3462 | pos = index_name_pos(istate, path, len); | |
3463 | if (pos < 0) { | |
3464 | /* | |
3465 | * We might be in the middle of a merge, in which | |
3466 | * case we would read stage #2 (ours). | |
3467 | */ | |
3468 | int i; | |
3469 | for (i = -pos - 1; | |
3470 | (pos < 0 && i < istate->cache_nr && | |
3471 | !strcmp(istate->cache[i]->name, path)); | |
3472 | i++) | |
3473 | if (ce_stage(istate->cache[i]) == 2) | |
3474 | pos = i; | |
3475 | } | |
3476 | if (pos < 0) | |
3477 | return NULL; | |
b4f5aca4 | 3478 | data = read_object_file(&istate->cache[pos]->oid, &type, &sz); |
29fb37b2 LF |
3479 | if (!data || type != OBJ_BLOB) { |
3480 | free(data); | |
3481 | return NULL; | |
3482 | } | |
ff366825 LF |
3483 | if (size) |
3484 | *size = sz; | |
29fb37b2 LF |
3485 | return data; |
3486 | } | |
38612532 MH |
3487 | |
3488 | void stat_validity_clear(struct stat_validity *sv) | |
3489 | { | |
6a83d902 | 3490 | FREE_AND_NULL(sv->sd); |
38612532 MH |
3491 | } |
3492 | ||
3493 | int stat_validity_check(struct stat_validity *sv, const char *path) | |
3494 | { | |
3495 | struct stat st; | |
3496 | ||
3497 | if (stat(path, &st) < 0) | |
3498 | return sv->sd == NULL; | |
3499 | if (!sv->sd) | |
3500 | return 0; | |
3501 | return S_ISREG(st.st_mode) && !match_stat_data(sv->sd, &st); | |
3502 | } | |
3503 | ||
3504 | void stat_validity_update(struct stat_validity *sv, int fd) | |
3505 | { | |
3506 | struct stat st; | |
3507 | ||
3508 | if (fstat(fd, &st) < 0 || !S_ISREG(st.st_mode)) | |
3509 | stat_validity_clear(sv); | |
3510 | else { | |
3511 | if (!sv->sd) | |
ca56dadb | 3512 | CALLOC_ARRAY(sv->sd, 1); |
38612532 MH |
3513 | fill_stat_data(sv->sd, &st); |
3514 | } | |
3515 | } | |
edf3b905 DT |
3516 | |
3517 | void move_index_extensions(struct index_state *dst, struct index_state *src) | |
3518 | { | |
3519 | dst->untracked = src->untracked; | |
3520 | src->untracked = NULL; | |
836ef2b6 NTND |
3521 | dst->cache_tree = src->cache_tree; |
3522 | src->cache_tree = NULL; | |
edf3b905 | 3523 | } |
a849735b | 3524 | |
8e72d675 JM |
3525 | struct cache_entry *dup_cache_entry(const struct cache_entry *ce, |
3526 | struct index_state *istate) | |
3527 | { | |
3528 | unsigned int size = ce_size(ce); | |
3529 | int mem_pool_allocated; | |
3530 | struct cache_entry *new_entry = make_empty_cache_entry(istate, ce_namelen(ce)); | |
3531 | mem_pool_allocated = new_entry->mem_pool_allocated; | |
3532 | ||
3533 | memcpy(new_entry, ce, size); | |
3534 | new_entry->mem_pool_allocated = mem_pool_allocated; | |
3535 | return new_entry; | |
3536 | } | |
3537 | ||
a849735b JM |
3538 | void discard_cache_entry(struct cache_entry *ce) |
3539 | { | |
8616a2d0 JM |
3540 | if (ce && should_validate_cache_entries()) |
3541 | memset(ce, 0xCD, cache_entry_size(ce->ce_namelen)); | |
3542 | ||
8e72d675 JM |
3543 | if (ce && ce->mem_pool_allocated) |
3544 | return; | |
3545 | ||
a849735b JM |
3546 | free(ce); |
3547 | } | |
8616a2d0 JM |
3548 | |
3549 | int should_validate_cache_entries(void) | |
3550 | { | |
3551 | static int validate_index_cache_entries = -1; | |
3552 | ||
3553 | if (validate_index_cache_entries < 0) { | |
3554 | if (getenv("GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES")) | |
3555 | validate_index_cache_entries = 1; | |
3556 | else | |
3557 | validate_index_cache_entries = 0; | |
3558 | } | |
3559 | ||
3560 | return validate_index_cache_entries; | |
3561 | } | |
3b1d9e04 BP |
3562 | |
3563 | #define EOIE_SIZE (4 + GIT_SHA1_RAWSZ) /* <4-byte offset> + <20-byte hash> */ | |
3564 | #define EOIE_SIZE_WITH_HEADER (4 + 4 + EOIE_SIZE) /* <4-byte signature> + <4-byte length> + EOIE_SIZE */ | |
3565 | ||
3566 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size) | |
3567 | { | |
3568 | /* | |
3569 | * The end of index entries (EOIE) extension is guaranteed to be last | |
3570 | * so that it can be found by scanning backwards from the EOF. | |
3571 | * | |
3572 | * "EOIE" | |
3573 | * <4-byte length> | |
3574 | * <4-byte offset> | |
3575 | * <20-byte hash> | |
3576 | */ | |
3577 | const char *index, *eoie; | |
3578 | uint32_t extsize; | |
3579 | size_t offset, src_offset; | |
3580 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3581 | git_hash_ctx c; | |
3582 | ||
3583 | /* ensure we have an index big enough to contain an EOIE extension */ | |
3584 | if (mmap_size < sizeof(struct cache_header) + EOIE_SIZE_WITH_HEADER + the_hash_algo->rawsz) | |
3585 | return 0; | |
3586 | ||
3587 | /* validate the extension signature */ | |
3588 | index = eoie = mmap + mmap_size - EOIE_SIZE_WITH_HEADER - the_hash_algo->rawsz; | |
3589 | if (CACHE_EXT(index) != CACHE_EXT_ENDOFINDEXENTRIES) | |
3590 | return 0; | |
3591 | index += sizeof(uint32_t); | |
3592 | ||
3593 | /* validate the extension size */ | |
3594 | extsize = get_be32(index); | |
3595 | if (extsize != EOIE_SIZE) | |
3596 | return 0; | |
3597 | index += sizeof(uint32_t); | |
3598 | ||
3599 | /* | |
3600 | * Validate the offset we're going to look for the first extension | |
3601 | * signature is after the index header and before the eoie extension. | |
3602 | */ | |
3603 | offset = get_be32(index); | |
3604 | if (mmap + offset < mmap + sizeof(struct cache_header)) | |
3605 | return 0; | |
3606 | if (mmap + offset >= eoie) | |
3607 | return 0; | |
3608 | index += sizeof(uint32_t); | |
3609 | ||
3610 | /* | |
3611 | * The hash is computed over extension types and their sizes (but not | |
3612 | * their contents). E.g. if we have "TREE" extension that is N-bytes | |
3613 | * long, "REUC" extension that is M-bytes long, followed by "EOIE", | |
3614 | * then the hash would be: | |
3615 | * | |
3616 | * SHA-1("TREE" + <binary representation of N> + | |
3617 | * "REUC" + <binary representation of M>) | |
3618 | */ | |
3619 | src_offset = offset; | |
3620 | the_hash_algo->init_fn(&c); | |
3621 | while (src_offset < mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) { | |
3622 | /* After an array of active_nr index entries, | |
3623 | * there can be arbitrary number of extended | |
3624 | * sections, each of which is prefixed with | |
3625 | * extension name (4-byte) and section length | |
3626 | * in 4-byte network byte order. | |
3627 | */ | |
3628 | uint32_t extsize; | |
3629 | memcpy(&extsize, mmap + src_offset + 4, 4); | |
3630 | extsize = ntohl(extsize); | |
3631 | ||
3632 | /* verify the extension size isn't so large it will wrap around */ | |
3633 | if (src_offset + 8 + extsize < src_offset) | |
3634 | return 0; | |
3635 | ||
3636 | the_hash_algo->update_fn(&c, mmap + src_offset, 8); | |
3637 | ||
3638 | src_offset += 8; | |
3639 | src_offset += extsize; | |
3640 | } | |
3641 | the_hash_algo->final_fn(hash, &c); | |
3642 | if (!hasheq(hash, (const unsigned char *)index)) | |
3643 | return 0; | |
3644 | ||
3645 | /* Validate that the extension offsets returned us back to the eoie extension. */ | |
3646 | if (src_offset != mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) | |
3647 | return 0; | |
3648 | ||
3649 | return offset; | |
3650 | } | |
3651 | ||
3652 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset) | |
3653 | { | |
3654 | uint32_t buffer; | |
3655 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3656 | ||
3657 | /* offset */ | |
3658 | put_be32(&buffer, offset); | |
3659 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3660 | ||
3661 | /* hash */ | |
3662 | the_hash_algo->final_fn(hash, eoie_context); | |
3663 | strbuf_add(sb, hash, the_hash_algo->rawsz); | |
3664 | } | |
3255089a | 3665 | |
3255089a BP |
3666 | #define IEOT_VERSION (1) |
3667 | ||
3668 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset) | |
3669 | { | |
ec36c42a NTND |
3670 | const char *index = NULL; |
3671 | uint32_t extsize, ext_version; | |
3672 | struct index_entry_offset_table *ieot; | |
3673 | int i, nr; | |
3674 | ||
3675 | /* find the IEOT extension */ | |
3676 | if (!offset) | |
3677 | return NULL; | |
3678 | while (offset <= mmap_size - the_hash_algo->rawsz - 8) { | |
3679 | extsize = get_be32(mmap + offset + 4); | |
3680 | if (CACHE_EXT((mmap + offset)) == CACHE_EXT_INDEXENTRYOFFSETTABLE) { | |
3681 | index = mmap + offset + 4 + 4; | |
3682 | break; | |
3683 | } | |
3684 | offset += 8; | |
3685 | offset += extsize; | |
3686 | } | |
3687 | if (!index) | |
3688 | return NULL; | |
3689 | ||
3690 | /* validate the version is IEOT_VERSION */ | |
3691 | ext_version = get_be32(index); | |
3692 | if (ext_version != IEOT_VERSION) { | |
3693 | error("invalid IEOT version %d", ext_version); | |
3694 | return NULL; | |
3695 | } | |
3696 | index += sizeof(uint32_t); | |
3697 | ||
3698 | /* extension size - version bytes / bytes per entry */ | |
3699 | nr = (extsize - sizeof(uint32_t)) / (sizeof(uint32_t) + sizeof(uint32_t)); | |
3700 | if (!nr) { | |
3701 | error("invalid number of IEOT entries %d", nr); | |
3702 | return NULL; | |
3703 | } | |
3704 | ieot = xmalloc(sizeof(struct index_entry_offset_table) | |
3705 | + (nr * sizeof(struct index_entry_offset))); | |
3706 | ieot->nr = nr; | |
3707 | for (i = 0; i < nr; i++) { | |
3708 | ieot->entries[i].offset = get_be32(index); | |
3709 | index += sizeof(uint32_t); | |
3710 | ieot->entries[i].nr = get_be32(index); | |
3711 | index += sizeof(uint32_t); | |
3712 | } | |
3713 | ||
3714 | return ieot; | |
3255089a BP |
3715 | } |
3716 | ||
3717 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot) | |
3718 | { | |
ec36c42a NTND |
3719 | uint32_t buffer; |
3720 | int i; | |
3255089a | 3721 | |
ec36c42a NTND |
3722 | /* version */ |
3723 | put_be32(&buffer, IEOT_VERSION); | |
3724 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3725 | |
ec36c42a NTND |
3726 | /* ieot */ |
3727 | for (i = 0; i < ieot->nr; i++) { | |
3255089a | 3728 | |
ec36c42a NTND |
3729 | /* offset */ |
3730 | put_be32(&buffer, ieot->entries[i].offset); | |
3731 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3732 | |
ec36c42a NTND |
3733 | /* count */ |
3734 | put_be32(&buffer, ieot->entries[i].nr); | |
3735 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3736 | } | |
3255089a | 3737 | } |
b2896d27 JT |
3738 | |
3739 | void prefetch_cache_entries(const struct index_state *istate, | |
3740 | must_prefetch_predicate must_prefetch) | |
3741 | { | |
3742 | int i; | |
3743 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
3744 | ||
3745 | for (i = 0; i < istate->cache_nr; i++) { | |
3746 | struct cache_entry *ce = istate->cache[i]; | |
3747 | ||
3748 | if (S_ISGITLINK(ce->ce_mode) || !must_prefetch(ce)) | |
3749 | continue; | |
3750 | if (!oid_object_info_extended(the_repository, &ce->oid, | |
3751 | NULL, | |
3752 | OBJECT_INFO_FOR_PREFETCH)) | |
3753 | continue; | |
3754 | oid_array_append(&to_fetch, &ce->oid); | |
3755 | } | |
3756 | promisor_remote_get_direct(the_repository, | |
3757 | to_fetch.oid, to_fetch.nr); | |
3758 | oid_array_clear(&to_fetch); | |
3759 | } |