]>
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" |
bad68ec9 | 31 | |
b60e188c TG |
32 | /* Mask for the name length in ce_flags in the on-disk index */ |
33 | ||
34 | #define CE_NAMEMASK (0x0fff) | |
35 | ||
bad68ec9 JH |
36 | /* Index extensions. |
37 | * | |
38 | * The first letter should be 'A'..'Z' for extensions that are not | |
39 | * necessary for a correct operation (i.e. optimization data). | |
40 | * When new extensions are added that _needs_ to be understood in | |
41 | * order to correctly interpret the index file, pick character that | |
42 | * is outside the range, to cause the reader to abort. | |
43 | */ | |
44 | ||
45 | #define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) ) | |
46 | #define CACHE_EXT_TREE 0x54524545 /* "TREE" */ | |
b659b49b | 47 | #define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */ |
5fc2fc8f | 48 | #define CACHE_EXT_LINK 0x6c696e6b /* "link" */ |
83c094ad | 49 | #define CACHE_EXT_UNTRACKED 0x554E5452 /* "UNTR" */ |
883e248b | 50 | #define CACHE_EXT_FSMONITOR 0x46534D4E /* "FSMN" */ |
3b1d9e04 | 51 | #define CACHE_EXT_ENDOFINDEXENTRIES 0x454F4945 /* "EOIE" */ |
3255089a | 52 | #define CACHE_EXT_INDEXENTRYOFFSETTABLE 0x49454F54 /* "IEOT" */ |
cd42415f | 53 | #define CACHE_EXT_SPARSE_DIRECTORIES 0x73646972 /* "sdir" */ |
5fc2fc8f NTND |
54 | |
55 | /* changes that can be kept in $GIT_DIR/index (basically all extensions) */ | |
e0cf0d7d | 56 | #define EXTMASK (RESOLVE_UNDO_CHANGED | CACHE_TREE_CHANGED | \ |
c18b80a0 | 57 | CE_ENTRY_ADDED | CE_ENTRY_REMOVED | CE_ENTRY_CHANGED | \ |
883e248b | 58 | SPLIT_INDEX_ORDERED | UNTRACKED_CHANGED | FSMONITOR_CHANGED) |
e83c5163 | 59 | |
8e72d675 JM |
60 | |
61 | /* | |
62 | * This is an estimate of the pathname length in the index. We use | |
63 | * this for V4 index files to guess the un-deltafied size of the index | |
64 | * in memory because of pathname deltafication. This is not required | |
65 | * for V2/V3 index formats because their pathnames are not compressed. | |
66 | * If the initial amount of memory set aside is not sufficient, the | |
67 | * mem pool will allocate extra memory. | |
68 | */ | |
69 | #define CACHE_ENTRY_PATH_LENGTH 80 | |
70 | ||
71 | static inline struct cache_entry *mem_pool__ce_alloc(struct mem_pool *mem_pool, size_t len) | |
72 | { | |
73 | struct cache_entry *ce; | |
74 | ce = mem_pool_alloc(mem_pool, cache_entry_size(len)); | |
75 | ce->mem_pool_allocated = 1; | |
76 | return ce; | |
77 | } | |
78 | ||
79 | static inline struct cache_entry *mem_pool__ce_calloc(struct mem_pool *mem_pool, size_t len) | |
80 | { | |
81 | struct cache_entry * ce; | |
82 | ce = mem_pool_calloc(mem_pool, 1, cache_entry_size(len)); | |
83 | ce->mem_pool_allocated = 1; | |
84 | return ce; | |
85 | } | |
86 | ||
87 | static struct mem_pool *find_mem_pool(struct index_state *istate) | |
88 | { | |
89 | struct mem_pool **pool_ptr; | |
90 | ||
91 | if (istate->split_index && istate->split_index->base) | |
92 | pool_ptr = &istate->split_index->base->ce_mem_pool; | |
93 | else | |
94 | pool_ptr = &istate->ce_mem_pool; | |
95 | ||
44c7e1a7 EN |
96 | if (!*pool_ptr) { |
97 | *pool_ptr = xmalloc(sizeof(**pool_ptr)); | |
98 | mem_pool_init(*pool_ptr, 0); | |
99 | } | |
8e72d675 JM |
100 | |
101 | return *pool_ptr; | |
102 | } | |
103 | ||
626f35c8 | 104 | static const char *alternate_index_output; |
8fd2cb40 | 105 | |
9cb76b8c JH |
106 | static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
107 | { | |
4300f844 DS |
108 | if (S_ISSPARSEDIR(ce->ce_mode)) |
109 | istate->sparse_index = 1; | |
110 | ||
9cb76b8c | 111 | istate->cache[nr] = ce; |
96872bc2 | 112 | add_name_hash(istate, ce); |
9cb76b8c JH |
113 | } |
114 | ||
cf558704 LT |
115 | static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
116 | { | |
117 | struct cache_entry *old = istate->cache[nr]; | |
118 | ||
078a58e8 | 119 | replace_index_entry_in_base(istate, old, ce); |
2092678c | 120 | remove_name_hash(istate, old); |
a849735b | 121 | discard_cache_entry(old); |
0e267b7a | 122 | ce->ce_flags &= ~CE_HASHED; |
a22c6371 | 123 | set_index_entry(istate, nr, ce); |
078a58e8 | 124 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 125 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 126 | istate->cache_changed |= CE_ENTRY_CHANGED; |
cf558704 LT |
127 | } |
128 | ||
81dc2307 PB |
129 | void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name) |
130 | { | |
285c2e25 | 131 | struct cache_entry *old_entry = istate->cache[nr], *new_entry; |
81dc2307 PB |
132 | int namelen = strlen(new_name); |
133 | ||
a849735b | 134 | new_entry = make_empty_cache_entry(istate, namelen); |
285c2e25 BW |
135 | copy_cache_entry(new_entry, old_entry); |
136 | new_entry->ce_flags &= ~CE_HASHED; | |
137 | new_entry->ce_namelen = namelen; | |
138 | new_entry->index = 0; | |
139 | memcpy(new_entry->name, new_name, namelen + 1); | |
81dc2307 | 140 | |
285c2e25 BW |
141 | cache_tree_invalidate_path(istate, old_entry->name); |
142 | untracked_cache_remove_from_index(istate, old_entry->name); | |
81dc2307 | 143 | remove_index_entry_at(istate, nr); |
285c2e25 | 144 | add_index_entry(istate, new_entry, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); |
81dc2307 PB |
145 | } |
146 | ||
c21d39d7 MH |
147 | void fill_stat_data(struct stat_data *sd, struct stat *st) |
148 | { | |
149 | sd->sd_ctime.sec = (unsigned int)st->st_ctime; | |
150 | sd->sd_mtime.sec = (unsigned int)st->st_mtime; | |
151 | sd->sd_ctime.nsec = ST_CTIME_NSEC(*st); | |
152 | sd->sd_mtime.nsec = ST_MTIME_NSEC(*st); | |
153 | sd->sd_dev = st->st_dev; | |
154 | sd->sd_ino = st->st_ino; | |
155 | sd->sd_uid = st->st_uid; | |
156 | sd->sd_gid = st->st_gid; | |
157 | sd->sd_size = st->st_size; | |
158 | } | |
159 | ||
160 | int match_stat_data(const struct stat_data *sd, struct stat *st) | |
161 | { | |
162 | int changed = 0; | |
163 | ||
164 | if (sd->sd_mtime.sec != (unsigned int)st->st_mtime) | |
165 | changed |= MTIME_CHANGED; | |
166 | if (trust_ctime && check_stat && | |
167 | sd->sd_ctime.sec != (unsigned int)st->st_ctime) | |
168 | changed |= CTIME_CHANGED; | |
169 | ||
170 | #ifdef USE_NSEC | |
171 | if (check_stat && sd->sd_mtime.nsec != ST_MTIME_NSEC(*st)) | |
172 | changed |= MTIME_CHANGED; | |
173 | if (trust_ctime && check_stat && | |
174 | sd->sd_ctime.nsec != ST_CTIME_NSEC(*st)) | |
175 | changed |= CTIME_CHANGED; | |
176 | #endif | |
177 | ||
178 | if (check_stat) { | |
179 | if (sd->sd_uid != (unsigned int) st->st_uid || | |
180 | sd->sd_gid != (unsigned int) st->st_gid) | |
181 | changed |= OWNER_CHANGED; | |
182 | if (sd->sd_ino != (unsigned int) st->st_ino) | |
183 | changed |= INODE_CHANGED; | |
184 | } | |
185 | ||
186 | #ifdef USE_STDEV | |
187 | /* | |
188 | * st_dev breaks on network filesystems where different | |
189 | * clients will have different views of what "device" | |
190 | * the filesystem is on | |
191 | */ | |
192 | if (check_stat && sd->sd_dev != (unsigned int) st->st_dev) | |
193 | changed |= INODE_CHANGED; | |
194 | #endif | |
195 | ||
196 | if (sd->sd_size != (unsigned int) st->st_size) | |
197 | changed |= DATA_CHANGED; | |
198 | ||
199 | return changed; | |
200 | } | |
201 | ||
415e96c8 JH |
202 | /* |
203 | * This only updates the "non-critical" parts of the directory | |
204 | * cache, ie the parts that aren't tracked by GIT, and only used | |
205 | * to validate the cache. | |
206 | */ | |
d4c0a3ac | 207 | void fill_stat_cache_info(struct index_state *istate, struct cache_entry *ce, struct stat *st) |
415e96c8 | 208 | { |
c21d39d7 | 209 | fill_stat_data(&ce->ce_stat_data, st); |
5f73076c JH |
210 | |
211 | if (assume_unchanged) | |
7a51ed66 | 212 | ce->ce_flags |= CE_VALID; |
eadb5831 | 213 | |
883e248b | 214 | if (S_ISREG(st->st_mode)) { |
eadb5831 | 215 | ce_mark_uptodate(ce); |
b5a81697 | 216 | mark_fsmonitor_valid(istate, ce); |
883e248b | 217 | } |
415e96c8 JH |
218 | } |
219 | ||
58bf2a4c NTND |
220 | static int ce_compare_data(struct index_state *istate, |
221 | const struct cache_entry *ce, | |
222 | struct stat *st) | |
29e4d363 JH |
223 | { |
224 | int match = -1; | |
1b8ac5ea | 225 | int fd = git_open_cloexec(ce->name, O_RDONLY); |
29e4d363 JH |
226 | |
227 | if (fd >= 0) { | |
bebfecb9 | 228 | struct object_id oid; |
58bf2a4c | 229 | if (!index_fd(istate, &oid, fd, st, OBJ_BLOB, ce->name, 0)) |
6a29d7b7 | 230 | match = !oideq(&oid, &ce->oid); |
7f8508e8 | 231 | /* index_fd() closed the file descriptor already */ |
29e4d363 JH |
232 | } |
233 | return match; | |
234 | } | |
235 | ||
21a6b9fa | 236 | static int ce_compare_link(const struct cache_entry *ce, size_t expected_size) |
29e4d363 JH |
237 | { |
238 | int match = -1; | |
29e4d363 JH |
239 | void *buffer; |
240 | unsigned long size; | |
21666f1a | 241 | enum object_type type; |
a60272b3 | 242 | struct strbuf sb = STRBUF_INIT; |
29e4d363 | 243 | |
a60272b3 | 244 | if (strbuf_readlink(&sb, ce->name, expected_size)) |
29e4d363 | 245 | return -1; |
a60272b3 | 246 | |
b4f5aca4 | 247 | buffer = read_object_file(&ce->oid, &type, &size); |
a60272b3 LT |
248 | if (buffer) { |
249 | if (size == sb.len) | |
250 | match = memcmp(buffer, sb.buf, size); | |
251 | free(buffer); | |
29e4d363 | 252 | } |
a60272b3 | 253 | strbuf_release(&sb); |
29e4d363 JH |
254 | return match; |
255 | } | |
256 | ||
21a6b9fa | 257 | static int ce_compare_gitlink(const struct cache_entry *ce) |
f35a6d3b | 258 | { |
1053fe82 | 259 | struct object_id oid; |
f35a6d3b LT |
260 | |
261 | /* | |
262 | * We don't actually require that the .git directory | |
302b9282 | 263 | * under GITLINK directory be a valid git directory. It |
f35a6d3b LT |
264 | * might even be missing (in case nobody populated that |
265 | * sub-project). | |
266 | * | |
267 | * If so, we consider it always to match. | |
268 | */ | |
a98e6101 | 269 | if (resolve_gitlink_ref(ce->name, "HEAD", &oid) < 0) |
f35a6d3b | 270 | return 0; |
6a29d7b7 | 271 | return !oideq(&oid, &ce->oid); |
f35a6d3b LT |
272 | } |
273 | ||
58bf2a4c NTND |
274 | static int ce_modified_check_fs(struct index_state *istate, |
275 | const struct cache_entry *ce, | |
276 | struct stat *st) | |
29e4d363 JH |
277 | { |
278 | switch (st->st_mode & S_IFMT) { | |
279 | case S_IFREG: | |
58bf2a4c | 280 | if (ce_compare_data(istate, ce, st)) |
29e4d363 JH |
281 | return DATA_CHANGED; |
282 | break; | |
283 | case S_IFLNK: | |
dc49cd76 | 284 | if (ce_compare_link(ce, xsize_t(st->st_size))) |
29e4d363 JH |
285 | return DATA_CHANGED; |
286 | break; | |
a8ee75bc | 287 | case S_IFDIR: |
7a51ed66 | 288 | if (S_ISGITLINK(ce->ce_mode)) |
c70115b4 | 289 | return ce_compare_gitlink(ce) ? DATA_CHANGED : 0; |
1cf01a34 | 290 | /* else fallthrough */ |
29e4d363 JH |
291 | default: |
292 | return TYPE_CHANGED; | |
293 | } | |
294 | return 0; | |
295 | } | |
296 | ||
21a6b9fa | 297 | static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st) |
734aab75 LT |
298 | { |
299 | unsigned int changed = 0; | |
300 | ||
7a51ed66 LT |
301 | if (ce->ce_flags & CE_REMOVE) |
302 | return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED; | |
303 | ||
304 | switch (ce->ce_mode & S_IFMT) { | |
8ae0a8c5 KS |
305 | case S_IFREG: |
306 | changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0; | |
3e09cdfd JH |
307 | /* We consider only the owner x bit to be relevant for |
308 | * "mode changes" | |
309 | */ | |
310 | if (trust_executable_bit && | |
7a51ed66 | 311 | (0100 & (ce->ce_mode ^ st->st_mode))) |
ffbe1add | 312 | changed |= MODE_CHANGED; |
8ae0a8c5 KS |
313 | break; |
314 | case S_IFLNK: | |
78a8d641 JS |
315 | if (!S_ISLNK(st->st_mode) && |
316 | (has_symlinks || !S_ISREG(st->st_mode))) | |
317 | changed |= TYPE_CHANGED; | |
8ae0a8c5 | 318 | break; |
302b9282 | 319 | case S_IFGITLINK: |
c70115b4 | 320 | /* We ignore most of the st_xxx fields for gitlinks */ |
f35a6d3b LT |
321 | if (!S_ISDIR(st->st_mode)) |
322 | changed |= TYPE_CHANGED; | |
323 | else if (ce_compare_gitlink(ce)) | |
324 | changed |= DATA_CHANGED; | |
a8ee75bc | 325 | return changed; |
8ae0a8c5 | 326 | default: |
391408e5 | 327 | BUG("unsupported ce_mode: %o", ce->ce_mode); |
8ae0a8c5 | 328 | } |
2cb45e95 | 329 | |
c21d39d7 | 330 | changed |= match_stat_data(&ce->ce_stat_data, st); |
b0391890 | 331 | |
f49c2c22 | 332 | /* Racily smudged entry? */ |
c21d39d7 | 333 | if (!ce->ce_stat_data.sd_size) { |
99d1a986 | 334 | if (!is_empty_blob_sha1(ce->oid.hash)) |
f49c2c22 LT |
335 | changed |= DATA_CHANGED; |
336 | } | |
337 | ||
407c8eb0 JH |
338 | return changed; |
339 | } | |
340 | ||
2bb4cda1 NTND |
341 | static int is_racy_stat(const struct index_state *istate, |
342 | const struct stat_data *sd) | |
6d91da6d | 343 | { |
2bb4cda1 | 344 | return (istate->timestamp.sec && |
fba2f38a KB |
345 | #ifdef USE_NSEC |
346 | /* nanosecond timestamped files can also be racy! */ | |
2bb4cda1 NTND |
347 | (istate->timestamp.sec < sd->sd_mtime.sec || |
348 | (istate->timestamp.sec == sd->sd_mtime.sec && | |
349 | istate->timestamp.nsec <= sd->sd_mtime.nsec)) | |
fba2f38a | 350 | #else |
2bb4cda1 | 351 | istate->timestamp.sec <= sd->sd_mtime.sec |
fba2f38a | 352 | #endif |
2bb4cda1 NTND |
353 | ); |
354 | } | |
355 | ||
5581a019 | 356 | int is_racy_timestamp(const struct index_state *istate, |
2bb4cda1 NTND |
357 | const struct cache_entry *ce) |
358 | { | |
359 | return (!S_ISGITLINK(ce->ce_mode) && | |
360 | is_racy_stat(istate, &ce->ce_stat_data)); | |
6d91da6d JH |
361 | } |
362 | ||
ed4efab1 NTND |
363 | int match_stat_data_racy(const struct index_state *istate, |
364 | const struct stat_data *sd, struct stat *st) | |
365 | { | |
366 | if (is_racy_stat(istate, sd)) | |
367 | return MTIME_CHANGED; | |
368 | return match_stat_data(sd, st); | |
6d91da6d JH |
369 | } |
370 | ||
883e248b | 371 | int ie_match_stat(struct index_state *istate, |
21a6b9fa | 372 | const struct cache_entry *ce, struct stat *st, |
4bd5b7da | 373 | unsigned int options) |
407c8eb0 | 374 | { |
5f73076c | 375 | unsigned int changed; |
4bd5b7da | 376 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 377 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
4bd5b7da | 378 | int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY; |
883e248b | 379 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
5f73076c | 380 | |
883e248b BP |
381 | if (!ignore_fsmonitor) |
382 | refresh_fsmonitor(istate); | |
5f73076c JH |
383 | /* |
384 | * If it's marked as always valid in the index, it's | |
385 | * valid whatever the checked-out copy says. | |
56cac48c NTND |
386 | * |
387 | * skip-worktree has the same effect with higher precedence | |
5f73076c | 388 | */ |
56cac48c NTND |
389 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) |
390 | return 0; | |
7a51ed66 | 391 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) |
5f73076c | 392 | return 0; |
883e248b BP |
393 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) |
394 | return 0; | |
5f73076c | 395 | |
331fcb59 JH |
396 | /* |
397 | * Intent-to-add entries have not been added, so the index entry | |
398 | * by definition never matches what is in the work tree until it | |
399 | * actually gets added. | |
400 | */ | |
895ff3b2 | 401 | if (ce_intent_to_add(ce)) |
331fcb59 JH |
402 | return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED; |
403 | ||
5f73076c | 404 | changed = ce_match_stat_basic(ce, st); |
407c8eb0 | 405 | |
29e4d363 JH |
406 | /* |
407 | * Within 1 second of this sequence: | |
408 | * echo xyzzy >file && git-update-index --add file | |
409 | * running this command: | |
410 | * echo frotz >file | |
411 | * would give a falsely clean cache entry. The mtime and | |
412 | * length match the cache, and other stat fields do not change. | |
413 | * | |
414 | * We could detect this at update-index time (the cache entry | |
415 | * being registered/updated records the same time as "now") | |
416 | * and delay the return from git-update-index, but that would | |
417 | * effectively mean we can make at most one commit per second, | |
418 | * which is not acceptable. Instead, we check cache entries | |
419 | * whose mtime are the same as the index file timestamp more | |
5f73076c | 420 | * carefully than others. |
29e4d363 | 421 | */ |
6d91da6d | 422 | if (!changed && is_racy_timestamp(istate, ce)) { |
42f77406 JH |
423 | if (assume_racy_is_modified) |
424 | changed |= DATA_CHANGED; | |
425 | else | |
58bf2a4c | 426 | changed |= ce_modified_check_fs(istate, ce, st); |
42f77406 | 427 | } |
b0391890 | 428 | |
29e4d363 | 429 | return changed; |
b0391890 JH |
430 | } |
431 | ||
883e248b | 432 | int ie_modified(struct index_state *istate, |
21a6b9fa RS |
433 | const struct cache_entry *ce, |
434 | struct stat *st, unsigned int options) | |
b0391890 | 435 | { |
29e4d363 | 436 | int changed, changed_fs; |
4bd5b7da JH |
437 | |
438 | changed = ie_match_stat(istate, ce, st, options); | |
b0391890 JH |
439 | if (!changed) |
440 | return 0; | |
b0391890 JH |
441 | /* |
442 | * If the mode or type has changed, there's no point in trying | |
443 | * to refresh the entry - it's not going to match | |
444 | */ | |
445 | if (changed & (MODE_CHANGED | TYPE_CHANGED)) | |
446 | return changed; | |
447 | ||
c70115b4 JH |
448 | /* |
449 | * Immediately after read-tree or update-index --cacheinfo, | |
450 | * the length field is zero, as we have never even read the | |
451 | * lstat(2) information once, and we cannot trust DATA_CHANGED | |
452 | * returned by ie_match_stat() which in turn was returned by | |
453 | * ce_match_stat_basic() to signal that the filesize of the | |
454 | * blob changed. We have to actually go to the filesystem to | |
455 | * see if the contents match, and if so, should answer "unchanged". | |
456 | * | |
457 | * The logic does not apply to gitlinks, as ce_match_stat_basic() | |
458 | * already has checked the actual HEAD from the filesystem in the | |
459 | * subproject. If ie_match_stat() already said it is different, | |
460 | * then we know it is. | |
b0391890 | 461 | */ |
c70115b4 | 462 | if ((changed & DATA_CHANGED) && |
c21d39d7 | 463 | (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0)) |
b0391890 JH |
464 | return changed; |
465 | ||
58bf2a4c | 466 | changed_fs = ce_modified_check_fs(istate, ce, st); |
29e4d363 JH |
467 | if (changed_fs) |
468 | return changed | changed_fs; | |
b0391890 JH |
469 | return 0; |
470 | } | |
471 | ||
958ba6c9 LT |
472 | int base_name_compare(const char *name1, int len1, int mode1, |
473 | const char *name2, int len2, int mode2) | |
474 | { | |
475 | unsigned char c1, c2; | |
476 | int len = len1 < len2 ? len1 : len2; | |
477 | int cmp; | |
478 | ||
479 | cmp = memcmp(name1, name2, len); | |
480 | if (cmp) | |
481 | return cmp; | |
482 | c1 = name1[len]; | |
483 | c2 = name2[len]; | |
1833a925 | 484 | if (!c1 && S_ISDIR(mode1)) |
958ba6c9 | 485 | c1 = '/'; |
1833a925 | 486 | if (!c2 && S_ISDIR(mode2)) |
958ba6c9 LT |
487 | c2 = '/'; |
488 | return (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0; | |
489 | } | |
490 | ||
0ab9e1e8 LT |
491 | /* |
492 | * df_name_compare() is identical to base_name_compare(), except it | |
493 | * compares conflicting directory/file entries as equal. Note that | |
494 | * while a directory name compares as equal to a regular file, they | |
495 | * then individually compare _differently_ to a filename that has | |
496 | * a dot after the basename (because '\0' < '.' < '/'). | |
497 | * | |
498 | * This is used by routines that want to traverse the git namespace | |
499 | * but then handle conflicting entries together when possible. | |
500 | */ | |
501 | int df_name_compare(const char *name1, int len1, int mode1, | |
502 | const char *name2, int len2, int mode2) | |
503 | { | |
504 | int len = len1 < len2 ? len1 : len2, cmp; | |
505 | unsigned char c1, c2; | |
506 | ||
507 | cmp = memcmp(name1, name2, len); | |
508 | if (cmp) | |
509 | return cmp; | |
510 | /* Directories and files compare equal (same length, same name) */ | |
511 | if (len1 == len2) | |
512 | return 0; | |
513 | c1 = name1[len]; | |
514 | if (!c1 && S_ISDIR(mode1)) | |
515 | c1 = '/'; | |
516 | c2 = name2[len]; | |
517 | if (!c2 && S_ISDIR(mode2)) | |
518 | c2 = '/'; | |
519 | if (c1 == '/' && !c2) | |
520 | return 0; | |
521 | if (c2 == '/' && !c1) | |
522 | return 0; | |
523 | return c1 - c2; | |
524 | } | |
525 | ||
ccdd4a0f | 526 | int name_compare(const char *name1, size_t len1, const char *name2, size_t len2) |
eb38c22f | 527 | { |
ccdd4a0f JM |
528 | size_t min_len = (len1 < len2) ? len1 : len2; |
529 | int cmp = memcmp(name1, name2, min_len); | |
eb38c22f LT |
530 | if (cmp) |
531 | return cmp; | |
532 | if (len1 < len2) | |
533 | return -1; | |
534 | if (len1 > len2) | |
535 | return 1; | |
ccdd4a0f JM |
536 | return 0; |
537 | } | |
538 | ||
539 | int cache_name_stage_compare(const char *name1, int len1, int stage1, const char *name2, int len2, int stage2) | |
540 | { | |
541 | int cmp; | |
542 | ||
543 | cmp = name_compare(name1, len1, name2, len2); | |
544 | if (cmp) | |
545 | return cmp; | |
5f73076c | 546 | |
b60e188c | 547 | if (stage1 < stage2) |
95fd5bf8 | 548 | return -1; |
b60e188c | 549 | if (stage1 > stage2) |
95fd5bf8 | 550 | return 1; |
eb38c22f LT |
551 | return 0; |
552 | } | |
553 | ||
847a9e5d | 554 | static int index_name_stage_pos(struct index_state *istate, const char *name, int namelen, int stage) |
eb38c22f LT |
555 | { |
556 | int first, last; | |
557 | ||
558 | first = 0; | |
4aab5b46 | 559 | last = istate->cache_nr; |
eb38c22f | 560 | while (last > first) { |
568a05c5 | 561 | int next = first + ((last - first) >> 1); |
4aab5b46 | 562 | struct cache_entry *ce = istate->cache[next]; |
b60e188c | 563 | int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce)); |
eb38c22f | 564 | if (!cmp) |
76e7f4ec | 565 | return next; |
eb38c22f LT |
566 | if (cmp < 0) { |
567 | last = next; | |
568 | continue; | |
569 | } | |
570 | first = next+1; | |
571 | } | |
95e0321c DS |
572 | |
573 | if (istate->sparse_index && | |
574 | first > 0) { | |
575 | /* Note: first <= istate->cache_nr */ | |
576 | struct cache_entry *ce = istate->cache[first - 1]; | |
577 | ||
578 | /* | |
579 | * If we are in a sparse-index _and_ the entry before the | |
580 | * insertion position is a sparse-directory entry that is | |
581 | * an ancestor of 'name', then we need to expand the index | |
582 | * and search again. This will only trigger once, because | |
583 | * thereafter the index is fully expanded. | |
584 | */ | |
585 | if (S_ISSPARSEDIR(ce->ce_mode) && | |
586 | ce_namelen(ce) < namelen && | |
587 | !strncmp(name, ce->name, ce_namelen(ce))) { | |
588 | ensure_full_index(istate); | |
589 | return index_name_stage_pos(istate, name, namelen, stage); | |
590 | } | |
591 | } | |
592 | ||
76e7f4ec | 593 | return -first-1; |
eb38c22f LT |
594 | } |
595 | ||
847a9e5d | 596 | int index_name_pos(struct index_state *istate, const char *name, int namelen) |
b60e188c TG |
597 | { |
598 | return index_name_stage_pos(istate, name, namelen, 0); | |
599 | } | |
600 | ||
4aab5b46 | 601 | int remove_index_entry_at(struct index_state *istate, int pos) |
7b937ca3 | 602 | { |
cf558704 LT |
603 | struct cache_entry *ce = istate->cache[pos]; |
604 | ||
cfc5789a | 605 | record_resolve_undo(istate, ce); |
2092678c | 606 | remove_name_hash(istate, ce); |
045113a5 | 607 | save_or_free_index_entry(istate, ce); |
e636a7b4 | 608 | istate->cache_changed |= CE_ENTRY_REMOVED; |
4aab5b46 JH |
609 | istate->cache_nr--; |
610 | if (pos >= istate->cache_nr) | |
7b937ca3 | 611 | return 0; |
f331ab9d RS |
612 | MOVE_ARRAY(istate->cache + pos, istate->cache + pos + 1, |
613 | istate->cache_nr - pos); | |
7b937ca3 LT |
614 | return 1; |
615 | } | |
616 | ||
36419c8e | 617 | /* |
98e023de | 618 | * Remove all cache entries marked for removal, that is where |
36419c8e KB |
619 | * CE_REMOVE is set in ce_flags. This is much more effective than |
620 | * calling remove_index_entry_at() for each entry to be removed. | |
621 | */ | |
6fdc2057 | 622 | void remove_marked_cache_entries(struct index_state *istate, int invalidate) |
36419c8e KB |
623 | { |
624 | struct cache_entry **ce_array = istate->cache; | |
625 | unsigned int i, j; | |
626 | ||
627 | for (i = j = 0; i < istate->cache_nr; i++) { | |
5699d17e | 628 | if (ce_array[i]->ce_flags & CE_REMOVE) { |
6fdc2057 TG |
629 | if (invalidate) { |
630 | cache_tree_invalidate_path(istate, | |
631 | ce_array[i]->name); | |
632 | untracked_cache_remove_from_index(istate, | |
633 | ce_array[i]->name); | |
634 | } | |
2092678c | 635 | remove_name_hash(istate, ce_array[i]); |
045113a5 | 636 | save_or_free_index_entry(istate, ce_array[i]); |
5699d17e | 637 | } |
36419c8e KB |
638 | else |
639 | ce_array[j++] = ce_array[i]; | |
640 | } | |
ad837d9e NTND |
641 | if (j == istate->cache_nr) |
642 | return; | |
e636a7b4 | 643 | istate->cache_changed |= CE_ENTRY_REMOVED; |
36419c8e KB |
644 | istate->cache_nr = j; |
645 | } | |
646 | ||
4aab5b46 | 647 | int remove_file_from_index(struct index_state *istate, const char *path) |
197ee8c9 | 648 | { |
4aab5b46 | 649 | int pos = index_name_pos(istate, path, strlen(path)); |
c4e3cca1 JH |
650 | if (pos < 0) |
651 | pos = -pos-1; | |
a5400efe | 652 | cache_tree_invalidate_path(istate, path); |
e931371a | 653 | untracked_cache_remove_from_index(istate, path); |
4aab5b46 JH |
654 | while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path)) |
655 | remove_index_entry_at(istate, pos); | |
197ee8c9 LT |
656 | return 0; |
657 | } | |
658 | ||
20314271 JS |
659 | static int compare_name(struct cache_entry *ce, const char *path, int namelen) |
660 | { | |
661 | return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen); | |
662 | } | |
663 | ||
664 | static int index_name_pos_also_unmerged(struct index_state *istate, | |
665 | const char *path, int namelen) | |
666 | { | |
667 | int pos = index_name_pos(istate, path, namelen); | |
668 | struct cache_entry *ce; | |
669 | ||
670 | if (pos >= 0) | |
671 | return pos; | |
672 | ||
673 | /* maybe unmerged? */ | |
674 | pos = -1 - pos; | |
675 | if (pos >= istate->cache_nr || | |
676 | compare_name((ce = istate->cache[pos]), path, namelen)) | |
677 | return -1; | |
678 | ||
679 | /* order of preference: stage 2, 1, 3 */ | |
680 | if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr && | |
681 | ce_stage((ce = istate->cache[pos + 1])) == 2 && | |
682 | !compare_name(ce, path, namelen)) | |
683 | pos++; | |
684 | return pos; | |
685 | } | |
686 | ||
1102952b LT |
687 | static int different_name(struct cache_entry *ce, struct cache_entry *alias) |
688 | { | |
689 | int len = ce_namelen(ce); | |
690 | return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len); | |
691 | } | |
692 | ||
693 | /* | |
694 | * If we add a filename that aliases in the cache, we will use the | |
695 | * name that we already have - but we don't want to update the same | |
696 | * alias twice, because that implies that there were actually two | |
697 | * different files with aliasing names! | |
698 | * | |
699 | * So we use the CE_ADDED flag to verify that the alias was an old | |
700 | * one before we accept it as | |
701 | */ | |
045113a5 NTND |
702 | static struct cache_entry *create_alias_ce(struct index_state *istate, |
703 | struct cache_entry *ce, | |
704 | struct cache_entry *alias) | |
1102952b LT |
705 | { |
706 | int len; | |
285c2e25 | 707 | struct cache_entry *new_entry; |
1102952b LT |
708 | |
709 | if (alias->ce_flags & CE_ADDED) | |
9d0a9e90 NTND |
710 | die(_("will not add file alias '%s' ('%s' already exists in index)"), |
711 | ce->name, alias->name); | |
1102952b LT |
712 | |
713 | /* Ok, create the new entry using the name of the existing alias */ | |
714 | len = ce_namelen(alias); | |
a849735b | 715 | new_entry = make_empty_cache_entry(istate, len); |
285c2e25 BW |
716 | memcpy(new_entry->name, alias->name, len); |
717 | copy_cache_entry(new_entry, ce); | |
045113a5 | 718 | save_or_free_index_entry(istate, ce); |
285c2e25 | 719 | return new_entry; |
1102952b LT |
720 | } |
721 | ||
b4b313f9 | 722 | void set_object_name_for_intent_to_add_entry(struct cache_entry *ce) |
39425819 | 723 | { |
a09c985e PO |
724 | struct object_id oid; |
725 | if (write_object_file("", 0, blob_type, &oid)) | |
9d0a9e90 | 726 | die(_("cannot create an empty blob in the object database")); |
a09c985e | 727 | oidcpy(&ce->oid, &oid); |
39425819 JH |
728 | } |
729 | ||
610d55af | 730 | int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags) |
11be42a4 | 731 | { |
a849735b | 732 | int namelen, was_same; |
d177cab0 | 733 | mode_t st_mode = st->st_mode; |
9472935d | 734 | struct cache_entry *ce, *alias = NULL; |
56cac48c | 735 | unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY; |
38ed1d89 JH |
736 | int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND); |
737 | int pretend = flags & ADD_CACHE_PRETEND; | |
39425819 JH |
738 | int intent_only = flags & ADD_CACHE_INTENT; |
739 | int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE| | |
740 | (intent_only ? ADD_CACHE_NEW_ONLY : 0)); | |
e578d031 | 741 | unsigned hash_flags = pretend ? 0 : HASH_WRITE_OBJECT; |
f937bc2f | 742 | struct object_id oid; |
9472935d | 743 | |
9e5da3d0 JK |
744 | if (flags & ADD_CACHE_RENORMALIZE) |
745 | hash_flags |= HASH_RENORMALIZE; | |
11be42a4 | 746 | |
d177cab0 | 747 | if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode)) |
9d0a9e90 | 748 | return error(_("%s: can only add regular files, symbolic links or git-directories"), path); |
11be42a4 JS |
749 | |
750 | namelen = strlen(path); | |
d177cab0 | 751 | if (S_ISDIR(st_mode)) { |
f937bc2f KM |
752 | if (resolve_gitlink_ref(path, "HEAD", &oid) < 0) |
753 | return error(_("'%s' does not have a commit checked out"), path); | |
09595258 LT |
754 | while (namelen && path[namelen-1] == '/') |
755 | namelen--; | |
756 | } | |
a849735b | 757 | ce = make_empty_cache_entry(istate, namelen); |
11be42a4 | 758 | memcpy(ce->name, path, namelen); |
b60e188c | 759 | ce->ce_namelen = namelen; |
39425819 | 760 | if (!intent_only) |
d4c0a3ac | 761 | fill_stat_cache_info(istate, ce, st); |
388b2acd JH |
762 | else |
763 | ce->ce_flags |= CE_INTENT_TO_ADD; | |
11be42a4 | 764 | |
610d55af TG |
765 | |
766 | if (trust_executable_bit && has_symlinks) { | |
d177cab0 | 767 | ce->ce_mode = create_ce_mode(st_mode); |
610d55af | 768 | } else { |
78a8d641 JS |
769 | /* If there is an existing entry, pick the mode bits and type |
770 | * from it, otherwise assume unexecutable regular file. | |
11be42a4 | 771 | */ |
185c975f | 772 | struct cache_entry *ent; |
20314271 | 773 | int pos = index_name_pos_also_unmerged(istate, path, namelen); |
185c975f | 774 | |
4aab5b46 | 775 | ent = (0 <= pos) ? istate->cache[pos] : NULL; |
d177cab0 | 776 | ce->ce_mode = ce_mode_from_stat(ent, st_mode); |
11be42a4 JS |
777 | } |
778 | ||
dc1ae704 JJ |
779 | /* When core.ignorecase=true, determine if a directory of the same name but differing |
780 | * case already exists within the Git repository. If it does, ensure the directory | |
781 | * case of the file being added to the repository matches (is folded into) the existing | |
782 | * entry's directory case. | |
783 | */ | |
784 | if (ignore_case) { | |
41284eb0 | 785 | adjust_dirname_case(istate, ce->name); |
dc1ae704 | 786 | } |
e2c2a375 | 787 | if (!(flags & ADD_CACHE_RENORMALIZE)) { |
9472935d TB |
788 | alias = index_file_exists(istate, ce->name, |
789 | ce_namelen(ce), ignore_case); | |
790 | if (alias && | |
791 | !ce_stage(alias) && | |
792 | !ie_match_stat(istate, alias, st, ce_option)) { | |
793 | /* Nothing changed, really */ | |
794 | if (!S_ISGITLINK(alias->ce_mode)) | |
795 | ce_mark_uptodate(alias); | |
796 | alias->ce_flags |= CE_ADDED; | |
dc1ae704 | 797 | |
a849735b | 798 | discard_cache_entry(ce); |
9472935d TB |
799 | return 0; |
800 | } | |
0781b8a9 | 801 | } |
39425819 | 802 | if (!intent_only) { |
1c418243 | 803 | if (index_path(istate, &ce->oid, path, st, hash_flags)) { |
a849735b | 804 | discard_cache_entry(ce); |
9d0a9e90 | 805 | return error(_("unable to index file '%s'"), path); |
2d9426b0 | 806 | } |
39425819 | 807 | } else |
b4b313f9 | 808 | set_object_name_for_intent_to_add_entry(ce); |
39425819 | 809 | |
1102952b | 810 | if (ignore_case && alias && different_name(ce, alias)) |
045113a5 | 811 | ce = create_alias_ce(istate, ce, alias); |
1102952b | 812 | ce->ce_flags |= CE_ADDED; |
38ed1d89 | 813 | |
3bf0dd1f | 814 | /* It was suspected to be racily clean, but it turns out to be Ok */ |
38ed1d89 JH |
815 | was_same = (alias && |
816 | !ce_stage(alias) && | |
4a7e27e9 | 817 | oideq(&alias->oid, &ce->oid) && |
38ed1d89 JH |
818 | ce->ce_mode == alias->ce_mode); |
819 | ||
820 | if (pretend) | |
a849735b | 821 | discard_cache_entry(ce); |
067178ed | 822 | else if (add_index_entry(istate, ce, add_option)) { |
a849735b | 823 | discard_cache_entry(ce); |
9d0a9e90 | 824 | return error(_("unable to add '%s' to index"), path); |
067178ed | 825 | } |
38ed1d89 | 826 | if (verbose && !was_same) |
11be42a4 | 827 | printf("add '%s'\n", path); |
11be42a4 JS |
828 | return 0; |
829 | } | |
830 | ||
610d55af | 831 | int add_file_to_index(struct index_state *istate, const char *path, int flags) |
d177cab0 LT |
832 | { |
833 | struct stat st; | |
834 | if (lstat(path, &st)) | |
9d0a9e90 | 835 | die_errno(_("unable to stat '%s'"), path); |
610d55af | 836 | return add_to_index(istate, path, &st, flags); |
d177cab0 LT |
837 | } |
838 | ||
a849735b JM |
839 | struct cache_entry *make_empty_cache_entry(struct index_state *istate, size_t len) |
840 | { | |
8e72d675 | 841 | return mem_pool__ce_calloc(find_mem_pool(istate), len); |
a849735b JM |
842 | } |
843 | ||
96168827 MT |
844 | struct cache_entry *make_empty_transient_cache_entry(size_t len, |
845 | struct mem_pool *ce_mem_pool) | |
a849735b | 846 | { |
96168827 MT |
847 | if (ce_mem_pool) |
848 | return mem_pool__ce_calloc(ce_mem_pool, len); | |
a849735b JM |
849 | return xcalloc(1, cache_entry_size(len)); |
850 | } | |
851 | ||
2a1ae649 RS |
852 | enum verify_path_result { |
853 | PATH_OK, | |
854 | PATH_INVALID, | |
855 | PATH_DIR_WITH_SEP, | |
856 | }; | |
857 | ||
858 | static enum verify_path_result verify_path_internal(const char *, unsigned); | |
859 | ||
860 | int verify_path(const char *path, unsigned mode) | |
861 | { | |
c8ad9d04 | 862 | return verify_path_internal(path, mode) == PATH_OK; |
2a1ae649 RS |
863 | } |
864 | ||
a849735b JM |
865 | struct cache_entry *make_cache_entry(struct index_state *istate, |
866 | unsigned int mode, | |
825ed4d9 JM |
867 | const struct object_id *oid, |
868 | const char *path, | |
869 | int stage, | |
870 | unsigned int refresh_options) | |
6640f881 | 871 | { |
bc1c2caa | 872 | struct cache_entry *ce, *ret; |
a849735b | 873 | int len; |
6640f881 | 874 | |
c8ad9d04 | 875 | if (verify_path_internal(path, mode) == PATH_INVALID) { |
9d0a9e90 | 876 | error(_("invalid path '%s'"), path); |
6640f881 | 877 | return NULL; |
7e7abea9 | 878 | } |
6640f881 CR |
879 | |
880 | len = strlen(path); | |
a849735b | 881 | ce = make_empty_cache_entry(istate, len); |
6640f881 | 882 | |
825ed4d9 | 883 | oidcpy(&ce->oid, oid); |
6640f881 | 884 | memcpy(ce->name, path, len); |
b60e188c TG |
885 | ce->ce_flags = create_ce_flags(stage); |
886 | ce->ce_namelen = len; | |
6640f881 CR |
887 | ce->ce_mode = create_ce_mode(mode); |
888 | ||
d7b665c3 | 889 | ret = refresh_cache_entry(istate, ce, refresh_options); |
915e44c6 | 890 | if (ret != ce) |
a849735b | 891 | discard_cache_entry(ce); |
915e44c6 | 892 | return ret; |
6640f881 CR |
893 | } |
894 | ||
96168827 MT |
895 | struct cache_entry *make_transient_cache_entry(unsigned int mode, |
896 | const struct object_id *oid, | |
897 | const char *path, | |
898 | int stage, | |
899 | struct mem_pool *ce_mem_pool) | |
a849735b JM |
900 | { |
901 | struct cache_entry *ce; | |
902 | int len; | |
903 | ||
904 | if (!verify_path(path, mode)) { | |
9d0a9e90 | 905 | error(_("invalid path '%s'"), path); |
a849735b JM |
906 | return NULL; |
907 | } | |
908 | ||
909 | len = strlen(path); | |
96168827 | 910 | ce = make_empty_transient_cache_entry(len, ce_mem_pool); |
a849735b JM |
911 | |
912 | oidcpy(&ce->oid, oid); | |
913 | memcpy(ce->name, path, len); | |
914 | ce->ce_flags = create_ce_flags(stage); | |
915 | ce->ce_namelen = len; | |
916 | ce->ce_mode = create_ce_mode(mode); | |
917 | ||
918 | return ce; | |
919 | } | |
920 | ||
d9d70966 TG |
921 | /* |
922 | * Chmod an index entry with either +x or -x. | |
923 | * | |
924 | * Returns -1 if the chmod for the particular cache entry failed (if it's | |
925 | * not a regular file), -2 if an invalid flip argument is passed in, 0 | |
926 | * otherwise. | |
927 | */ | |
928 | int chmod_index_entry(struct index_state *istate, struct cache_entry *ce, | |
929 | char flip) | |
930 | { | |
931 | if (!S_ISREG(ce->ce_mode)) | |
932 | return -1; | |
933 | switch (flip) { | |
934 | case '+': | |
935 | ce->ce_mode |= 0111; | |
936 | break; | |
937 | case '-': | |
938 | ce->ce_mode &= ~0111; | |
939 | break; | |
940 | default: | |
941 | return -2; | |
942 | } | |
943 | cache_tree_invalidate_path(istate, ce->name); | |
944 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
883e248b | 945 | mark_fsmonitor_invalid(istate, ce); |
d9d70966 TG |
946 | istate->cache_changed |= CE_ENTRY_CHANGED; |
947 | ||
948 | return 0; | |
949 | } | |
950 | ||
9c5e6c80 | 951 | int ce_same_name(const struct cache_entry *a, const struct cache_entry *b) |
7b937ca3 LT |
952 | { |
953 | int len = ce_namelen(a); | |
954 | return ce_namelen(b) == len && !memcmp(a->name, b->name, len); | |
955 | } | |
956 | ||
8dcf39c4 LT |
957 | /* |
958 | * We fundamentally don't like some paths: we don't want | |
959 | * dot or dot-dot anywhere, and for obvious reasons don't | |
960 | * want to recurse into ".git" either. | |
961 | * | |
962 | * Also, we don't want double slashes or slashes at the | |
963 | * end that can make pathnames ambiguous. | |
964 | */ | |
10ecfa76 | 965 | static int verify_dotfile(const char *rest, unsigned mode) |
8dcf39c4 LT |
966 | { |
967 | /* | |
968 | * The first character was '.', but that | |
969 | * has already been discarded, we now test | |
970 | * the rest. | |
971 | */ | |
e0f530ff | 972 | |
8dcf39c4 | 973 | /* "." is not allowed */ |
e0f530ff | 974 | if (*rest == '\0' || is_dir_sep(*rest)) |
8dcf39c4 LT |
975 | return 0; |
976 | ||
e0f530ff | 977 | switch (*rest) { |
8dcf39c4 | 978 | /* |
641084b6 JK |
979 | * ".git" followed by NUL or slash is bad. Note that we match |
980 | * case-insensitively here, even if ignore_case is not set. | |
981 | * This outlaws ".GIT" everywhere out of an abundance of caution, | |
982 | * since there's really no good reason to allow it. | |
10ecfa76 JK |
983 | * |
984 | * Once we've seen ".git", we can also find ".gitmodules", etc (also | |
985 | * case-insensitively). | |
8dcf39c4 LT |
986 | */ |
987 | case 'g': | |
cc2fc7c2 JK |
988 | case 'G': |
989 | if (rest[1] != 'i' && rest[1] != 'I') | |
8dcf39c4 | 990 | break; |
cc2fc7c2 | 991 | if (rest[2] != 't' && rest[2] != 'T') |
8dcf39c4 | 992 | break; |
e19e5e66 JK |
993 | if (rest[3] == '\0' || is_dir_sep(rest[3])) |
994 | return 0; | |
10ecfa76 JK |
995 | if (S_ISLNK(mode)) { |
996 | rest += 3; | |
997 | if (skip_iprefix(rest, "modules", &rest) && | |
998 | (*rest == '\0' || is_dir_sep(*rest))) | |
999 | return 0; | |
1000 | } | |
e19e5e66 | 1001 | break; |
8dcf39c4 | 1002 | case '.': |
e0f530ff | 1003 | if (rest[1] == '\0' || is_dir_sep(rest[1])) |
8dcf39c4 LT |
1004 | return 0; |
1005 | } | |
1006 | return 1; | |
1007 | } | |
1008 | ||
2a1ae649 RS |
1009 | static enum verify_path_result verify_path_internal(const char *path, |
1010 | unsigned mode) | |
8dcf39c4 | 1011 | { |
49e268e2 | 1012 | char c = 0; |
8dcf39c4 | 1013 | |
56948cb6 | 1014 | if (has_dos_drive_prefix(path)) |
2a1ae649 | 1015 | return PATH_INVALID; |
56948cb6 | 1016 | |
d2c84dad | 1017 | if (!is_valid_path(path)) |
2a1ae649 | 1018 | return PATH_INVALID; |
d2c84dad | 1019 | |
8dcf39c4 LT |
1020 | goto inside; |
1021 | for (;;) { | |
1022 | if (!c) | |
2a1ae649 | 1023 | return PATH_OK; |
56948cb6 | 1024 | if (is_dir_sep(c)) { |
8dcf39c4 | 1025 | inside: |
10ecfa76 | 1026 | if (protect_hfs) { |
49e268e2 | 1027 | |
10ecfa76 | 1028 | if (is_hfs_dotgit(path)) |
2a1ae649 | 1029 | return PATH_INVALID; |
10ecfa76 JK |
1030 | if (S_ISLNK(mode)) { |
1031 | if (is_hfs_dotgitmodules(path)) | |
2a1ae649 | 1032 | return PATH_INVALID; |
10ecfa76 JK |
1033 | } |
1034 | } | |
1035 | if (protect_ntfs) { | |
bccc37fd | 1036 | #if defined GIT_WINDOWS_NATIVE || defined __CYGWIN__ |
49e268e2 | 1037 | if (c == '\\') |
2a1ae649 | 1038 | return PATH_INVALID; |
49e268e2 | 1039 | #endif |
10ecfa76 | 1040 | if (is_ntfs_dotgit(path)) |
2a1ae649 | 1041 | return PATH_INVALID; |
10ecfa76 JK |
1042 | if (S_ISLNK(mode)) { |
1043 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 1044 | return PATH_INVALID; |
10ecfa76 JK |
1045 | } |
1046 | } | |
1047 | ||
8dcf39c4 | 1048 | c = *path++; |
10ecfa76 | 1049 | if ((c == '.' && !verify_dotfile(path, mode)) || |
6e773527 | 1050 | is_dir_sep(c)) |
2a1ae649 | 1051 | return PATH_INVALID; |
6e773527 DS |
1052 | /* |
1053 | * allow terminating directory separators for | |
1054 | * sparse directory entries. | |
1055 | */ | |
1056 | if (c == '\0') | |
2a1ae649 RS |
1057 | return S_ISDIR(mode) ? PATH_DIR_WITH_SEP : |
1058 | PATH_INVALID; | |
288a74bc JS |
1059 | } else if (c == '\\' && protect_ntfs) { |
1060 | if (is_ntfs_dotgit(path)) | |
2a1ae649 | 1061 | return PATH_INVALID; |
288a74bc JS |
1062 | if (S_ISLNK(mode)) { |
1063 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 1064 | return PATH_INVALID; |
288a74bc | 1065 | } |
8dcf39c4 | 1066 | } |
288a74bc | 1067 | |
8dcf39c4 LT |
1068 | c = *path++; |
1069 | } | |
1070 | } | |
1071 | ||
12676608 LT |
1072 | /* |
1073 | * Do we have another file that has the beginning components being a | |
1074 | * proper superset of the name we're trying to add? | |
0f1e4f04 | 1075 | */ |
4aab5b46 JH |
1076 | static int has_file_name(struct index_state *istate, |
1077 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
0f1e4f04 | 1078 | { |
12676608 LT |
1079 | int retval = 0; |
1080 | int len = ce_namelen(ce); | |
b155725d | 1081 | int stage = ce_stage(ce); |
12676608 | 1082 | const char *name = ce->name; |
0f1e4f04 | 1083 | |
4aab5b46 JH |
1084 | while (pos < istate->cache_nr) { |
1085 | struct cache_entry *p = istate->cache[pos++]; | |
0f1e4f04 | 1086 | |
12676608 | 1087 | if (len >= ce_namelen(p)) |
0f1e4f04 | 1088 | break; |
12676608 LT |
1089 | if (memcmp(name, p->name, len)) |
1090 | break; | |
b155725d JH |
1091 | if (ce_stage(p) != stage) |
1092 | continue; | |
12676608 LT |
1093 | if (p->name[len] != '/') |
1094 | continue; | |
7a51ed66 | 1095 | if (p->ce_flags & CE_REMOVE) |
21cd8d00 | 1096 | continue; |
12676608 LT |
1097 | retval = -1; |
1098 | if (!ok_to_replace) | |
1099 | break; | |
4aab5b46 | 1100 | remove_index_entry_at(istate, --pos); |
0f1e4f04 | 1101 | } |
12676608 LT |
1102 | return retval; |
1103 | } | |
0f1e4f04 | 1104 | |
a6db3fbb JH |
1105 | |
1106 | /* | |
1107 | * Like strcmp(), but also return the offset of the first change. | |
1108 | * If strings are equal, return the length. | |
1109 | */ | |
1110 | int strcmp_offset(const char *s1, const char *s2, size_t *first_change) | |
1111 | { | |
1112 | size_t k; | |
1113 | ||
1114 | if (!first_change) | |
1115 | return strcmp(s1, s2); | |
1116 | ||
1117 | for (k = 0; s1[k] == s2[k]; k++) | |
1118 | if (s1[k] == '\0') | |
1119 | break; | |
1120 | ||
1121 | *first_change = k; | |
1122 | return (unsigned char)s1[k] - (unsigned char)s2[k]; | |
1123 | } | |
1124 | ||
12676608 LT |
1125 | /* |
1126 | * Do we have another file with a pathname that is a proper | |
1127 | * subset of the name we're trying to add? | |
06b6d81b JH |
1128 | * |
1129 | * That is, is there another file in the index with a path | |
1130 | * that matches a sub-directory in the given entry? | |
12676608 | 1131 | */ |
4aab5b46 JH |
1132 | static int has_dir_name(struct index_state *istate, |
1133 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
12676608 LT |
1134 | { |
1135 | int retval = 0; | |
b155725d | 1136 | int stage = ce_stage(ce); |
12676608 LT |
1137 | const char *name = ce->name; |
1138 | const char *slash = name + ce_namelen(ce); | |
06b6d81b JH |
1139 | size_t len_eq_last; |
1140 | int cmp_last = 0; | |
1141 | ||
1142 | /* | |
1143 | * We are frequently called during an iteration on a sorted | |
1144 | * list of pathnames and while building a new index. Therefore, | |
1145 | * there is a high probability that this entry will eventually | |
1146 | * be appended to the index, rather than inserted in the middle. | |
1147 | * If we can confirm that, we can avoid binary searches on the | |
1148 | * components of the pathname. | |
1149 | * | |
1150 | * Compare the entry's full path with the last path in the index. | |
1151 | */ | |
1152 | if (istate->cache_nr > 0) { | |
1153 | cmp_last = strcmp_offset(name, | |
1154 | istate->cache[istate->cache_nr - 1]->name, | |
1155 | &len_eq_last); | |
1156 | if (cmp_last > 0) { | |
1157 | if (len_eq_last == 0) { | |
1158 | /* | |
1159 | * The entry sorts AFTER the last one in the | |
1160 | * index and their paths have no common prefix, | |
1161 | * so there cannot be a F/D conflict. | |
1162 | */ | |
1163 | return retval; | |
1164 | } else { | |
1165 | /* | |
1166 | * The entry sorts AFTER the last one in the | |
1167 | * index, but has a common prefix. Fall through | |
1168 | * to the loop below to disect the entry's path | |
1169 | * and see where the difference is. | |
1170 | */ | |
1171 | } | |
1172 | } else if (cmp_last == 0) { | |
1173 | /* | |
1174 | * The entry exactly matches the last one in the | |
1175 | * index, but because of multiple stage and CE_REMOVE | |
1176 | * items, we fall through and let the regular search | |
1177 | * code handle it. | |
1178 | */ | |
1179 | } | |
1180 | } | |
0f1e4f04 | 1181 | |
12676608 | 1182 | for (;;) { |
b986df5c | 1183 | size_t len; |
0f1e4f04 | 1184 | |
12676608 LT |
1185 | for (;;) { |
1186 | if (*--slash == '/') | |
1187 | break; | |
1188 | if (slash <= ce->name) | |
1189 | return retval; | |
1190 | } | |
1191 | len = slash - name; | |
0f1e4f04 | 1192 | |
b986df5c JH |
1193 | if (cmp_last > 0) { |
1194 | /* | |
1195 | * (len + 1) is a directory boundary (including | |
1196 | * the trailing slash). And since the loop is | |
1197 | * decrementing "slash", the first iteration is | |
1198 | * the longest directory prefix; subsequent | |
1199 | * iterations consider parent directories. | |
1200 | */ | |
1201 | ||
1202 | if (len + 1 <= len_eq_last) { | |
1203 | /* | |
1204 | * The directory prefix (including the trailing | |
1205 | * slash) also appears as a prefix in the last | |
1206 | * entry, so the remainder cannot collide (because | |
1207 | * strcmp said the whole path was greater). | |
1208 | * | |
1209 | * EQ: last: xxx/A | |
1210 | * this: xxx/B | |
1211 | * | |
1212 | * LT: last: xxx/file_A | |
1213 | * this: xxx/file_B | |
1214 | */ | |
1215 | return retval; | |
1216 | } | |
1217 | ||
1218 | if (len > len_eq_last) { | |
1219 | /* | |
1220 | * This part of the directory prefix (excluding | |
1221 | * the trailing slash) is longer than the known | |
1222 | * equal portions, so this sub-directory cannot | |
1223 | * collide with a file. | |
1224 | * | |
1225 | * GT: last: xxxA | |
1226 | * this: xxxB/file | |
1227 | */ | |
1228 | return retval; | |
1229 | } | |
1230 | ||
b986df5c JH |
1231 | /* |
1232 | * This is a possible collision. Fall through and | |
1233 | * let the regular search code handle it. | |
1234 | * | |
1235 | * last: xxx | |
1236 | * this: xxx/file | |
1237 | */ | |
1238 | } | |
1239 | ||
b60e188c | 1240 | pos = index_name_stage_pos(istate, name, len, stage); |
12676608 | 1241 | if (pos >= 0) { |
21cd8d00 JH |
1242 | /* |
1243 | * Found one, but not so fast. This could | |
1244 | * be a marker that says "I was here, but | |
1245 | * I am being removed". Such an entry is | |
1246 | * not a part of the resulting tree, and | |
1247 | * it is Ok to have a directory at the same | |
1248 | * path. | |
1249 | */ | |
077c48df | 1250 | if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) { |
21cd8d00 JH |
1251 | retval = -1; |
1252 | if (!ok_to_replace) | |
1253 | break; | |
4aab5b46 | 1254 | remove_index_entry_at(istate, pos); |
21cd8d00 JH |
1255 | continue; |
1256 | } | |
12676608 | 1257 | } |
21cd8d00 JH |
1258 | else |
1259 | pos = -pos-1; | |
12676608 LT |
1260 | |
1261 | /* | |
1262 | * Trivial optimization: if we find an entry that | |
1263 | * already matches the sub-directory, then we know | |
b155725d | 1264 | * we're ok, and we can exit. |
12676608 | 1265 | */ |
4aab5b46 JH |
1266 | while (pos < istate->cache_nr) { |
1267 | struct cache_entry *p = istate->cache[pos]; | |
b155725d JH |
1268 | if ((ce_namelen(p) <= len) || |
1269 | (p->name[len] != '/') || | |
1270 | memcmp(p->name, name, len)) | |
1271 | break; /* not our subdirectory */ | |
077c48df JH |
1272 | if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE)) |
1273 | /* | |
1274 | * p is at the same stage as our entry, and | |
b155725d JH |
1275 | * is a subdirectory of what we are looking |
1276 | * at, so we cannot have conflicts at our | |
1277 | * level or anything shorter. | |
1278 | */ | |
1279 | return retval; | |
1280 | pos++; | |
192268c1 | 1281 | } |
0f1e4f04 | 1282 | } |
12676608 LT |
1283 | return retval; |
1284 | } | |
1285 | ||
1286 | /* We may be in a situation where we already have path/file and path | |
1287 | * is being added, or we already have path and path/file is being | |
1288 | * added. Either one would result in a nonsense tree that has path | |
1289 | * twice when git-write-tree tries to write it out. Prevent it. | |
a6080a0a | 1290 | * |
12676608 LT |
1291 | * If ok-to-replace is specified, we remove the conflicting entries |
1292 | * from the cache so the caller should recompute the insert position. | |
1293 | * When this happens, we return non-zero. | |
1294 | */ | |
4aab5b46 JH |
1295 | static int check_file_directory_conflict(struct index_state *istate, |
1296 | const struct cache_entry *ce, | |
1297 | int pos, int ok_to_replace) | |
12676608 | 1298 | { |
21cd8d00 JH |
1299 | int retval; |
1300 | ||
1301 | /* | |
1302 | * When ce is an "I am going away" entry, we allow it to be added | |
1303 | */ | |
7a51ed66 | 1304 | if (ce->ce_flags & CE_REMOVE) |
21cd8d00 JH |
1305 | return 0; |
1306 | ||
12676608 LT |
1307 | /* |
1308 | * We check if the path is a sub-path of a subsequent pathname | |
1309 | * first, since removing those will not change the position | |
21cd8d00 | 1310 | * in the array. |
12676608 | 1311 | */ |
4aab5b46 | 1312 | retval = has_file_name(istate, ce, pos, ok_to_replace); |
21cd8d00 | 1313 | |
12676608 LT |
1314 | /* |
1315 | * Then check if the path might have a clashing sub-directory | |
1316 | * before it. | |
1317 | */ | |
4aab5b46 | 1318 | return retval + has_dir_name(istate, ce, pos, ok_to_replace); |
0f1e4f04 JH |
1319 | } |
1320 | ||
af3785dc | 1321 | static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option) |
197ee8c9 LT |
1322 | { |
1323 | int pos; | |
192268c1 JH |
1324 | int ok_to_add = option & ADD_CACHE_OK_TO_ADD; |
1325 | int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE; | |
b155725d | 1326 | int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK; |
39425819 | 1327 | int new_only = option & ADD_CACHE_NEW_ONLY; |
5f73076c | 1328 | |
ce7c614b NTND |
1329 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1330 | cache_tree_invalidate_path(istate, ce->name); | |
e5494631 JH |
1331 | |
1332 | /* | |
1333 | * If this entry's path sorts after the last entry in the index, | |
1334 | * we can avoid searching for it. | |
1335 | */ | |
1336 | if (istate->cache_nr > 0 && | |
1337 | strcmp(ce->name, istate->cache[istate->cache_nr - 1]->name) > 0) | |
c097b95a | 1338 | pos = index_pos_to_insert_pos(istate->cache_nr); |
e5494631 JH |
1339 | else |
1340 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce)); | |
197ee8c9 | 1341 | |
3e09cdfd | 1342 | /* existing match? Just replace it. */ |
76e7f4ec | 1343 | if (pos >= 0) { |
39425819 JH |
1344 | if (!new_only) |
1345 | replace_index_entry(istate, pos, ce); | |
197ee8c9 LT |
1346 | return 0; |
1347 | } | |
76e7f4ec | 1348 | pos = -pos-1; |
197ee8c9 | 1349 | |
ffcc9ba7 NTND |
1350 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1351 | untracked_cache_add_to_index(istate, ce->name); | |
e931371a | 1352 | |
7b937ca3 LT |
1353 | /* |
1354 | * Inserting a merged entry ("stage 0") into the index | |
1355 | * will always replace all non-merged entries.. | |
1356 | */ | |
4aab5b46 JH |
1357 | if (pos < istate->cache_nr && ce_stage(ce) == 0) { |
1358 | while (ce_same_name(istate->cache[pos], ce)) { | |
7b937ca3 | 1359 | ok_to_add = 1; |
4aab5b46 | 1360 | if (!remove_index_entry_at(istate, pos)) |
7b937ca3 LT |
1361 | break; |
1362 | } | |
1363 | } | |
1364 | ||
121481ab LT |
1365 | if (!ok_to_add) |
1366 | return -1; | |
c8ad9d04 | 1367 | if (verify_path_internal(ce->name, ce->ce_mode) == PATH_INVALID) |
9d0a9e90 | 1368 | return error(_("invalid path '%s'"), ce->name); |
121481ab | 1369 | |
3e09cdfd | 1370 | if (!skip_df_check && |
4aab5b46 | 1371 | check_file_directory_conflict(istate, ce, pos, ok_to_replace)) { |
192268c1 | 1372 | if (!ok_to_replace) |
9d0a9e90 | 1373 | return error(_("'%s' appears as both a file and as a directory"), |
4aab5b46 | 1374 | ce->name); |
b60e188c | 1375 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce)); |
192268c1 JH |
1376 | pos = -pos-1; |
1377 | } | |
af3785dc JH |
1378 | return pos + 1; |
1379 | } | |
1380 | ||
1381 | int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option) | |
1382 | { | |
1383 | int pos; | |
1384 | ||
1385 | if (option & ADD_CACHE_JUST_APPEND) | |
1386 | pos = istate->cache_nr; | |
1387 | else { | |
1388 | int ret; | |
1389 | ret = add_index_entry_with_check(istate, ce, option); | |
1390 | if (ret <= 0) | |
1391 | return ret; | |
1392 | pos = ret - 1; | |
1393 | } | |
0f1e4f04 | 1394 | |
197ee8c9 | 1395 | /* Make sure the array is big enough .. */ |
999f5660 | 1396 | ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc); |
197ee8c9 LT |
1397 | |
1398 | /* Add it in.. */ | |
4aab5b46 | 1399 | istate->cache_nr++; |
af3785dc | 1400 | if (istate->cache_nr > pos + 1) |
f919ffeb SG |
1401 | MOVE_ARRAY(istate->cache + pos + 1, istate->cache + pos, |
1402 | istate->cache_nr - pos - 1); | |
cf558704 | 1403 | set_index_entry(istate, pos, ce); |
e636a7b4 | 1404 | istate->cache_changed |= CE_ENTRY_ADDED; |
197ee8c9 LT |
1405 | return 0; |
1406 | } | |
1407 | ||
405e5b2f LT |
1408 | /* |
1409 | * "refresh" does not calculate a new sha1 file or bring the | |
1410 | * cache up-to-date for mode/content changes. But what it | |
1411 | * _does_ do is to "re-match" the stat information of a file | |
1412 | * with the cache, so that you can refresh the cache for a | |
1413 | * file that hasn't been changed but where the stat entry is | |
1414 | * out of date. | |
1415 | * | |
1416 | * For example, you'd want to do this after doing a "git-read-tree", | |
1417 | * to link up the stat cache details with the proper files. | |
1418 | */ | |
4aab5b46 | 1419 | static struct cache_entry *refresh_cache_ent(struct index_state *istate, |
4bd5b7da | 1420 | struct cache_entry *ce, |
d05e6970 | 1421 | unsigned int options, int *err, |
a98e0f2d | 1422 | int *changed_ret, |
15268d12 JH |
1423 | int *t2_did_lstat, |
1424 | int *t2_did_scan) | |
405e5b2f LT |
1425 | { |
1426 | struct stat st; | |
1427 | struct cache_entry *updated; | |
a849735b | 1428 | int changed; |
25762726 | 1429 | int refresh = options & CE_MATCH_REFRESH; |
4bd5b7da | 1430 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 1431 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
2e2e7ec1 | 1432 | int ignore_missing = options & CE_MATCH_IGNORE_MISSING; |
883e248b | 1433 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
405e5b2f | 1434 | |
25762726 | 1435 | if (!refresh || ce_uptodate(ce)) |
eadb5831 JH |
1436 | return ce; |
1437 | ||
883e248b BP |
1438 | if (!ignore_fsmonitor) |
1439 | refresh_fsmonitor(istate); | |
aa9349d4 | 1440 | /* |
56cac48c NTND |
1441 | * CE_VALID or CE_SKIP_WORKTREE means the user promised us |
1442 | * that the change to the work tree does not matter and told | |
1443 | * us not to worry. | |
aa9349d4 | 1444 | */ |
56cac48c NTND |
1445 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) { |
1446 | ce_mark_uptodate(ce); | |
1447 | return ce; | |
1448 | } | |
aa9349d4 MSO |
1449 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) { |
1450 | ce_mark_uptodate(ce); | |
1451 | return ce; | |
1452 | } | |
883e248b BP |
1453 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) { |
1454 | ce_mark_uptodate(ce); | |
1455 | return ce; | |
1456 | } | |
aa9349d4 | 1457 | |
ccad42d4 RS |
1458 | if (has_symlink_leading_path(ce->name, ce_namelen(ce))) { |
1459 | if (ignore_missing) | |
1460 | return ce; | |
1461 | if (err) | |
1462 | *err = ENOENT; | |
1463 | return NULL; | |
1464 | } | |
1465 | ||
a98e0f2d JH |
1466 | if (t2_did_lstat) |
1467 | *t2_did_lstat = 1; | |
8fd2cb40 | 1468 | if (lstat(ce->name, &st) < 0) { |
2e2e7ec1 BK |
1469 | if (ignore_missing && errno == ENOENT) |
1470 | return ce; | |
ec0cc704 JH |
1471 | if (err) |
1472 | *err = errno; | |
8fd2cb40 JS |
1473 | return NULL; |
1474 | } | |
405e5b2f | 1475 | |
4bd5b7da | 1476 | changed = ie_match_stat(istate, ce, &st, options); |
d05e6970 JK |
1477 | if (changed_ret) |
1478 | *changed_ret = changed; | |
405e5b2f | 1479 | if (!changed) { |
4bd5b7da JH |
1480 | /* |
1481 | * The path is unchanged. If we were told to ignore | |
1482 | * valid bit, then we did the actual stat check and | |
1483 | * found that the entry is unmodified. If the entry | |
1484 | * is not marked VALID, this is the place to mark it | |
1485 | * valid again, under "assume unchanged" mode. | |
1486 | */ | |
1487 | if (ignore_valid && assume_unchanged && | |
7a51ed66 | 1488 | !(ce->ce_flags & CE_VALID)) |
405e5b2f | 1489 | ; /* mark this one VALID again */ |
eadb5831 JH |
1490 | else { |
1491 | /* | |
1492 | * We do not mark the index itself "modified" | |
1493 | * because CE_UPTODATE flag is in-core only; | |
1494 | * we are not going to write this change out. | |
1495 | */ | |
883e248b | 1496 | if (!S_ISGITLINK(ce->ce_mode)) { |
125fd984 | 1497 | ce_mark_uptodate(ce); |
b5a81697 | 1498 | mark_fsmonitor_valid(istate, ce); |
883e248b | 1499 | } |
8fd2cb40 | 1500 | return ce; |
eadb5831 | 1501 | } |
405e5b2f LT |
1502 | } |
1503 | ||
15268d12 JH |
1504 | if (t2_did_scan) |
1505 | *t2_did_scan = 1; | |
4bd5b7da | 1506 | if (ie_modified(istate, ce, &st, options)) { |
ec0cc704 JH |
1507 | if (err) |
1508 | *err = EINVAL; | |
8fd2cb40 JS |
1509 | return NULL; |
1510 | } | |
405e5b2f | 1511 | |
a849735b | 1512 | updated = make_empty_cache_entry(istate, ce_namelen(ce)); |
0e267b7a BP |
1513 | copy_cache_entry(updated, ce); |
1514 | memcpy(updated->name, ce->name, ce->ce_namelen + 1); | |
d4c0a3ac | 1515 | fill_stat_cache_info(istate, updated, &st); |
4bd5b7da JH |
1516 | /* |
1517 | * If ignore_valid is not set, we should leave CE_VALID bit | |
1518 | * alone. Otherwise, paths marked with --no-assume-unchanged | |
1519 | * (i.e. things to be edited) will reacquire CE_VALID bit | |
1520 | * automatically, which is not really what we want. | |
405e5b2f | 1521 | */ |
4bd5b7da | 1522 | if (!ignore_valid && assume_unchanged && |
7a51ed66 LT |
1523 | !(ce->ce_flags & CE_VALID)) |
1524 | updated->ce_flags &= ~CE_VALID; | |
405e5b2f | 1525 | |
e636a7b4 | 1526 | /* istate->cache_changed is updated in the caller */ |
405e5b2f LT |
1527 | return updated; |
1528 | } | |
1529 | ||
3deffc52 | 1530 | static void show_file(const char * fmt, const char * name, int in_porcelain, |
046613c5 | 1531 | int * first, const char *header_msg) |
3deffc52 MM |
1532 | { |
1533 | if (in_porcelain && *first && header_msg) { | |
1534 | printf("%s\n", header_msg); | |
cd2b8ae9 | 1535 | *first = 0; |
3deffc52 MM |
1536 | } |
1537 | printf(fmt, name); | |
1538 | } | |
1539 | ||
22184497 TG |
1540 | int repo_refresh_and_write_index(struct repository *repo, |
1541 | unsigned int refresh_flags, | |
1542 | unsigned int write_flags, | |
1543 | int gentle, | |
1544 | const struct pathspec *pathspec, | |
1545 | char *seen, const char *header_msg) | |
1546 | { | |
1547 | struct lock_file lock_file = LOCK_INIT; | |
1548 | int fd, ret = 0; | |
1549 | ||
1550 | fd = repo_hold_locked_index(repo, &lock_file, 0); | |
1551 | if (!gentle && fd < 0) | |
1552 | return -1; | |
1553 | if (refresh_index(repo->index, refresh_flags, pathspec, seen, header_msg)) | |
1554 | ret = 1; | |
1555 | if (0 <= fd && write_locked_index(repo->index, &lock_file, COMMIT_LOCK | write_flags)) | |
1556 | ret = -1; | |
1557 | return ret; | |
1558 | } | |
1559 | ||
1560 | ||
9b2d6149 NTND |
1561 | int refresh_index(struct index_state *istate, unsigned int flags, |
1562 | const struct pathspec *pathspec, | |
046613c5 | 1563 | char *seen, const char *header_msg) |
405e5b2f LT |
1564 | { |
1565 | int i; | |
1566 | int has_errors = 0; | |
1567 | int really = (flags & REFRESH_REALLY) != 0; | |
1568 | int allow_unmerged = (flags & REFRESH_UNMERGED) != 0; | |
1569 | int quiet = (flags & REFRESH_QUIET) != 0; | |
1570 | int not_new = (flags & REFRESH_IGNORE_MISSING) != 0; | |
5fdeacb0 | 1571 | int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0; |
b243012c | 1572 | int ignore_skip_worktree = (flags & REFRESH_IGNORE_SKIP_WORKTREE) != 0; |
3deffc52 MM |
1573 | int first = 1; |
1574 | int in_porcelain = (flags & REFRESH_IN_PORCELAIN); | |
25762726 BK |
1575 | unsigned int options = (CE_MATCH_REFRESH | |
1576 | (really ? CE_MATCH_IGNORE_VALID : 0) | | |
2e2e7ec1 | 1577 | (not_new ? CE_MATCH_IGNORE_MISSING : 0)); |
4bd4e730 | 1578 | const char *modified_fmt; |
73b7eae6 JK |
1579 | const char *deleted_fmt; |
1580 | const char *typechange_fmt; | |
1581 | const char *added_fmt; | |
4bd4e730 | 1582 | const char *unmerged_fmt; |
ae9af122 | 1583 | struct progress *progress = NULL; |
a98e0f2d | 1584 | int t2_sum_lstat = 0; |
15268d12 | 1585 | int t2_sum_scan = 0; |
ae9af122 NTND |
1586 | |
1587 | if (flags & REFRESH_PROGRESS && isatty(2)) | |
1588 | progress = start_delayed_progress(_("Refresh index"), | |
1589 | istate->cache_nr); | |
405e5b2f | 1590 | |
c46c406a | 1591 | trace_performance_enter(); |
a71806a7 NTND |
1592 | modified_fmt = in_porcelain ? "M\t%s\n" : "%s: needs update\n"; |
1593 | deleted_fmt = in_porcelain ? "D\t%s\n" : "%s: needs update\n"; | |
1594 | typechange_fmt = in_porcelain ? "T\t%s\n" : "%s: needs update\n"; | |
1595 | added_fmt = in_porcelain ? "A\t%s\n" : "%s: needs update\n"; | |
1596 | unmerged_fmt = in_porcelain ? "U\t%s\n" : "%s: needs merge\n"; | |
99ce720c BP |
1597 | /* |
1598 | * Use the multi-threaded preload_index() to refresh most of the | |
1599 | * cache entries quickly then in the single threaded loop below, | |
1600 | * we only have to do the special cases that are left. | |
1601 | */ | |
1602 | preload_index(istate, pathspec, 0); | |
a98e0f2d | 1603 | trace2_region_enter("index", "refresh", NULL); |
d76723ee | 1604 | |
4aab5b46 | 1605 | for (i = 0; i < istate->cache_nr; i++) { |
285c2e25 | 1606 | struct cache_entry *ce, *new_entry; |
ec0cc704 | 1607 | int cache_errno = 0; |
73b7eae6 | 1608 | int changed = 0; |
3d1f148c | 1609 | int filtered = 0; |
a98e0f2d | 1610 | int t2_did_lstat = 0; |
15268d12 | 1611 | int t2_did_scan = 0; |
ec0cc704 | 1612 | |
4aab5b46 | 1613 | ce = istate->cache[i]; |
5fdeacb0 JS |
1614 | if (ignore_submodules && S_ISGITLINK(ce->ce_mode)) |
1615 | continue; | |
b243012c MT |
1616 | if (ignore_skip_worktree && ce_skip_worktree(ce)) |
1617 | continue; | |
5fdeacb0 | 1618 | |
d76723ee DS |
1619 | /* |
1620 | * If this entry is a sparse directory, then there isn't | |
1621 | * any stat() information to update. Ignore the entry. | |
1622 | */ | |
1623 | if (S_ISSPARSEDIR(ce->ce_mode)) | |
1624 | continue; | |
1625 | ||
d7b665c3 | 1626 | if (pathspec && !ce_path_match(istate, ce, pathspec, seen)) |
3d1f148c JH |
1627 | filtered = 1; |
1628 | ||
405e5b2f | 1629 | if (ce_stage(ce)) { |
4aab5b46 JH |
1630 | while ((i < istate->cache_nr) && |
1631 | ! strcmp(istate->cache[i]->name, ce->name)) | |
405e5b2f LT |
1632 | i++; |
1633 | i--; | |
1634 | if (allow_unmerged) | |
1635 | continue; | |
3d1f148c JH |
1636 | if (!filtered) |
1637 | show_file(unmerged_fmt, ce->name, in_porcelain, | |
1638 | &first, header_msg); | |
405e5b2f LT |
1639 | has_errors = 1; |
1640 | continue; | |
1641 | } | |
1642 | ||
3d1f148c | 1643 | if (filtered) |
d616813d AJ |
1644 | continue; |
1645 | ||
a98e0f2d JH |
1646 | new_entry = refresh_cache_ent(istate, ce, options, |
1647 | &cache_errno, &changed, | |
15268d12 | 1648 | &t2_did_lstat, &t2_did_scan); |
a98e0f2d | 1649 | t2_sum_lstat += t2_did_lstat; |
15268d12 | 1650 | t2_sum_scan += t2_did_scan; |
285c2e25 | 1651 | if (new_entry == ce) |
405e5b2f | 1652 | continue; |
b7b793d1 | 1653 | display_progress(progress, i); |
285c2e25 | 1654 | if (!new_entry) { |
73b7eae6 JK |
1655 | const char *fmt; |
1656 | ||
8fd2cb40 | 1657 | if (really && cache_errno == EINVAL) { |
405e5b2f LT |
1658 | /* If we are doing --really-refresh that |
1659 | * means the index is not valid anymore. | |
1660 | */ | |
7a51ed66 | 1661 | ce->ce_flags &= ~CE_VALID; |
078a58e8 | 1662 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 1663 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 1664 | istate->cache_changed |= CE_ENTRY_CHANGED; |
405e5b2f LT |
1665 | } |
1666 | if (quiet) | |
1667 | continue; | |
73b7eae6 JK |
1668 | |
1669 | if (cache_errno == ENOENT) | |
1670 | fmt = deleted_fmt; | |
895ff3b2 | 1671 | else if (ce_intent_to_add(ce)) |
73b7eae6 JK |
1672 | fmt = added_fmt; /* must be before other checks */ |
1673 | else if (changed & TYPE_CHANGED) | |
1674 | fmt = typechange_fmt; | |
1675 | else | |
1676 | fmt = modified_fmt; | |
1677 | show_file(fmt, | |
1678 | ce->name, in_porcelain, &first, header_msg); | |
405e5b2f LT |
1679 | has_errors = 1; |
1680 | continue; | |
1681 | } | |
cf558704 | 1682 | |
285c2e25 | 1683 | replace_index_entry(istate, i, new_entry); |
405e5b2f | 1684 | } |
a98e0f2d | 1685 | trace2_data_intmax("index", NULL, "refresh/sum_lstat", t2_sum_lstat); |
15268d12 | 1686 | trace2_data_intmax("index", NULL, "refresh/sum_scan", t2_sum_scan); |
a98e0f2d | 1687 | trace2_region_leave("index", "refresh", NULL); |
b7b793d1 ÆAB |
1688 | display_progress(progress, istate->cache_nr); |
1689 | stop_progress(&progress); | |
c46c406a | 1690 | trace_performance_leave("refresh index"); |
405e5b2f LT |
1691 | return has_errors; |
1692 | } | |
1693 | ||
768d7965 JM |
1694 | struct cache_entry *refresh_cache_entry(struct index_state *istate, |
1695 | struct cache_entry *ce, | |
1696 | unsigned int options) | |
ec0cc704 | 1697 | { |
15268d12 | 1698 | return refresh_cache_ent(istate, ce, options, NULL, NULL, NULL, NULL); |
ec0cc704 JH |
1699 | } |
1700 | ||
db3b313c JH |
1701 | |
1702 | /***************************************************************** | |
1703 | * Index File I/O | |
1704 | *****************************************************************/ | |
1705 | ||
9d227781 JH |
1706 | #define INDEX_FORMAT_DEFAULT 3 |
1707 | ||
7211b9e7 | 1708 | static unsigned int get_index_format_default(struct repository *r) |
136347d7 TG |
1709 | { |
1710 | char *envversion = getenv("GIT_INDEX_VERSION"); | |
3c09d684 TG |
1711 | char *endp; |
1712 | unsigned int version = INDEX_FORMAT_DEFAULT; | |
1713 | ||
136347d7 | 1714 | if (!envversion) { |
7211b9e7 DS |
1715 | prepare_repo_settings(r); |
1716 | ||
1717 | if (r->settings.index_version >= 0) | |
1718 | version = r->settings.index_version; | |
3c09d684 TG |
1719 | if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { |
1720 | warning(_("index.version set, but the value is invalid.\n" | |
136347d7 | 1721 | "Using version %i"), INDEX_FORMAT_DEFAULT); |
3c09d684 | 1722 | return INDEX_FORMAT_DEFAULT; |
136347d7 TG |
1723 | } |
1724 | return version; | |
1725 | } | |
3c09d684 TG |
1726 | |
1727 | version = strtoul(envversion, &endp, 10); | |
1728 | if (*endp || | |
1729 | version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { | |
1730 | warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n" | |
1731 | "Using version %i"), INDEX_FORMAT_DEFAULT); | |
1732 | version = INDEX_FORMAT_DEFAULT; | |
1733 | } | |
1734 | return version; | |
136347d7 TG |
1735 | } |
1736 | ||
db3b313c JH |
1737 | /* |
1738 | * dev/ino/uid/gid/size are also just tracked to the low 32 bits | |
1739 | * Again - this is just a (very strong in practice) heuristic that | |
1740 | * the inode hasn't changed. | |
1741 | * | |
1742 | * We save the fields in big-endian order to allow using the | |
1743 | * index file over NFS transparently. | |
1744 | */ | |
1745 | struct ondisk_cache_entry { | |
1746 | struct cache_time ctime; | |
1747 | struct cache_time mtime; | |
7800c1eb TG |
1748 | uint32_t dev; |
1749 | uint32_t ino; | |
1750 | uint32_t mode; | |
1751 | uint32_t uid; | |
1752 | uint32_t gid; | |
1753 | uint32_t size; | |
575fa8a3 | 1754 | /* |
1755 | * unsigned char hash[hashsz]; | |
1756 | * uint16_t flags; | |
1757 | * if (flags & CE_EXTENDED) | |
1758 | * uint16_t flags2; | |
1759 | */ | |
1760 | unsigned char data[GIT_MAX_RAWSZ + 2 * sizeof(uint16_t)]; | |
1761 | char name[FLEX_ARRAY]; | |
db3b313c JH |
1762 | }; |
1763 | ||
6c9cd161 | 1764 | /* These are only used for v3 or lower */ |
ce012deb | 1765 | #define align_padding_size(size, len) ((size + (len) + 8) & ~7) - (size + len) |
575fa8a3 | 1766 | #define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,data) + (len) + 8) & ~7) |
db3b313c | 1767 | #define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len) |
575fa8a3 | 1768 | #define ondisk_data_size(flags, len) (the_hash_algo->rawsz + \ |
1769 | ((flags & CE_EXTENDED) ? 2 : 1) * sizeof(uint16_t) + len) | |
1770 | #define ondisk_data_size_max(len) (ondisk_data_size(CE_EXTENDED, len)) | |
1771 | #define ondisk_ce_size(ce) (ondisk_cache_entry_size(ondisk_data_size((ce)->ce_flags, ce_namelen(ce)))) | |
db3b313c | 1772 | |
a33fc72f JH |
1773 | /* Allow fsck to force verification of the index checksum. */ |
1774 | int verify_index_checksum; | |
1775 | ||
00ec50e5 BP |
1776 | /* Allow fsck to force verification of the cache entry order. */ |
1777 | int verify_ce_order; | |
1778 | ||
371ed0de | 1779 | static int verify_hdr(const struct cache_header *hdr, unsigned long size) |
e83c5163 | 1780 | { |
aab61359 | 1781 | git_hash_ctx c; |
1782 | unsigned char hash[GIT_MAX_RAWSZ]; | |
0136bac9 | 1783 | int hdr_version; |
e83c5163 | 1784 | |
ccc4feb5 | 1785 | if (hdr->hdr_signature != htonl(CACHE_SIGNATURE)) |
9d0a9e90 | 1786 | return error(_("bad signature 0x%08x"), hdr->hdr_signature); |
0136bac9 | 1787 | hdr_version = ntohl(hdr->hdr_version); |
b82a7b5b | 1788 | if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version) |
9d0a9e90 | 1789 | return error(_("bad index version %d"), hdr_version); |
a33fc72f JH |
1790 | |
1791 | if (!verify_index_checksum) | |
1792 | return 0; | |
1793 | ||
aab61359 | 1794 | the_hash_algo->init_fn(&c); |
1795 | the_hash_algo->update_fn(&c, hdr, size - the_hash_algo->rawsz); | |
1796 | the_hash_algo->final_fn(hash, &c); | |
67947c34 | 1797 | if (!hasheq(hash, (unsigned char *)hdr + size - the_hash_algo->rawsz)) |
9d0a9e90 | 1798 | return error(_("bad index file sha1 signature")); |
e83c5163 LT |
1799 | return 0; |
1800 | } | |
1801 | ||
4aab5b46 | 1802 | static int read_index_extension(struct index_state *istate, |
371ed0de | 1803 | const char *ext, const char *data, unsigned long sz) |
bad68ec9 JH |
1804 | { |
1805 | switch (CACHE_EXT(ext)) { | |
1806 | case CACHE_EXT_TREE: | |
4aab5b46 | 1807 | istate->cache_tree = cache_tree_read(data, sz); |
bad68ec9 | 1808 | break; |
cfc5789a JH |
1809 | case CACHE_EXT_RESOLVE_UNDO: |
1810 | istate->resolve_undo = resolve_undo_read(data, sz); | |
1811 | break; | |
5fc2fc8f NTND |
1812 | case CACHE_EXT_LINK: |
1813 | if (read_link_extension(istate, data, sz)) | |
1814 | return -1; | |
1815 | break; | |
f9e6c649 NTND |
1816 | case CACHE_EXT_UNTRACKED: |
1817 | istate->untracked = read_untracked_extension(data, sz); | |
1818 | break; | |
883e248b BP |
1819 | case CACHE_EXT_FSMONITOR: |
1820 | read_fsmonitor_extension(istate, data, sz); | |
1821 | break; | |
3b1d9e04 | 1822 | case CACHE_EXT_ENDOFINDEXENTRIES: |
3255089a | 1823 | case CACHE_EXT_INDEXENTRYOFFSETTABLE: |
3b1d9e04 BP |
1824 | /* already handled in do_read_index() */ |
1825 | break; | |
cd42415f DS |
1826 | case CACHE_EXT_SPARSE_DIRECTORIES: |
1827 | /* no content, only an indicator */ | |
1828 | istate->sparse_index = 1; | |
1829 | break; | |
bad68ec9 JH |
1830 | default: |
1831 | if (*ext < 'A' || 'Z' < *ext) | |
9d0a9e90 | 1832 | return error(_("index uses %.4s extension, which we do not understand"), |
bad68ec9 | 1833 | ext); |
9d0a9e90 | 1834 | fprintf_ln(stderr, _("ignoring %.4s extension"), ext); |
bad68ec9 JH |
1835 | break; |
1836 | } | |
1837 | return 0; | |
1838 | } | |
1839 | ||
77ff1127 BP |
1840 | static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool, |
1841 | unsigned int version, | |
a849735b | 1842 | struct ondisk_cache_entry *ondisk, |
6c9cd161 | 1843 | unsigned long *ent_size, |
252d079c | 1844 | const struct cache_entry *previous_ce) |
7a51ed66 | 1845 | { |
debed2a6 | 1846 | struct cache_entry *ce; |
7fec10b7 | 1847 | size_t len; |
06aaaa0b | 1848 | const char *name; |
575fa8a3 | 1849 | const unsigned hashsz = the_hash_algo->rawsz; |
1850 | const uint16_t *flagsp = (const uint16_t *)(ondisk->data + hashsz); | |
debed2a6 | 1851 | unsigned int flags; |
f5c4a9af | 1852 | size_t copy_len = 0; |
252d079c NTND |
1853 | /* |
1854 | * Adjacent cache entries tend to share the leading paths, so it makes | |
1855 | * sense to only store the differences in later entries. In the v4 | |
1856 | * on-disk format of the index, each on-disk cache entry stores the | |
1857 | * number of bytes to be stripped from the end of the previous name, | |
1858 | * and the bytes to append to the result, to come up with its name. | |
1859 | */ | |
77ff1127 | 1860 | int expand_name_field = version == 4; |
7fec10b7 | 1861 | |
7a51ed66 | 1862 | /* On-disk flags are just 16 bits */ |
575fa8a3 | 1863 | flags = get_be16(flagsp); |
debed2a6 | 1864 | len = flags & CE_NAMEMASK; |
7fec10b7 | 1865 | |
debed2a6 | 1866 | if (flags & CE_EXTENDED) { |
06aaaa0b | 1867 | int extended_flags; |
575fa8a3 | 1868 | extended_flags = get_be16(flagsp + 1) << 16; |
06aaaa0b NTND |
1869 | /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */ |
1870 | if (extended_flags & ~CE_EXTENDED_FLAGS) | |
9d0a9e90 | 1871 | die(_("unknown index entry format 0x%08x"), extended_flags); |
debed2a6 | 1872 | flags |= extended_flags; |
575fa8a3 | 1873 | name = (const char *)(flagsp + 2); |
06aaaa0b NTND |
1874 | } |
1875 | else | |
575fa8a3 | 1876 | name = (const char *)(flagsp + 1); |
06aaaa0b | 1877 | |
252d079c NTND |
1878 | if (expand_name_field) { |
1879 | const unsigned char *cp = (const unsigned char *)name; | |
1880 | size_t strip_len, previous_len; | |
6c9cd161 | 1881 | |
15beaaa3 | 1882 | /* If we're at the beginning of a block, ignore the previous name */ |
252d079c | 1883 | strip_len = decode_varint(&cp); |
77ff1127 BP |
1884 | if (previous_ce) { |
1885 | previous_len = previous_ce->ce_namelen; | |
1886 | if (previous_len < strip_len) | |
252d079c | 1887 | die(_("malformed name field in the index, near path '%s'"), |
77ff1127 BP |
1888 | previous_ce->name); |
1889 | copy_len = previous_len - strip_len; | |
252d079c | 1890 | } |
252d079c NTND |
1891 | name = (const char *)cp; |
1892 | } | |
6c9cd161 | 1893 | |
252d079c NTND |
1894 | if (len == CE_NAMEMASK) { |
1895 | len = strlen(name); | |
1896 | if (expand_name_field) | |
1897 | len += copy_len; | |
1898 | } | |
6c9cd161 | 1899 | |
77ff1127 | 1900 | ce = mem_pool__ce_alloc(ce_mem_pool, len); |
252d079c NTND |
1901 | |
1902 | ce->ce_stat_data.sd_ctime.sec = get_be32(&ondisk->ctime.sec); | |
1903 | ce->ce_stat_data.sd_mtime.sec = get_be32(&ondisk->mtime.sec); | |
1904 | ce->ce_stat_data.sd_ctime.nsec = get_be32(&ondisk->ctime.nsec); | |
1905 | ce->ce_stat_data.sd_mtime.nsec = get_be32(&ondisk->mtime.nsec); | |
1906 | ce->ce_stat_data.sd_dev = get_be32(&ondisk->dev); | |
1907 | ce->ce_stat_data.sd_ino = get_be32(&ondisk->ino); | |
1908 | ce->ce_mode = get_be32(&ondisk->mode); | |
1909 | ce->ce_stat_data.sd_uid = get_be32(&ondisk->uid); | |
1910 | ce->ce_stat_data.sd_gid = get_be32(&ondisk->gid); | |
1911 | ce->ce_stat_data.sd_size = get_be32(&ondisk->size); | |
1912 | ce->ce_flags = flags & ~CE_NAMEMASK; | |
1913 | ce->ce_namelen = len; | |
1914 | ce->index = 0; | |
92e2cab9 | 1915 | oidread(&ce->oid, ondisk->data); |
575fa8a3 | 1916 | memcpy(ce->name, name, len); |
1917 | ce->name[len] = '\0'; | |
6c9cd161 | 1918 | |
252d079c NTND |
1919 | if (expand_name_field) { |
1920 | if (copy_len) | |
1921 | memcpy(ce->name, previous_ce->name, copy_len); | |
1922 | memcpy(ce->name + copy_len, name, len + 1 - copy_len); | |
1923 | *ent_size = (name - ((char *)ondisk)) + len + 1 - copy_len; | |
1924 | } else { | |
1925 | memcpy(ce->name, name, len + 1); | |
1926 | *ent_size = ondisk_ce_size(ce); | |
6c9cd161 | 1927 | } |
debed2a6 | 1928 | return ce; |
cf558704 LT |
1929 | } |
1930 | ||
03f15a79 | 1931 | static void check_ce_order(struct index_state *istate) |
15999d0b | 1932 | { |
03f15a79 TG |
1933 | unsigned int i; |
1934 | ||
00ec50e5 BP |
1935 | if (!verify_ce_order) |
1936 | return; | |
1937 | ||
03f15a79 TG |
1938 | for (i = 1; i < istate->cache_nr; i++) { |
1939 | struct cache_entry *ce = istate->cache[i - 1]; | |
1940 | struct cache_entry *next_ce = istate->cache[i]; | |
1941 | int name_compare = strcmp(ce->name, next_ce->name); | |
1942 | ||
1943 | if (0 < name_compare) | |
9d0a9e90 | 1944 | die(_("unordered stage entries in index")); |
03f15a79 TG |
1945 | if (!name_compare) { |
1946 | if (!ce_stage(ce)) | |
9d0a9e90 | 1947 | die(_("multiple stage entries for merged file '%s'"), |
03f15a79 TG |
1948 | ce->name); |
1949 | if (ce_stage(ce) > ce_stage(next_ce)) | |
9d0a9e90 | 1950 | die(_("unordered stage entries for '%s'"), |
03f15a79 TG |
1951 | ce->name); |
1952 | } | |
15999d0b JSP |
1953 | } |
1954 | } | |
1955 | ||
435ec090 CC |
1956 | static void tweak_untracked_cache(struct index_state *istate) |
1957 | { | |
ad0fb659 DS |
1958 | struct repository *r = the_repository; |
1959 | ||
1960 | prepare_repo_settings(r); | |
1961 | ||
f1bee828 ÆAB |
1962 | switch (r->settings.core_untracked_cache) { |
1963 | case UNTRACKED_CACHE_REMOVE: | |
435ec090 | 1964 | remove_untracked_cache(istate); |
f1bee828 ÆAB |
1965 | break; |
1966 | case UNTRACKED_CACHE_WRITE: | |
ad0fb659 | 1967 | add_untracked_cache(istate); |
f1bee828 ÆAB |
1968 | break; |
1969 | case UNTRACKED_CACHE_KEEP: | |
3050b6df ÆAB |
1970 | /* |
1971 | * Either an explicit "core.untrackedCache=keep", the | |
1972 | * default if "core.untrackedCache" isn't configured, | |
1973 | * or a fallback on an unknown "core.untrackedCache" | |
1974 | * value. | |
1975 | */ | |
f1bee828 | 1976 | break; |
f1bee828 | 1977 | } |
435ec090 CC |
1978 | } |
1979 | ||
43925312 CC |
1980 | static void tweak_split_index(struct index_state *istate) |
1981 | { | |
1982 | switch (git_config_get_split_index()) { | |
1983 | case -1: /* unset: do nothing */ | |
1984 | break; | |
1985 | case 0: /* false */ | |
1986 | remove_split_index(istate); | |
1987 | break; | |
1988 | case 1: /* true */ | |
1989 | add_split_index(istate); | |
1990 | break; | |
1991 | default: /* unknown value: do nothing */ | |
1992 | break; | |
1993 | } | |
1994 | } | |
1995 | ||
435ec090 CC |
1996 | static void post_read_index_from(struct index_state *istate) |
1997 | { | |
1998 | check_ce_order(istate); | |
1999 | tweak_untracked_cache(istate); | |
43925312 | 2000 | tweak_split_index(istate); |
883e248b | 2001 | tweak_fsmonitor(istate); |
435ec090 CC |
2002 | } |
2003 | ||
8e72d675 JM |
2004 | static size_t estimate_cache_size_from_compressed(unsigned int entries) |
2005 | { | |
2006 | return entries * (sizeof(struct cache_entry) + CACHE_ENTRY_PATH_LENGTH); | |
2007 | } | |
2008 | ||
2009 | static size_t estimate_cache_size(size_t ondisk_size, unsigned int entries) | |
2010 | { | |
2011 | long per_entry = sizeof(struct cache_entry) - sizeof(struct ondisk_cache_entry); | |
2012 | ||
2013 | /* | |
2014 | * Account for potential alignment differences. | |
2015 | */ | |
c097b95a | 2016 | per_entry += align_padding_size(per_entry, 0); |
8e72d675 JM |
2017 | return ondisk_size + entries * per_entry; |
2018 | } | |
2019 | ||
3255089a BP |
2020 | struct index_entry_offset |
2021 | { | |
2022 | /* starting byte offset into index file, count of index entries in this block */ | |
2023 | int offset, nr; | |
2024 | }; | |
2025 | ||
2026 | struct index_entry_offset_table | |
2027 | { | |
2028 | int nr; | |
2029 | struct index_entry_offset entries[FLEX_ARRAY]; | |
2030 | }; | |
2031 | ||
3255089a BP |
2032 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset); |
2033 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot); | |
3255089a | 2034 | |
3b1d9e04 BP |
2035 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size); |
2036 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset); | |
2037 | ||
abb4bb83 BP |
2038 | struct load_index_extensions |
2039 | { | |
abb4bb83 | 2040 | pthread_t pthread; |
abb4bb83 BP |
2041 | struct index_state *istate; |
2042 | const char *mmap; | |
2043 | size_t mmap_size; | |
2044 | unsigned long src_offset; | |
2045 | }; | |
2046 | ||
2047 | static void *load_index_extensions(void *_data) | |
2048 | { | |
2049 | struct load_index_extensions *p = _data; | |
2050 | unsigned long src_offset = p->src_offset; | |
2051 | ||
2052 | while (src_offset <= p->mmap_size - the_hash_algo->rawsz - 8) { | |
2053 | /* After an array of active_nr index entries, | |
2054 | * there can be arbitrary number of extended | |
2055 | * sections, each of which is prefixed with | |
2056 | * extension name (4-byte) and section length | |
2057 | * in 4-byte network byte order. | |
2058 | */ | |
2059 | uint32_t extsize = get_be32(p->mmap + src_offset + 4); | |
2060 | if (read_index_extension(p->istate, | |
2061 | p->mmap + src_offset, | |
2062 | p->mmap + src_offset + 8, | |
2063 | extsize) < 0) { | |
2064 | munmap((void *)p->mmap, p->mmap_size); | |
2065 | die(_("index file corrupt")); | |
2066 | } | |
2067 | src_offset += 8; | |
2068 | src_offset += extsize; | |
2069 | } | |
2070 | ||
2071 | return NULL; | |
2072 | } | |
2073 | ||
77ff1127 BP |
2074 | /* |
2075 | * A helper function that will load the specified range of cache entries | |
2076 | * from the memory mapped file and add them to the given index. | |
2077 | */ | |
2078 | static unsigned long load_cache_entry_block(struct index_state *istate, | |
2079 | struct mem_pool *ce_mem_pool, int offset, int nr, const char *mmap, | |
2080 | unsigned long start_offset, const struct cache_entry *previous_ce) | |
2081 | { | |
2082 | int i; | |
2083 | unsigned long src_offset = start_offset; | |
2084 | ||
2085 | for (i = offset; i < offset + nr; i++) { | |
2086 | struct ondisk_cache_entry *disk_ce; | |
2087 | struct cache_entry *ce; | |
2088 | unsigned long consumed; | |
2089 | ||
2090 | disk_ce = (struct ondisk_cache_entry *)(mmap + src_offset); | |
2091 | ce = create_from_disk(ce_mem_pool, istate->version, disk_ce, &consumed, previous_ce); | |
2092 | set_index_entry(istate, i, ce); | |
2093 | ||
2094 | src_offset += consumed; | |
2095 | previous_ce = ce; | |
2096 | } | |
2097 | return src_offset - start_offset; | |
2098 | } | |
2099 | ||
2100 | static unsigned long load_all_cache_entries(struct index_state *istate, | |
2101 | const char *mmap, size_t mmap_size, unsigned long src_offset) | |
2102 | { | |
2103 | unsigned long consumed; | |
2104 | ||
44c7e1a7 | 2105 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2106 | if (istate->version == 4) { |
44c7e1a7 | 2107 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2108 | estimate_cache_size_from_compressed(istate->cache_nr)); |
2109 | } else { | |
44c7e1a7 | 2110 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2111 | estimate_cache_size(mmap_size, istate->cache_nr)); |
2112 | } | |
2113 | ||
2114 | consumed = load_cache_entry_block(istate, istate->ce_mem_pool, | |
2115 | 0, istate->cache_nr, mmap, src_offset, NULL); | |
2116 | return consumed; | |
2117 | } | |
2118 | ||
3255089a BP |
2119 | /* |
2120 | * Mostly randomly chosen maximum thread counts: we | |
2121 | * cap the parallelism to online_cpus() threads, and we want | |
2122 | * to have at least 10000 cache entries per thread for it to | |
2123 | * be worth starting a thread. | |
2124 | */ | |
2125 | ||
2126 | #define THREAD_COST (10000) | |
2127 | ||
77ff1127 BP |
2128 | struct load_cache_entries_thread_data |
2129 | { | |
2130 | pthread_t pthread; | |
2131 | struct index_state *istate; | |
2132 | struct mem_pool *ce_mem_pool; | |
2133 | int offset; | |
2134 | const char *mmap; | |
2135 | struct index_entry_offset_table *ieot; | |
2136 | int ieot_start; /* starting index into the ieot array */ | |
2137 | int ieot_blocks; /* count of ieot entries to process */ | |
2138 | unsigned long consumed; /* return # of bytes in index file processed */ | |
2139 | }; | |
2140 | ||
2141 | /* | |
2142 | * A thread proc to run the load_cache_entries() computation | |
2143 | * across multiple background threads. | |
2144 | */ | |
2145 | static void *load_cache_entries_thread(void *_data) | |
2146 | { | |
2147 | struct load_cache_entries_thread_data *p = _data; | |
2148 | int i; | |
2149 | ||
2150 | /* iterate across all ieot blocks assigned to this thread */ | |
2151 | for (i = p->ieot_start; i < p->ieot_start + p->ieot_blocks; i++) { | |
2152 | p->consumed += load_cache_entry_block(p->istate, p->ce_mem_pool, | |
2153 | p->offset, p->ieot->entries[i].nr, p->mmap, p->ieot->entries[i].offset, NULL); | |
2154 | p->offset += p->ieot->entries[i].nr; | |
2155 | } | |
2156 | return NULL; | |
2157 | } | |
2158 | ||
2159 | static unsigned long load_cache_entries_threaded(struct index_state *istate, const char *mmap, size_t mmap_size, | |
7bd9631b | 2160 | int nr_threads, struct index_entry_offset_table *ieot) |
77ff1127 BP |
2161 | { |
2162 | int i, offset, ieot_blocks, ieot_start, err; | |
2163 | struct load_cache_entries_thread_data *data; | |
2164 | unsigned long consumed = 0; | |
2165 | ||
2166 | /* a little sanity checking */ | |
2167 | if (istate->name_hash_initialized) | |
2168 | BUG("the name hash isn't thread safe"); | |
2169 | ||
44c7e1a7 EN |
2170 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
2171 | mem_pool_init(istate->ce_mem_pool, 0); | |
77ff1127 BP |
2172 | |
2173 | /* ensure we have no more threads than we have blocks to process */ | |
2174 | if (nr_threads > ieot->nr) | |
2175 | nr_threads = ieot->nr; | |
ca56dadb | 2176 | CALLOC_ARRAY(data, nr_threads); |
77ff1127 BP |
2177 | |
2178 | offset = ieot_start = 0; | |
2179 | ieot_blocks = DIV_ROUND_UP(ieot->nr, nr_threads); | |
2180 | for (i = 0; i < nr_threads; i++) { | |
2181 | struct load_cache_entries_thread_data *p = &data[i]; | |
2182 | int nr, j; | |
2183 | ||
2184 | if (ieot_start + ieot_blocks > ieot->nr) | |
2185 | ieot_blocks = ieot->nr - ieot_start; | |
2186 | ||
2187 | p->istate = istate; | |
2188 | p->offset = offset; | |
2189 | p->mmap = mmap; | |
2190 | p->ieot = ieot; | |
2191 | p->ieot_start = ieot_start; | |
2192 | p->ieot_blocks = ieot_blocks; | |
2193 | ||
2194 | /* create a mem_pool for each thread */ | |
2195 | nr = 0; | |
2196 | for (j = p->ieot_start; j < p->ieot_start + p->ieot_blocks; j++) | |
2197 | nr += p->ieot->entries[j].nr; | |
bcd2c5ee | 2198 | p->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2199 | if (istate->version == 4) { |
44c7e1a7 | 2200 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2201 | estimate_cache_size_from_compressed(nr)); |
2202 | } else { | |
44c7e1a7 | 2203 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2204 | estimate_cache_size(mmap_size, nr)); |
2205 | } | |
2206 | ||
2207 | err = pthread_create(&p->pthread, NULL, load_cache_entries_thread, p); | |
2208 | if (err) | |
2209 | die(_("unable to create load_cache_entries thread: %s"), strerror(err)); | |
2210 | ||
2211 | /* increment by the number of cache entries in the ieot block being processed */ | |
2212 | for (j = 0; j < ieot_blocks; j++) | |
2213 | offset += ieot->entries[ieot_start + j].nr; | |
2214 | ieot_start += ieot_blocks; | |
2215 | } | |
2216 | ||
2217 | for (i = 0; i < nr_threads; i++) { | |
2218 | struct load_cache_entries_thread_data *p = &data[i]; | |
2219 | ||
2220 | err = pthread_join(p->pthread, NULL); | |
2221 | if (err) | |
2222 | die(_("unable to join load_cache_entries thread: %s"), strerror(err)); | |
2223 | mem_pool_combine(istate->ce_mem_pool, p->ce_mem_pool); | |
2224 | consumed += p->consumed; | |
2225 | } | |
2226 | ||
2227 | free(data); | |
2228 | ||
2229 | return consumed; | |
2230 | } | |
77ff1127 | 2231 | |
8fd2cb40 | 2232 | /* remember to discard_cache() before reading a different cache! */ |
3e52f70b | 2233 | int do_read_index(struct index_state *istate, const char *path, int must_exist) |
e83c5163 | 2234 | { |
77ff1127 | 2235 | int fd; |
e83c5163 | 2236 | struct stat st; |
debed2a6 | 2237 | unsigned long src_offset; |
371ed0de BP |
2238 | const struct cache_header *hdr; |
2239 | const char *mmap; | |
7a51ed66 | 2240 | size_t mmap_size; |
abb4bb83 BP |
2241 | struct load_index_extensions p; |
2242 | size_t extension_offset = 0; | |
77ff1127 BP |
2243 | int nr_threads, cpus; |
2244 | struct index_entry_offset_table *ieot = NULL; | |
e83c5163 | 2245 | |
913e0e99 | 2246 | if (istate->initialized) |
4aab5b46 | 2247 | return istate->cache_nr; |
5d1a5c02 | 2248 | |
fba2f38a KB |
2249 | istate->timestamp.sec = 0; |
2250 | istate->timestamp.nsec = 0; | |
8fd2cb40 | 2251 | fd = open(path, O_RDONLY); |
5d1a5c02 | 2252 | if (fd < 0) { |
5fc2fc8f | 2253 | if (!must_exist && errno == ENOENT) |
5d1a5c02 | 2254 | return 0; |
9d0a9e90 | 2255 | die_errno(_("%s: index file open failed"), path); |
5d1a5c02 | 2256 | } |
e83c5163 | 2257 | |
3511a377 | 2258 | if (fstat(fd, &st)) |
9d0a9e90 | 2259 | die_errno(_("%s: cannot stat the open index"), path); |
3511a377 | 2260 | |
7a51ed66 | 2261 | mmap_size = xsize_t(st.st_size); |
aab61359 | 2262 | if (mmap_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
9d0a9e90 | 2263 | die(_("%s: index file smaller than expected"), path); |
3511a377 | 2264 | |
02638d1e | 2265 | mmap = xmmap_gently(NULL, mmap_size, PROT_READ, MAP_PRIVATE, fd, 0); |
7a51ed66 | 2266 | if (mmap == MAP_FAILED) |
dc059294 EW |
2267 | die_errno(_("%s: unable to map index file%s"), path, |
2268 | mmap_os_err()); | |
57d84f8d | 2269 | close(fd); |
e83c5163 | 2270 | |
371ed0de | 2271 | hdr = (const struct cache_header *)mmap; |
7a51ed66 | 2272 | if (verify_hdr(hdr, mmap_size) < 0) |
e83c5163 LT |
2273 | goto unmap; |
2274 | ||
92e2cab9 | 2275 | oidread(&istate->oid, (const unsigned char *)hdr + mmap_size - the_hash_algo->rawsz); |
9d227781 | 2276 | istate->version = ntohl(hdr->hdr_version); |
4aab5b46 JH |
2277 | istate->cache_nr = ntohl(hdr->hdr_entries); |
2278 | istate->cache_alloc = alloc_nr(istate->cache_nr); | |
ca56dadb | 2279 | CALLOC_ARRAY(istate->cache, istate->cache_alloc); |
913e0e99 | 2280 | istate->initialized = 1; |
7a51ed66 | 2281 | |
abb4bb83 BP |
2282 | p.istate = istate; |
2283 | p.mmap = mmap; | |
2284 | p.mmap_size = mmap_size; | |
6c9cd161 | 2285 | |
7a51ed66 | 2286 | src_offset = sizeof(*hdr); |
4aab5b46 | 2287 | |
2a9dedef JN |
2288 | if (git_config_get_index_threads(&nr_threads)) |
2289 | nr_threads = 1; | |
7a51ed66 | 2290 | |
77ff1127 BP |
2291 | /* TODO: does creating more threads than cores help? */ |
2292 | if (!nr_threads) { | |
2293 | nr_threads = istate->cache_nr / THREAD_COST; | |
2294 | cpus = online_cpus(); | |
2295 | if (nr_threads > cpus) | |
2296 | nr_threads = cpus; | |
e83c5163 | 2297 | } |
abb4bb83 | 2298 | |
88168b9b NTND |
2299 | if (!HAVE_THREADS) |
2300 | nr_threads = 1; | |
2301 | ||
abb4bb83 BP |
2302 | if (nr_threads > 1) { |
2303 | extension_offset = read_eoie_extension(mmap, mmap_size); | |
2304 | if (extension_offset) { | |
2305 | int err; | |
2306 | ||
2307 | p.src_offset = extension_offset; | |
2308 | err = pthread_create(&p.pthread, NULL, load_index_extensions, &p); | |
2309 | if (err) | |
2310 | die(_("unable to create load_index_extensions thread: %s"), strerror(err)); | |
2311 | ||
2312 | nr_threads--; | |
2313 | } | |
2314 | } | |
abb4bb83 | 2315 | |
77ff1127 BP |
2316 | /* |
2317 | * Locate and read the index entry offset table so that we can use it | |
2318 | * to multi-thread the reading of the cache entries. | |
2319 | */ | |
2320 | if (extension_offset && nr_threads > 1) | |
2321 | ieot = read_ieot_extension(mmap, mmap_size, extension_offset); | |
2322 | ||
2323 | if (ieot) { | |
7bd9631b | 2324 | src_offset += load_cache_entries_threaded(istate, mmap, mmap_size, nr_threads, ieot); |
77ff1127 | 2325 | free(ieot); |
8e72d675 | 2326 | } else { |
77ff1127 | 2327 | src_offset += load_all_cache_entries(istate, mmap, mmap_size, src_offset); |
8e72d675 | 2328 | } |
6c9cd161 | 2329 | |
fba2f38a | 2330 | istate->timestamp.sec = st.st_mtime; |
c06ff490 | 2331 | istate->timestamp.nsec = ST_MTIME_NSEC(st); |
fba2f38a | 2332 | |
abb4bb83 | 2333 | /* if we created a thread, join it otherwise load the extensions on the primary thread */ |
abb4bb83 BP |
2334 | if (extension_offset) { |
2335 | int ret = pthread_join(p.pthread, NULL); | |
2336 | if (ret) | |
2337 | die(_("unable to join load_index_extensions thread: %s"), strerror(ret)); | |
88168b9b | 2338 | } else { |
abb4bb83 BP |
2339 | p.src_offset = src_offset; |
2340 | load_index_extensions(&p); | |
bad68ec9 | 2341 | } |
371ed0de | 2342 | munmap((void *)mmap, mmap_size); |
42fee7a3 JH |
2343 | |
2344 | /* | |
2345 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2346 | * that is associated with the given "istate". | |
2347 | */ | |
2348 | trace2_data_intmax("index", the_repository, "read/version", | |
2349 | istate->version); | |
2350 | trace2_data_intmax("index", the_repository, "read/cache_nr", | |
2351 | istate->cache_nr); | |
2352 | ||
4300f844 DS |
2353 | if (!istate->repo) |
2354 | istate->repo = the_repository; | |
2355 | prepare_repo_settings(istate->repo); | |
2356 | if (istate->repo->settings.command_requires_full_index) | |
2357 | ensure_full_index(istate); | |
2358 | ||
4aab5b46 | 2359 | return istate->cache_nr; |
e83c5163 LT |
2360 | |
2361 | unmap: | |
371ed0de | 2362 | munmap((void *)mmap, mmap_size); |
9d0a9e90 | 2363 | die(_("index file corrupt")); |
e83c5163 LT |
2364 | } |
2365 | ||
0d59ffb4 CC |
2366 | /* |
2367 | * Signal that the shared index is used by updating its mtime. | |
2368 | * | |
2369 | * This way, shared index can be removed if they have not been used | |
2370 | * for some time. | |
2371 | */ | |
a125a223 | 2372 | static void freshen_shared_index(const char *shared_index, int warn) |
0d59ffb4 | 2373 | { |
0d59ffb4 | 2374 | if (!check_and_freshen_file(shared_index, 1) && warn) |
9d0a9e90 | 2375 | warning(_("could not freshen shared index '%s'"), shared_index); |
0d59ffb4 CC |
2376 | } |
2377 | ||
a125a223 TG |
2378 | int read_index_from(struct index_state *istate, const char *path, |
2379 | const char *gitdir) | |
5fc2fc8f NTND |
2380 | { |
2381 | struct split_index *split_index; | |
2382 | int ret; | |
2182abd9 | 2383 | char *base_oid_hex; |
a125a223 | 2384 | char *base_path; |
5fc2fc8f NTND |
2385 | |
2386 | /* istate->initialized covers both .git/index and .git/sharedindex.xxx */ | |
2387 | if (istate->initialized) | |
2388 | return istate->cache_nr; | |
2389 | ||
42fee7a3 JH |
2390 | /* |
2391 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2392 | * that is associated with the given "istate". | |
2393 | */ | |
2394 | trace2_region_enter_printf("index", "do_read_index", the_repository, | |
2395 | "%s", path); | |
c46c406a | 2396 | trace_performance_enter(); |
5fc2fc8f | 2397 | ret = do_read_index(istate, path, 0); |
c46c406a | 2398 | trace_performance_leave("read cache %s", path); |
42fee7a3 JH |
2399 | trace2_region_leave_printf("index", "do_read_index", the_repository, |
2400 | "%s", path); | |
435ec090 | 2401 | |
5fc2fc8f | 2402 | split_index = istate->split_index; |
2182abd9 | 2403 | if (!split_index || is_null_oid(&split_index->base_oid)) { |
435ec090 | 2404 | post_read_index_from(istate); |
5fc2fc8f | 2405 | return ret; |
03f15a79 | 2406 | } |
5fc2fc8f | 2407 | |
c46c406a | 2408 | trace_performance_enter(); |
5fc2fc8f NTND |
2409 | if (split_index->base) |
2410 | discard_index(split_index->base); | |
2411 | else | |
ca56dadb | 2412 | CALLOC_ARRAY(split_index->base, 1); |
de6ae5f9 | 2413 | |
2182abd9 | 2414 | base_oid_hex = oid_to_hex(&split_index->base_oid); |
2415 | base_path = xstrfmt("%s/sharedindex.%s", gitdir, base_oid_hex); | |
42fee7a3 JH |
2416 | trace2_region_enter_printf("index", "shared/do_read_index", |
2417 | the_repository, "%s", base_path); | |
998330ac | 2418 | ret = do_read_index(split_index->base, base_path, 0); |
42fee7a3 JH |
2419 | trace2_region_leave_printf("index", "shared/do_read_index", |
2420 | the_repository, "%s", base_path); | |
998330ac SG |
2421 | if (!ret) { |
2422 | char *path_copy = xstrdup(path); | |
2423 | const char *base_path2 = xstrfmt("%s/sharedindex.%s", | |
2424 | dirname(path_copy), | |
2425 | base_oid_hex); | |
2426 | free(path_copy); | |
2427 | trace2_region_enter_printf("index", "shared/do_read_index", | |
2428 | the_repository, "%s", base_path2); | |
2429 | ret = do_read_index(split_index->base, base_path2, 1); | |
2430 | trace2_region_leave_printf("index", "shared/do_read_index", | |
2431 | the_repository, "%s", base_path2); | |
2432 | } | |
9001dc2a | 2433 | if (!oideq(&split_index->base_oid, &split_index->base->oid)) |
9d0a9e90 | 2434 | die(_("broken index, expect %s in %s, got %s"), |
2182abd9 | 2435 | base_oid_hex, base_path, |
75691ea3 | 2436 | oid_to_hex(&split_index->base->oid)); |
de6ae5f9 | 2437 | |
a125a223 | 2438 | freshen_shared_index(base_path, 0); |
5fc2fc8f | 2439 | merge_base_index(istate); |
435ec090 | 2440 | post_read_index_from(istate); |
c46c406a | 2441 | trace_performance_leave("read cache %s", base_path); |
b42ad7d5 | 2442 | free(base_path); |
5fc2fc8f NTND |
2443 | return ret; |
2444 | } | |
2445 | ||
fa7b3c2f JH |
2446 | int is_index_unborn(struct index_state *istate) |
2447 | { | |
debed2a6 | 2448 | return (!istate->cache_nr && !istate->timestamp.sec); |
fa7b3c2f JH |
2449 | } |
2450 | ||
4aab5b46 | 2451 | int discard_index(struct index_state *istate) |
6d297f81 | 2452 | { |
8e72d675 JM |
2453 | /* |
2454 | * Cache entries in istate->cache[] should have been allocated | |
2455 | * from the memory pool associated with this index, or from an | |
2456 | * associated split_index. There is no need to free individual | |
8616a2d0 JM |
2457 | * cache entries. validate_cache_entries can detect when this |
2458 | * assertion does not hold. | |
8e72d675 | 2459 | */ |
8616a2d0 | 2460 | validate_cache_entries(istate); |
debed2a6 | 2461 | |
cfc5789a | 2462 | resolve_undo_clear_index(istate); |
4aab5b46 JH |
2463 | istate->cache_nr = 0; |
2464 | istate->cache_changed = 0; | |
fba2f38a KB |
2465 | istate->timestamp.sec = 0; |
2466 | istate->timestamp.nsec = 0; | |
2092678c | 2467 | free_name_hash(istate); |
4aab5b46 | 2468 | cache_tree_free(&(istate->cache_tree)); |
913e0e99 | 2469 | istate->initialized = 0; |
398a3b08 | 2470 | istate->fsmonitor_has_run_once = 0; |
4abc5784 | 2471 | FREE_AND_NULL(istate->fsmonitor_last_update); |
6a83d902 | 2472 | FREE_AND_NULL(istate->cache); |
a0fc4db0 | 2473 | istate->cache_alloc = 0; |
5fc2fc8f | 2474 | discard_split_index(istate); |
f9e6c649 NTND |
2475 | free_untracked_cache(istate->untracked); |
2476 | istate->untracked = NULL; | |
8e72d675 JM |
2477 | |
2478 | if (istate->ce_mem_pool) { | |
8616a2d0 | 2479 | mem_pool_discard(istate->ce_mem_pool, should_validate_cache_entries()); |
44c7e1a7 | 2480 | FREE_AND_NULL(istate->ce_mem_pool); |
8e72d675 JM |
2481 | } |
2482 | ||
7a51ed66 | 2483 | return 0; |
6d297f81 JS |
2484 | } |
2485 | ||
8616a2d0 JM |
2486 | /* |
2487 | * Validate the cache entries of this index. | |
2488 | * All cache entries associated with this index | |
2489 | * should have been allocated by the memory pool | |
2490 | * associated with this index, or by a referenced | |
2491 | * split index. | |
2492 | */ | |
2493 | void validate_cache_entries(const struct index_state *istate) | |
2494 | { | |
2495 | int i; | |
2496 | ||
2497 | if (!should_validate_cache_entries() ||!istate || !istate->initialized) | |
2498 | return; | |
2499 | ||
2500 | for (i = 0; i < istate->cache_nr; i++) { | |
2501 | if (!istate) { | |
391408e5 | 2502 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2503 | } else if (!istate->ce_mem_pool || |
2504 | !mem_pool_contains(istate->ce_mem_pool, istate->cache[i])) { | |
2505 | if (!istate->split_index || | |
2506 | !istate->split_index->base || | |
2507 | !istate->split_index->base->ce_mem_pool || | |
2508 | !mem_pool_contains(istate->split_index->base->ce_mem_pool, istate->cache[i])) { | |
391408e5 | 2509 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2510 | } |
2511 | } | |
2512 | } | |
2513 | ||
2514 | if (istate->split_index) | |
2515 | validate_cache_entries(istate->split_index->base); | |
2516 | } | |
2517 | ||
d1f128b0 | 2518 | int unmerged_index(const struct index_state *istate) |
94a5728c DB |
2519 | { |
2520 | int i; | |
2521 | for (i = 0; i < istate->cache_nr; i++) { | |
2522 | if (ce_stage(istate->cache[i])) | |
2523 | return 1; | |
2524 | } | |
2525 | return 0; | |
2526 | } | |
2527 | ||
150fe065 NTND |
2528 | int repo_index_has_changes(struct repository *repo, |
2529 | struct tree *tree, | |
2530 | struct strbuf *sb) | |
cffbfad5 | 2531 | { |
150fe065 | 2532 | struct index_state *istate = repo->index; |
e1f8694f | 2533 | struct object_id cmp; |
cffbfad5 EN |
2534 | int i; |
2535 | ||
e1f8694f EN |
2536 | if (tree) |
2537 | cmp = tree->object.oid; | |
2538 | if (tree || !get_oid_tree("HEAD", &cmp)) { | |
cffbfad5 EN |
2539 | struct diff_options opt; |
2540 | ||
150fe065 | 2541 | repo_diff_setup(repo, &opt); |
cffbfad5 EN |
2542 | opt.flags.exit_with_status = 1; |
2543 | if (!sb) | |
2544 | opt.flags.quick = 1; | |
8d833e93 | 2545 | diff_setup_done(&opt); |
e1f8694f | 2546 | do_diff_cache(&cmp, &opt); |
cffbfad5 EN |
2547 | diffcore_std(&opt); |
2548 | for (i = 0; sb && i < diff_queued_diff.nr; i++) { | |
2549 | if (i) | |
2550 | strbuf_addch(sb, ' '); | |
2551 | strbuf_addstr(sb, diff_queued_diff.queue[i]->two->path); | |
2552 | } | |
2553 | diff_flush(&opt); | |
2554 | return opt.flags.has_changes != 0; | |
2555 | } else { | |
0c18c059 DS |
2556 | /* TODO: audit for interaction with sparse-index. */ |
2557 | ensure_full_index(istate); | |
1b9fbefb | 2558 | for (i = 0; sb && i < istate->cache_nr; i++) { |
cffbfad5 EN |
2559 | if (i) |
2560 | strbuf_addch(sb, ' '); | |
1b9fbefb | 2561 | strbuf_addstr(sb, istate->cache[i]->name); |
cffbfad5 | 2562 | } |
1b9fbefb | 2563 | return !!istate->cache_nr; |
cffbfad5 EN |
2564 | } |
2565 | } | |
2566 | ||
410334ed DS |
2567 | static int write_index_ext_header(struct hashfile *f, |
2568 | git_hash_ctx *eoie_f, | |
2569 | unsigned int ext, | |
2570 | unsigned int sz) | |
bad68ec9 | 2571 | { |
410334ed DS |
2572 | hashwrite_be32(f, ext); |
2573 | hashwrite_be32(f, sz); | |
2574 | ||
2575 | if (eoie_f) { | |
2576 | ext = htonl(ext); | |
2577 | sz = htonl(sz); | |
2578 | the_hash_algo->update_fn(eoie_f, &ext, sizeof(ext)); | |
2579 | the_hash_algo->update_fn(eoie_f, &sz, sizeof(sz)); | |
3b1d9e04 | 2580 | } |
410334ed | 2581 | return 0; |
bad68ec9 JH |
2582 | } |
2583 | ||
58bf2a4c NTND |
2584 | static void ce_smudge_racily_clean_entry(struct index_state *istate, |
2585 | struct cache_entry *ce) | |
407c8eb0 JH |
2586 | { |
2587 | /* | |
2588 | * The only thing we care about in this function is to smudge the | |
2589 | * falsely clean entry due to touch-update-touch race, so we leave | |
2590 | * everything else as they are. We are called for entries whose | |
c21d39d7 | 2591 | * ce_stat_data.sd_mtime match the index file mtime. |
c70115b4 JH |
2592 | * |
2593 | * Note that this actually does not do much for gitlinks, for | |
2594 | * which ce_match_stat_basic() always goes to the actual | |
2595 | * contents. The caller checks with is_racy_timestamp() which | |
2596 | * always says "no" for gitlinks, so we are not called for them ;-) | |
407c8eb0 JH |
2597 | */ |
2598 | struct stat st; | |
2599 | ||
2600 | if (lstat(ce->name, &st) < 0) | |
2601 | return; | |
2602 | if (ce_match_stat_basic(ce, &st)) | |
2603 | return; | |
58bf2a4c | 2604 | if (ce_modified_check_fs(istate, ce, &st)) { |
4b3511b0 JH |
2605 | /* This is "racily clean"; smudge it. Note that this |
2606 | * is a tricky code. At first glance, it may appear | |
2607 | * that it can break with this sequence: | |
2608 | * | |
2609 | * $ echo xyzzy >frotz | |
2610 | * $ git-update-index --add frotz | |
2611 | * $ : >frotz | |
2612 | * $ sleep 3 | |
2613 | * $ echo filfre >nitfol | |
2614 | * $ git-update-index --add nitfol | |
2615 | * | |
b7e58b17 | 2616 | * but it does not. When the second update-index runs, |
4b3511b0 JH |
2617 | * it notices that the entry "frotz" has the same timestamp |
2618 | * as index, and if we were to smudge it by resetting its | |
2619 | * size to zero here, then the object name recorded | |
2620 | * in index is the 6-byte file but the cached stat information | |
2621 | * becomes zero --- which would then match what we would | |
a6080a0a | 2622 | * obtain from the filesystem next time we stat("frotz"). |
4b3511b0 JH |
2623 | * |
2624 | * However, the second update-index, before calling | |
2625 | * this function, notices that the cached size is 6 | |
2626 | * bytes and what is on the filesystem is an empty | |
2627 | * file, and never calls us, so the cached size information | |
2628 | * for "frotz" stays 6 which does not match the filesystem. | |
2629 | */ | |
c21d39d7 | 2630 | ce->ce_stat_data.sd_size = 0; |
407c8eb0 JH |
2631 | } |
2632 | } | |
2633 | ||
f136f7bf | 2634 | /* Copy miscellaneous fields but not the name */ |
ce012deb | 2635 | static void copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk, |
f136f7bf | 2636 | struct cache_entry *ce) |
7a51ed66 | 2637 | { |
b60e188c | 2638 | short flags; |
575fa8a3 | 2639 | const unsigned hashsz = the_hash_algo->rawsz; |
2640 | uint16_t *flagsp = (uint16_t *)(ondisk->data + hashsz); | |
b60e188c | 2641 | |
c21d39d7 MH |
2642 | ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec); |
2643 | ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec); | |
2644 | ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec); | |
2645 | ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec); | |
2646 | ondisk->dev = htonl(ce->ce_stat_data.sd_dev); | |
2647 | ondisk->ino = htonl(ce->ce_stat_data.sd_ino); | |
7a51ed66 | 2648 | ondisk->mode = htonl(ce->ce_mode); |
c21d39d7 MH |
2649 | ondisk->uid = htonl(ce->ce_stat_data.sd_uid); |
2650 | ondisk->gid = htonl(ce->ce_stat_data.sd_gid); | |
2651 | ondisk->size = htonl(ce->ce_stat_data.sd_size); | |
575fa8a3 | 2652 | hashcpy(ondisk->data, ce->oid.hash); |
b60e188c | 2653 | |
ce51bf09 | 2654 | flags = ce->ce_flags & ~CE_NAMEMASK; |
b60e188c | 2655 | flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce)); |
575fa8a3 | 2656 | flagsp[0] = htons(flags); |
06aaaa0b | 2657 | if (ce->ce_flags & CE_EXTENDED) { |
575fa8a3 | 2658 | flagsp[1] = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16); |
f136f7bf JH |
2659 | } |
2660 | } | |
2661 | ||
410334ed | 2662 | static int ce_write_entry(struct hashfile *f, struct cache_entry *ce, |
ce012deb | 2663 | struct strbuf *previous_name, struct ondisk_cache_entry *ondisk) |
f136f7bf | 2664 | { |
9d227781 | 2665 | int size; |
00a4b035 RJ |
2666 | unsigned int saved_namelen; |
2667 | int stripped_name = 0; | |
ce012deb | 2668 | static unsigned char padding[8] = { 0x00 }; |
f136f7bf | 2669 | |
b3c96fb1 NTND |
2670 | if (ce->ce_flags & CE_STRIP_NAME) { |
2671 | saved_namelen = ce_namelen(ce); | |
2672 | ce->ce_namelen = 0; | |
00a4b035 | 2673 | stripped_name = 1; |
b3c96fb1 NTND |
2674 | } |
2675 | ||
575fa8a3 | 2676 | size = offsetof(struct ondisk_cache_entry,data) + ondisk_data_size(ce->ce_flags, 0); |
ce012deb | 2677 | |
9d227781 | 2678 | if (!previous_name) { |
ce012deb KW |
2679 | int len = ce_namelen(ce); |
2680 | copy_cache_entry_to_ondisk(ondisk, ce); | |
410334ed DS |
2681 | hashwrite(f, ondisk, size); |
2682 | hashwrite(f, ce->name, len); | |
2683 | hashwrite(f, padding, align_padding_size(size, len)); | |
9d227781 JH |
2684 | } else { |
2685 | int common, to_remove, prefix_size; | |
2686 | unsigned char to_remove_vi[16]; | |
2687 | for (common = 0; | |
2688 | (ce->name[common] && | |
2689 | common < previous_name->len && | |
2690 | ce->name[common] == previous_name->buf[common]); | |
2691 | common++) | |
2692 | ; /* still matching */ | |
2693 | to_remove = previous_name->len - common; | |
2694 | prefix_size = encode_varint(to_remove, to_remove_vi); | |
2695 | ||
ce012deb | 2696 | copy_cache_entry_to_ondisk(ondisk, ce); |
410334ed DS |
2697 | hashwrite(f, ondisk, size); |
2698 | hashwrite(f, to_remove_vi, prefix_size); | |
2699 | hashwrite(f, ce->name + common, ce_namelen(ce) - common); | |
2700 | hashwrite(f, padding, 1); | |
9d227781 JH |
2701 | |
2702 | strbuf_splice(previous_name, common, to_remove, | |
2703 | ce->name + common, ce_namelen(ce) - common); | |
06aaaa0b | 2704 | } |
00a4b035 | 2705 | if (stripped_name) { |
b3c96fb1 NTND |
2706 | ce->ce_namelen = saved_namelen; |
2707 | ce->ce_flags &= ~CE_STRIP_NAME; | |
2708 | } | |
7a51ed66 | 2709 | |
410334ed | 2710 | return 0; |
7a51ed66 LT |
2711 | } |
2712 | ||
426ddeea YM |
2713 | /* |
2714 | * This function verifies if index_state has the correct sha1 of the | |
2715 | * index file. Don't die if we have any other failure, just return 0. | |
2716 | */ | |
2717 | static int verify_index_from(const struct index_state *istate, const char *path) | |
2718 | { | |
2719 | int fd; | |
2720 | ssize_t n; | |
2721 | struct stat st; | |
aab61359 | 2722 | unsigned char hash[GIT_MAX_RAWSZ]; |
426ddeea YM |
2723 | |
2724 | if (!istate->initialized) | |
2725 | return 0; | |
2726 | ||
2727 | fd = open(path, O_RDONLY); | |
2728 | if (fd < 0) | |
2729 | return 0; | |
2730 | ||
2731 | if (fstat(fd, &st)) | |
2732 | goto out; | |
2733 | ||
aab61359 | 2734 | if (st.st_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
426ddeea YM |
2735 | goto out; |
2736 | ||
aab61359 | 2737 | n = pread_in_full(fd, hash, the_hash_algo->rawsz, st.st_size - the_hash_algo->rawsz); |
2738 | if (n != the_hash_algo->rawsz) | |
426ddeea YM |
2739 | goto out; |
2740 | ||
67947c34 | 2741 | if (!hasheq(istate->oid.hash, hash)) |
426ddeea YM |
2742 | goto out; |
2743 | ||
2744 | close(fd); | |
2745 | return 1; | |
2746 | ||
2747 | out: | |
2748 | close(fd); | |
2749 | return 0; | |
2750 | } | |
2751 | ||
1b0d968b | 2752 | static int repo_verify_index(struct repository *repo) |
426ddeea | 2753 | { |
1b0d968b | 2754 | return verify_index_from(repo->index, repo->index_file); |
426ddeea YM |
2755 | } |
2756 | ||
483fbe2b JH |
2757 | static int has_racy_timestamp(struct index_state *istate) |
2758 | { | |
2759 | int entries = istate->cache_nr; | |
2760 | int i; | |
2761 | ||
2762 | for (i = 0; i < entries; i++) { | |
2763 | struct cache_entry *ce = istate->cache[i]; | |
2764 | if (is_racy_timestamp(istate, ce)) | |
2765 | return 1; | |
2766 | } | |
2767 | return 0; | |
2768 | } | |
2769 | ||
1b0d968b NTND |
2770 | void repo_update_index_if_able(struct repository *repo, |
2771 | struct lock_file *lockfile) | |
ccdc4ec3 | 2772 | { |
1b0d968b NTND |
2773 | if ((repo->index->cache_changed || |
2774 | has_racy_timestamp(repo->index)) && | |
2775 | repo_verify_index(repo)) | |
2776 | write_locked_index(repo->index, lockfile, COMMIT_LOCK); | |
b74c90fb MÅ |
2777 | else |
2778 | rollback_lock_file(lockfile); | |
ccdc4ec3 JH |
2779 | } |
2780 | ||
d8465500 JN |
2781 | static int record_eoie(void) |
2782 | { | |
2783 | int val; | |
2784 | ||
2785 | if (!git_config_get_bool("index.recordendofindexentries", &val)) | |
2786 | return val; | |
2a9dedef JN |
2787 | |
2788 | /* | |
2789 | * As a convenience, the end of index entries extension | |
2790 | * used for threading is written by default if the user | |
2791 | * explicitly requested threaded index reads. | |
2792 | */ | |
2793 | return !git_config_get_index_threads(&val) && val != 1; | |
d8465500 JN |
2794 | } |
2795 | ||
42916054 JN |
2796 | static int record_ieot(void) |
2797 | { | |
2798 | int val; | |
2799 | ||
2800 | if (!git_config_get_bool("index.recordoffsettable", &val)) | |
2801 | return val; | |
2a9dedef JN |
2802 | |
2803 | /* | |
2804 | * As a convenience, the offset table used for threading is | |
2805 | * written by default if the user explicitly requested | |
2806 | * threaded index reads. | |
2807 | */ | |
2808 | return !git_config_get_index_threads(&val) && val != 1; | |
42916054 JN |
2809 | } |
2810 | ||
812d6b00 MÅ |
2811 | /* |
2812 | * On success, `tempfile` is closed. If it is the temporary file | |
2813 | * of a `struct lock_file`, we will therefore effectively perform | |
2814 | * a 'close_lock_file_gently()`. Since that is an implementation | |
2815 | * detail of lockfiles, callers of `do_write_index()` should not | |
2816 | * rely on it. | |
2817 | */ | |
9f41c7a6 | 2818 | static int do_write_index(struct index_state *istate, struct tempfile *tempfile, |
c18b80a0 | 2819 | int strip_extensions) |
197ee8c9 | 2820 | { |
ca54d9ba | 2821 | uint64_t start = getnanotime(); |
410334ed DS |
2822 | struct hashfile *f; |
2823 | git_hash_ctx *eoie_c = NULL; | |
197ee8c9 | 2824 | struct cache_header hdr; |
b50386c7 | 2825 | int i, err = 0, removed, extended, hdr_version; |
4aab5b46 JH |
2826 | struct cache_entry **cache = istate->cache; |
2827 | int entries = istate->cache_nr; | |
e1afca4f | 2828 | struct stat st; |
575fa8a3 | 2829 | struct ondisk_cache_entry ondisk; |
9d227781 | 2830 | struct strbuf previous_name_buf = STRBUF_INIT, *previous_name; |
4bddd983 | 2831 | int drop_cache_tree = istate->drop_cache_tree; |
3b1d9e04 | 2832 | off_t offset; |
77ff1127 | 2833 | int ieot_entries = 1; |
3255089a BP |
2834 | struct index_entry_offset_table *ieot = NULL; |
2835 | int nr, nr_threads; | |
025a0709 | 2836 | |
410334ed DS |
2837 | f = hashfd(tempfile->fd, tempfile->filename.buf); |
2838 | ||
06aaaa0b | 2839 | for (i = removed = extended = 0; i < entries; i++) { |
7a51ed66 | 2840 | if (cache[i]->ce_flags & CE_REMOVE) |
025a0709 | 2841 | removed++; |
197ee8c9 | 2842 | |
06aaaa0b NTND |
2843 | /* reduce extended entries if possible */ |
2844 | cache[i]->ce_flags &= ~CE_EXTENDED; | |
2845 | if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) { | |
2846 | extended++; | |
2847 | cache[i]->ce_flags |= CE_EXTENDED; | |
2848 | } | |
2849 | } | |
2850 | ||
e8ffd034 | 2851 | if (!istate->version) |
7211b9e7 | 2852 | istate->version = get_index_format_default(the_repository); |
9d227781 JH |
2853 | |
2854 | /* demote version 3 to version 2 when the latter suffices */ | |
2855 | if (istate->version == 3 || istate->version == 2) | |
2856 | istate->version = extended ? 3 : 2; | |
2857 | ||
2858 | hdr_version = istate->version; | |
2859 | ||
ccc4feb5 | 2860 | hdr.hdr_signature = htonl(CACHE_SIGNATURE); |
9d227781 | 2861 | hdr.hdr_version = htonl(hdr_version); |
025a0709 | 2862 | hdr.hdr_entries = htonl(entries - removed); |
197ee8c9 | 2863 | |
410334ed | 2864 | hashwrite(f, &hdr, sizeof(hdr)); |
197ee8c9 | 2865 | |
2a9dedef | 2866 | if (!HAVE_THREADS || git_config_get_index_threads(&nr_threads)) |
88168b9b | 2867 | nr_threads = 1; |
62e5ee81 | 2868 | |
42916054 | 2869 | if (nr_threads != 1 && record_ieot()) { |
3255089a BP |
2870 | int ieot_blocks, cpus; |
2871 | ||
2872 | /* | |
2873 | * ensure default number of ieot blocks maps evenly to the | |
2874 | * default number of threads that will process them leaving | |
2875 | * room for the thread to load the index extensions. | |
2876 | */ | |
2877 | if (!nr_threads) { | |
2878 | ieot_blocks = istate->cache_nr / THREAD_COST; | |
2879 | cpus = online_cpus(); | |
2880 | if (ieot_blocks > cpus - 1) | |
2881 | ieot_blocks = cpus - 1; | |
2882 | } else { | |
2883 | ieot_blocks = nr_threads; | |
77ff1127 BP |
2884 | if (ieot_blocks > istate->cache_nr) |
2885 | ieot_blocks = istate->cache_nr; | |
3255089a BP |
2886 | } |
2887 | ||
2888 | /* | |
2889 | * no reason to write out the IEOT extension if we don't | |
2890 | * have enough blocks to utilize multi-threading | |
2891 | */ | |
2892 | if (ieot_blocks > 1) { | |
2893 | ieot = xcalloc(1, sizeof(struct index_entry_offset_table) | |
2894 | + (ieot_blocks * sizeof(struct index_entry_offset))); | |
77ff1127 | 2895 | ieot_entries = DIV_ROUND_UP(entries, ieot_blocks); |
3255089a BP |
2896 | } |
2897 | } | |
3255089a | 2898 | |
410334ed DS |
2899 | offset = hashfile_total(f); |
2900 | ||
3255089a | 2901 | nr = 0; |
9d227781 | 2902 | previous_name = (hdr_version == 4) ? &previous_name_buf : NULL; |
ce012deb | 2903 | |
197ee8c9 LT |
2904 | for (i = 0; i < entries; i++) { |
2905 | struct cache_entry *ce = cache[i]; | |
7a51ed66 | 2906 | if (ce->ce_flags & CE_REMOVE) |
aa16021e | 2907 | continue; |
e06c43c7 | 2908 | if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce)) |
58bf2a4c | 2909 | ce_smudge_racily_clean_entry(istate, ce); |
99d1a986 | 2910 | if (is_null_oid(&ce->oid)) { |
83bd7437 JK |
2911 | static const char msg[] = "cache entry has null sha1: %s"; |
2912 | static int allow = -1; | |
2913 | ||
2914 | if (allow < 0) | |
2915 | allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0); | |
2916 | if (allow) | |
2917 | warning(msg, ce->name); | |
2918 | else | |
b50386c7 | 2919 | err = error(msg, ce->name); |
a96d3cc3 JK |
2920 | |
2921 | drop_cache_tree = 1; | |
83bd7437 | 2922 | } |
77ff1127 | 2923 | if (ieot && i && (i % ieot_entries == 0)) { |
3255089a BP |
2924 | ieot->entries[ieot->nr].nr = nr; |
2925 | ieot->entries[ieot->nr].offset = offset; | |
2926 | ieot->nr++; | |
2927 | /* | |
2928 | * If we have a V4 index, set the first byte to an invalid | |
2929 | * character to ensure there is nothing common with the previous | |
2930 | * entry | |
2931 | */ | |
2932 | if (previous_name) | |
2933 | previous_name->buf[0] = 0; | |
2934 | nr = 0; | |
410334ed DS |
2935 | |
2936 | offset = hashfile_total(f); | |
3255089a | 2937 | } |
410334ed | 2938 | if (ce_write_entry(f, ce, previous_name, (struct ondisk_cache_entry *)&ondisk) < 0) |
b50386c7 KW |
2939 | err = -1; |
2940 | ||
2941 | if (err) | |
2942 | break; | |
3255089a BP |
2943 | nr++; |
2944 | } | |
2945 | if (ieot && nr) { | |
2946 | ieot->entries[ieot->nr].nr = nr; | |
2947 | ieot->entries[ieot->nr].offset = offset; | |
2948 | ieot->nr++; | |
197ee8c9 | 2949 | } |
9d227781 | 2950 | strbuf_release(&previous_name_buf); |
1af1c2b6 | 2951 | |
3255089a BP |
2952 | if (err) { |
2953 | free(ieot); | |
b50386c7 | 2954 | return err; |
3255089a | 2955 | } |
b50386c7 | 2956 | |
410334ed DS |
2957 | offset = hashfile_total(f); |
2958 | ||
2959 | /* | |
2960 | * The extension headers must be hashed on their own for the | |
2961 | * EOIE extension. Create a hashfile here to compute that hash. | |
2962 | */ | |
2963 | if (offset && record_eoie()) { | |
2964 | CALLOC_ARRAY(eoie_c, 1); | |
2965 | the_hash_algo->init_fn(eoie_c); | |
3255089a | 2966 | } |
3b1d9e04 | 2967 | |
3255089a BP |
2968 | /* |
2969 | * Lets write out CACHE_EXT_INDEXENTRYOFFSETTABLE first so that we | |
2970 | * can minimize the number of extensions we have to scan through to | |
2971 | * find it during load. Write it out regardless of the | |
2972 | * strip_extensions parameter as we need it when loading the shared | |
2973 | * index. | |
2974 | */ | |
3255089a BP |
2975 | if (ieot) { |
2976 | struct strbuf sb = STRBUF_INIT; | |
2977 | ||
2978 | write_ieot_extension(&sb, ieot); | |
410334ed DS |
2979 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_INDEXENTRYOFFSETTABLE, sb.len) < 0; |
2980 | hashwrite(f, sb.buf, sb.len); | |
3255089a BP |
2981 | strbuf_release(&sb); |
2982 | free(ieot); | |
2983 | if (err) | |
2984 | return -1; | |
2985 | } | |
3255089a | 2986 | |
6e37c8ed NTND |
2987 | if (!strip_extensions && istate->split_index && |
2988 | !is_null_oid(&istate->split_index->base_oid)) { | |
5fc2fc8f NTND |
2989 | struct strbuf sb = STRBUF_INIT; |
2990 | ||
2991 | err = write_link_extension(&sb, istate) < 0 || | |
410334ed DS |
2992 | write_index_ext_header(f, eoie_c, CACHE_EXT_LINK, |
2993 | sb.len) < 0; | |
2994 | hashwrite(f, sb.buf, sb.len); | |
5fc2fc8f NTND |
2995 | strbuf_release(&sb); |
2996 | if (err) | |
2997 | return -1; | |
2998 | } | |
a96d3cc3 | 2999 | if (!strip_extensions && !drop_cache_tree && istate->cache_tree) { |
f285a2d7 | 3000 | struct strbuf sb = STRBUF_INIT; |
1dffb8fa | 3001 | |
1dffb8fa | 3002 | cache_tree_write(&sb, istate->cache_tree); |
410334ed DS |
3003 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_TREE, sb.len) < 0; |
3004 | hashwrite(f, sb.buf, sb.len); | |
1dffb8fa PH |
3005 | strbuf_release(&sb); |
3006 | if (err) | |
bad68ec9 | 3007 | return -1; |
bad68ec9 | 3008 | } |
c18b80a0 | 3009 | if (!strip_extensions && istate->resolve_undo) { |
cfc5789a JH |
3010 | struct strbuf sb = STRBUF_INIT; |
3011 | ||
3012 | resolve_undo_write(&sb, istate->resolve_undo); | |
410334ed DS |
3013 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_RESOLVE_UNDO, |
3014 | sb.len) < 0; | |
3015 | hashwrite(f, sb.buf, sb.len); | |
cfc5789a JH |
3016 | strbuf_release(&sb); |
3017 | if (err) | |
3018 | return -1; | |
3019 | } | |
83c094ad NTND |
3020 | if (!strip_extensions && istate->untracked) { |
3021 | struct strbuf sb = STRBUF_INIT; | |
3022 | ||
3023 | write_untracked_extension(&sb, istate->untracked); | |
410334ed DS |
3024 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_UNTRACKED, |
3025 | sb.len) < 0; | |
3026 | hashwrite(f, sb.buf, sb.len); | |
83c094ad NTND |
3027 | strbuf_release(&sb); |
3028 | if (err) | |
3029 | return -1; | |
3030 | } | |
883e248b BP |
3031 | if (!strip_extensions && istate->fsmonitor_last_update) { |
3032 | struct strbuf sb = STRBUF_INIT; | |
3033 | ||
3034 | write_fsmonitor_extension(&sb, istate); | |
410334ed DS |
3035 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_FSMONITOR, sb.len) < 0; |
3036 | hashwrite(f, sb.buf, sb.len); | |
3b1d9e04 BP |
3037 | strbuf_release(&sb); |
3038 | if (err) | |
3039 | return -1; | |
3040 | } | |
cd42415f | 3041 | if (istate->sparse_index) { |
410334ed | 3042 | if (write_index_ext_header(f, eoie_c, CACHE_EXT_SPARSE_DIRECTORIES, 0) < 0) |
cd42415f DS |
3043 | return -1; |
3044 | } | |
3b1d9e04 BP |
3045 | |
3046 | /* | |
3047 | * CACHE_EXT_ENDOFINDEXENTRIES must be written as the last entry before the SHA1 | |
3048 | * so that it can be found and processed before all the index entries are | |
3049 | * read. Write it out regardless of the strip_extensions parameter as we need it | |
3050 | * when loading the shared index. | |
3051 | */ | |
410334ed | 3052 | if (eoie_c) { |
3b1d9e04 BP |
3053 | struct strbuf sb = STRBUF_INIT; |
3054 | ||
410334ed DS |
3055 | write_eoie_extension(&sb, eoie_c, offset); |
3056 | err = write_index_ext_header(f, NULL, CACHE_EXT_ENDOFINDEXENTRIES, sb.len) < 0; | |
3057 | hashwrite(f, sb.buf, sb.len); | |
883e248b BP |
3058 | strbuf_release(&sb); |
3059 | if (err) | |
3060 | return -1; | |
3061 | } | |
e1afca4f | 3062 | |
410334ed | 3063 | finalize_hashfile(f, istate->oid.hash, CSUM_HASH_IN_STREAM); |
49bd0fc2 | 3064 | if (close_tempfile_gently(tempfile)) { |
6a8c89d0 | 3065 | error(_("could not close '%s'"), get_tempfile_path(tempfile)); |
49bd0fc2 JK |
3066 | return -1; |
3067 | } | |
6a8c89d0 | 3068 | if (stat(get_tempfile_path(tempfile), &st)) |
e1afca4f | 3069 | return -1; |
5bcf109c KB |
3070 | istate->timestamp.sec = (unsigned int)st.st_mtime; |
3071 | istate->timestamp.nsec = ST_MTIME_NSEC(st); | |
ca54d9ba | 3072 | trace_performance_since(start, "write index, changed mask = %x", istate->cache_changed); |
42fee7a3 JH |
3073 | |
3074 | /* | |
3075 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3076 | * that is associated with the given "istate". | |
3077 | */ | |
3078 | trace2_data_intmax("index", the_repository, "write/version", | |
3079 | istate->version); | |
3080 | trace2_data_intmax("index", the_repository, "write/cache_nr", | |
3081 | istate->cache_nr); | |
3082 | ||
e1afca4f | 3083 | return 0; |
197ee8c9 | 3084 | } |
e46bbcf6 | 3085 | |
626f35c8 NTND |
3086 | void set_alternate_index_output(const char *name) |
3087 | { | |
3088 | alternate_index_output = name; | |
3089 | } | |
3090 | ||
3091 | static int commit_locked_index(struct lock_file *lk) | |
3092 | { | |
751baced MH |
3093 | if (alternate_index_output) |
3094 | return commit_lock_file_to(lk, alternate_index_output); | |
3095 | else | |
626f35c8 | 3096 | return commit_lock_file(lk); |
626f35c8 NTND |
3097 | } |
3098 | ||
03b86647 NTND |
3099 | static int do_write_locked_index(struct index_state *istate, struct lock_file *lock, |
3100 | unsigned flags) | |
3101 | { | |
42fee7a3 | 3102 | int ret; |
6e773527 DS |
3103 | int was_full = !istate->sparse_index; |
3104 | ||
ce7a9f01 | 3105 | ret = convert_to_sparse(istate, 0); |
6e773527 DS |
3106 | |
3107 | if (ret) { | |
3108 | warning(_("failed to convert to a sparse-index")); | |
3109 | return ret; | |
3110 | } | |
42fee7a3 JH |
3111 | |
3112 | /* | |
3113 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3114 | * that is associated with the given "istate". | |
3115 | */ | |
3116 | trace2_region_enter_printf("index", "do_write_index", the_repository, | |
6a8c89d0 | 3117 | "%s", get_lock_file_path(lock)); |
42fee7a3 JH |
3118 | ret = do_write_index(istate, lock->tempfile, 0); |
3119 | trace2_region_leave_printf("index", "do_write_index", the_repository, | |
6a8c89d0 | 3120 | "%s", get_lock_file_path(lock)); |
42fee7a3 | 3121 | |
6e773527 DS |
3122 | if (was_full) |
3123 | ensure_full_index(istate); | |
3124 | ||
03b86647 NTND |
3125 | if (ret) |
3126 | return ret; | |
03b86647 | 3127 | if (flags & COMMIT_LOCK) |
1956ecd0 BP |
3128 | ret = commit_locked_index(lock); |
3129 | else | |
3130 | ret = close_lock_file_gently(lock); | |
3131 | ||
3132 | run_hook_le(NULL, "post-index-change", | |
3133 | istate->updated_workdir ? "1" : "0", | |
3134 | istate->updated_skipworktree ? "1" : "0", NULL); | |
3135 | istate->updated_workdir = 0; | |
3136 | istate->updated_skipworktree = 0; | |
3137 | ||
3138 | return ret; | |
03b86647 NTND |
3139 | } |
3140 | ||
5fc2fc8f NTND |
3141 | static int write_split_index(struct index_state *istate, |
3142 | struct lock_file *lock, | |
3143 | unsigned flags) | |
3144 | { | |
3145 | int ret; | |
3146 | prepare_to_write_split_index(istate); | |
3147 | ret = do_write_locked_index(istate, lock, flags); | |
3148 | finish_writing_split_index(istate); | |
3149 | return ret; | |
3150 | } | |
3151 | ||
b9683722 CC |
3152 | static const char *shared_index_expire = "2.weeks.ago"; |
3153 | ||
3154 | static unsigned long get_shared_index_expire_date(void) | |
3155 | { | |
3156 | static unsigned long shared_index_expire_date; | |
3157 | static int shared_index_expire_date_prepared; | |
3158 | ||
3159 | if (!shared_index_expire_date_prepared) { | |
3160 | git_config_get_expiry("splitindex.sharedindexexpire", | |
3161 | &shared_index_expire); | |
3162 | shared_index_expire_date = approxidate(shared_index_expire); | |
3163 | shared_index_expire_date_prepared = 1; | |
3164 | } | |
3165 | ||
3166 | return shared_index_expire_date; | |
3167 | } | |
3168 | ||
3169 | static int should_delete_shared_index(const char *shared_index_path) | |
3170 | { | |
3171 | struct stat st; | |
3172 | unsigned long expiration; | |
3173 | ||
3174 | /* Check timestamp */ | |
3175 | expiration = get_shared_index_expire_date(); | |
3176 | if (!expiration) | |
3177 | return 0; | |
3178 | if (stat(shared_index_path, &st)) | |
78bde923 | 3179 | return error_errno(_("could not stat '%s'"), shared_index_path); |
b9683722 CC |
3180 | if (st.st_mtime > expiration) |
3181 | return 0; | |
3182 | ||
3183 | return 1; | |
3184 | } | |
3185 | ||
3186 | static int clean_shared_index_files(const char *current_hex) | |
3187 | { | |
3188 | struct dirent *de; | |
3189 | DIR *dir = opendir(get_git_dir()); | |
3190 | ||
3191 | if (!dir) | |
3192 | return error_errno(_("unable to open git dir: %s"), get_git_dir()); | |
3193 | ||
3194 | while ((de = readdir(dir)) != NULL) { | |
3195 | const char *sha1_hex; | |
3196 | const char *shared_index_path; | |
3197 | if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex)) | |
3198 | continue; | |
3199 | if (!strcmp(sha1_hex, current_hex)) | |
3200 | continue; | |
3201 | shared_index_path = git_path("%s", de->d_name); | |
3202 | if (should_delete_shared_index(shared_index_path) > 0 && | |
3203 | unlink(shared_index_path)) | |
3204 | warning_errno(_("unable to unlink: %s"), shared_index_path); | |
3205 | } | |
3206 | closedir(dir); | |
3207 | ||
3208 | return 0; | |
3209 | } | |
3210 | ||
a0a96756 | 3211 | static int write_shared_index(struct index_state *istate, |
59f9d2dd | 3212 | struct tempfile **temp) |
c18b80a0 NTND |
3213 | { |
3214 | struct split_index *si = istate->split_index; | |
6e773527 | 3215 | int ret, was_full = !istate->sparse_index; |
c18b80a0 | 3216 | |
c18b80a0 | 3217 | move_cache_to_base_index(istate); |
ce7a9f01 | 3218 | convert_to_sparse(istate, 0); |
42fee7a3 JH |
3219 | |
3220 | trace2_region_enter_printf("index", "shared/do_write_index", | |
6a8c89d0 | 3221 | the_repository, "%s", get_tempfile_path(*temp)); |
7db2d08c | 3222 | ret = do_write_index(si->base, *temp, 1); |
c173542c | 3223 | trace2_region_leave_printf("index", "shared/do_write_index", |
6a8c89d0 | 3224 | the_repository, "%s", get_tempfile_path(*temp)); |
42fee7a3 | 3225 | |
6e773527 DS |
3226 | if (was_full) |
3227 | ensure_full_index(istate); | |
3228 | ||
59f9d2dd | 3229 | if (ret) |
c18b80a0 | 3230 | return ret; |
7db2d08c | 3231 | ret = adjust_shared_perm(get_tempfile_path(*temp)); |
df801f3f | 3232 | if (ret) { |
9d0a9e90 | 3233 | error(_("cannot fix permission bits on '%s'"), get_tempfile_path(*temp)); |
df801f3f CC |
3234 | return ret; |
3235 | } | |
7db2d08c | 3236 | ret = rename_tempfile(temp, |
75691ea3 | 3237 | git_path("sharedindex.%s", oid_to_hex(&si->base->oid))); |
b9683722 | 3238 | if (!ret) { |
75691ea3 | 3239 | oidcpy(&si->base_oid, &si->base->oid); |
3240 | clean_shared_index_files(oid_to_hex(&si->base->oid)); | |
b9683722 CC |
3241 | } |
3242 | ||
c18b80a0 NTND |
3243 | return ret; |
3244 | } | |
3245 | ||
e6a1dd77 CC |
3246 | static const int default_max_percent_split_change = 20; |
3247 | ||
3248 | static int too_many_not_shared_entries(struct index_state *istate) | |
3249 | { | |
3250 | int i, not_shared = 0; | |
3251 | int max_split = git_config_get_max_percent_split_change(); | |
3252 | ||
3253 | switch (max_split) { | |
3254 | case -1: | |
3255 | /* not or badly configured: use the default value */ | |
3256 | max_split = default_max_percent_split_change; | |
3257 | break; | |
3258 | case 0: | |
3259 | return 1; /* 0% means always write a new shared index */ | |
3260 | case 100: | |
3261 | return 0; /* 100% means never write a new shared index */ | |
3262 | default: | |
3263 | break; /* just use the configured value */ | |
3264 | } | |
3265 | ||
3266 | /* Count not shared entries */ | |
3267 | for (i = 0; i < istate->cache_nr; i++) { | |
3268 | struct cache_entry *ce = istate->cache[i]; | |
3269 | if (!ce->index) | |
3270 | not_shared++; | |
3271 | } | |
3272 | ||
3273 | return (int64_t)istate->cache_nr * max_split < (int64_t)not_shared * 100; | |
3274 | } | |
3275 | ||
03b86647 NTND |
3276 | int write_locked_index(struct index_state *istate, struct lock_file *lock, |
3277 | unsigned flags) | |
3278 | { | |
e8ffd034 | 3279 | int new_shared_index, ret, test_split_index_env; |
5fc2fc8f NTND |
3280 | struct split_index *si = istate->split_index; |
3281 | ||
4592e608 | 3282 | if (git_env_bool("GIT_TEST_CHECK_CACHE_TREE", 0)) |
c207e9e1 | 3283 | cache_tree_verify(the_repository, istate); |
4592e608 | 3284 | |
61000814 MÅ |
3285 | if ((flags & SKIP_IF_UNCHANGED) && !istate->cache_changed) { |
3286 | if (flags & COMMIT_LOCK) | |
3287 | rollback_lock_file(lock); | |
3288 | return 0; | |
3289 | } | |
3290 | ||
3bd28eb2 AV |
3291 | if (istate->fsmonitor_last_update) |
3292 | fill_fsmonitor_bitmap(istate); | |
3293 | ||
e8ffd034 SG |
3294 | test_split_index_env = git_env_bool("GIT_TEST_SPLIT_INDEX", 0); |
3295 | ||
3296 | if ((!si && !test_split_index_env) || | |
3297 | alternate_index_output || | |
5165dd59 | 3298 | (istate->cache_changed & ~EXTMASK)) { |
5fc2fc8f | 3299 | if (si) |
2182abd9 | 3300 | oidclr(&si->base_oid); |
df60cf57 MÅ |
3301 | ret = do_write_locked_index(istate, lock, flags); |
3302 | goto out; | |
5fc2fc8f NTND |
3303 | } |
3304 | ||
e8ffd034 SG |
3305 | if (test_split_index_env) { |
3306 | if (!si) { | |
3307 | si = init_split_index(istate); | |
d6e3c181 | 3308 | istate->cache_changed |= SPLIT_INDEX_ORDERED; |
e8ffd034 SG |
3309 | } else { |
3310 | int v = si->base_oid.hash[0]; | |
3311 | if ((v & 15) < 6) | |
3312 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
3313 | } | |
d6e3c181 | 3314 | } |
e6a1dd77 CC |
3315 | if (too_many_not_shared_entries(istate)) |
3316 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
0d59ffb4 CC |
3317 | |
3318 | new_shared_index = istate->cache_changed & SPLIT_INDEX_ORDERED; | |
3319 | ||
3320 | if (new_shared_index) { | |
59f9d2dd NTND |
3321 | struct tempfile *temp; |
3322 | int saved_errno; | |
3323 | ||
c9d6c788 ÆAB |
3324 | /* Same initial permissions as the main .git/index file */ |
3325 | temp = mks_tempfile_sm(git_path("sharedindex_XXXXXX"), 0, 0666); | |
59f9d2dd | 3326 | if (!temp) { |
2182abd9 | 3327 | oidclr(&si->base_oid); |
59f9d2dd | 3328 | ret = do_write_locked_index(istate, lock, flags); |
ef5b3a6c NTND |
3329 | goto out; |
3330 | } | |
3331 | ret = write_shared_index(istate, &temp); | |
59f9d2dd NTND |
3332 | |
3333 | saved_errno = errno; | |
3334 | if (is_tempfile_active(temp)) | |
3335 | delete_tempfile(&temp); | |
3336 | errno = saved_errno; | |
3337 | ||
c18b80a0 | 3338 | if (ret) |
df60cf57 | 3339 | goto out; |
c18b80a0 NTND |
3340 | } |
3341 | ||
0d59ffb4 CC |
3342 | ret = write_split_index(istate, lock, flags); |
3343 | ||
3344 | /* Freshen the shared index only if the split-index was written */ | |
6e37c8ed | 3345 | if (!ret && !new_shared_index && !is_null_oid(&si->base_oid)) { |
a125a223 | 3346 | const char *shared_index = git_path("sharedindex.%s", |
2182abd9 | 3347 | oid_to_hex(&si->base_oid)); |
a125a223 TG |
3348 | freshen_shared_index(shared_index, 1); |
3349 | } | |
0d59ffb4 | 3350 | |
df60cf57 MÅ |
3351 | out: |
3352 | if (flags & COMMIT_LOCK) | |
3353 | rollback_lock_file(lock); | |
0d59ffb4 | 3354 | return ret; |
03b86647 NTND |
3355 | } |
3356 | ||
e46bbcf6 MV |
3357 | /* |
3358 | * Read the index file that is potentially unmerged into given | |
ad376204 EN |
3359 | * index_state, dropping any unmerged entries to stage #0 (potentially |
3360 | * resulting in a path appearing as both a file and a directory in the | |
3361 | * index; the caller is responsible to clear out the extra entries | |
3362 | * before writing the index to a tree). Returns true if the index is | |
3363 | * unmerged. Callers who want to refuse to work from an unmerged | |
3364 | * state can call this and check its return value, instead of calling | |
3365 | * read_cache(). | |
e46bbcf6 | 3366 | */ |
e1ff0a32 | 3367 | int repo_read_index_unmerged(struct repository *repo) |
e46bbcf6 | 3368 | { |
e1ff0a32 | 3369 | struct index_state *istate; |
e46bbcf6 | 3370 | int i; |
d1a43f2a | 3371 | int unmerged = 0; |
e46bbcf6 | 3372 | |
e1ff0a32 NTND |
3373 | repo_read_index(repo); |
3374 | istate = repo->index; | |
e46bbcf6 MV |
3375 | for (i = 0; i < istate->cache_nr; i++) { |
3376 | struct cache_entry *ce = istate->cache[i]; | |
d1a43f2a | 3377 | struct cache_entry *new_ce; |
a849735b | 3378 | int len; |
d1a43f2a JH |
3379 | |
3380 | if (!ce_stage(ce)) | |
e46bbcf6 | 3381 | continue; |
d1a43f2a | 3382 | unmerged = 1; |
68c4f6a5 | 3383 | len = ce_namelen(ce); |
a849735b | 3384 | new_ce = make_empty_cache_entry(istate, len); |
d1a43f2a | 3385 | memcpy(new_ce->name, ce->name, len); |
b60e188c TG |
3386 | new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED; |
3387 | new_ce->ce_namelen = len; | |
d1a43f2a | 3388 | new_ce->ce_mode = ce->ce_mode; |
ad376204 | 3389 | if (add_index_entry(istate, new_ce, ADD_CACHE_SKIP_DFCHECK)) |
9d0a9e90 | 3390 | return error(_("%s: cannot drop to stage #0"), |
5699d17e | 3391 | new_ce->name); |
e46bbcf6 | 3392 | } |
d1a43f2a | 3393 | return unmerged; |
e46bbcf6 | 3394 | } |
041aee31 | 3395 | |
98fa4738 JK |
3396 | /* |
3397 | * Returns 1 if the path is an "other" path with respect to | |
3398 | * the index; that is, the path is not mentioned in the index at all, | |
3399 | * either as a file, a directory with some files in the index, | |
3400 | * or as an unmerged entry. | |
3401 | * | |
3402 | * We helpfully remove a trailing "/" from directories so that | |
3403 | * the output of read_directory can be used as-is. | |
3404 | */ | |
847a9e5d DS |
3405 | int index_name_is_other(struct index_state *istate, const char *name, |
3406 | int namelen) | |
98fa4738 JK |
3407 | { |
3408 | int pos; | |
3409 | if (namelen && name[namelen - 1] == '/') | |
3410 | namelen--; | |
3411 | pos = index_name_pos(istate, name, namelen); | |
3412 | if (0 <= pos) | |
3413 | return 0; /* exact match */ | |
3414 | pos = -pos - 1; | |
3415 | if (pos < istate->cache_nr) { | |
3416 | struct cache_entry *ce = istate->cache[pos]; | |
3417 | if (ce_namelen(ce) == namelen && | |
3418 | !memcmp(ce->name, name, namelen)) | |
3419 | return 0; /* Yup, this one exists unmerged */ | |
3420 | } | |
3421 | return 1; | |
3422 | } | |
29fb37b2 | 3423 | |
847a9e5d | 3424 | void *read_blob_data_from_index(struct index_state *istate, |
87542508 | 3425 | const char *path, unsigned long *size) |
29fb37b2 LF |
3426 | { |
3427 | int pos, len; | |
3428 | unsigned long sz; | |
3429 | enum object_type type; | |
3430 | void *data; | |
3431 | ||
3432 | len = strlen(path); | |
3433 | pos = index_name_pos(istate, path, len); | |
3434 | if (pos < 0) { | |
3435 | /* | |
3436 | * We might be in the middle of a merge, in which | |
3437 | * case we would read stage #2 (ours). | |
3438 | */ | |
3439 | int i; | |
3440 | for (i = -pos - 1; | |
3441 | (pos < 0 && i < istate->cache_nr && | |
3442 | !strcmp(istate->cache[i]->name, path)); | |
3443 | i++) | |
3444 | if (ce_stage(istate->cache[i]) == 2) | |
3445 | pos = i; | |
3446 | } | |
3447 | if (pos < 0) | |
3448 | return NULL; | |
b4f5aca4 | 3449 | data = read_object_file(&istate->cache[pos]->oid, &type, &sz); |
29fb37b2 LF |
3450 | if (!data || type != OBJ_BLOB) { |
3451 | free(data); | |
3452 | return NULL; | |
3453 | } | |
ff366825 LF |
3454 | if (size) |
3455 | *size = sz; | |
29fb37b2 LF |
3456 | return data; |
3457 | } | |
38612532 MH |
3458 | |
3459 | void stat_validity_clear(struct stat_validity *sv) | |
3460 | { | |
6a83d902 | 3461 | FREE_AND_NULL(sv->sd); |
38612532 MH |
3462 | } |
3463 | ||
3464 | int stat_validity_check(struct stat_validity *sv, const char *path) | |
3465 | { | |
3466 | struct stat st; | |
3467 | ||
3468 | if (stat(path, &st) < 0) | |
3469 | return sv->sd == NULL; | |
3470 | if (!sv->sd) | |
3471 | return 0; | |
3472 | return S_ISREG(st.st_mode) && !match_stat_data(sv->sd, &st); | |
3473 | } | |
3474 | ||
3475 | void stat_validity_update(struct stat_validity *sv, int fd) | |
3476 | { | |
3477 | struct stat st; | |
3478 | ||
3479 | if (fstat(fd, &st) < 0 || !S_ISREG(st.st_mode)) | |
3480 | stat_validity_clear(sv); | |
3481 | else { | |
3482 | if (!sv->sd) | |
ca56dadb | 3483 | CALLOC_ARRAY(sv->sd, 1); |
38612532 MH |
3484 | fill_stat_data(sv->sd, &st); |
3485 | } | |
3486 | } | |
edf3b905 DT |
3487 | |
3488 | void move_index_extensions(struct index_state *dst, struct index_state *src) | |
3489 | { | |
3490 | dst->untracked = src->untracked; | |
3491 | src->untracked = NULL; | |
836ef2b6 NTND |
3492 | dst->cache_tree = src->cache_tree; |
3493 | src->cache_tree = NULL; | |
edf3b905 | 3494 | } |
a849735b | 3495 | |
8e72d675 JM |
3496 | struct cache_entry *dup_cache_entry(const struct cache_entry *ce, |
3497 | struct index_state *istate) | |
3498 | { | |
3499 | unsigned int size = ce_size(ce); | |
3500 | int mem_pool_allocated; | |
3501 | struct cache_entry *new_entry = make_empty_cache_entry(istate, ce_namelen(ce)); | |
3502 | mem_pool_allocated = new_entry->mem_pool_allocated; | |
3503 | ||
3504 | memcpy(new_entry, ce, size); | |
3505 | new_entry->mem_pool_allocated = mem_pool_allocated; | |
3506 | return new_entry; | |
3507 | } | |
3508 | ||
a849735b JM |
3509 | void discard_cache_entry(struct cache_entry *ce) |
3510 | { | |
8616a2d0 JM |
3511 | if (ce && should_validate_cache_entries()) |
3512 | memset(ce, 0xCD, cache_entry_size(ce->ce_namelen)); | |
3513 | ||
8e72d675 JM |
3514 | if (ce && ce->mem_pool_allocated) |
3515 | return; | |
3516 | ||
a849735b JM |
3517 | free(ce); |
3518 | } | |
8616a2d0 JM |
3519 | |
3520 | int should_validate_cache_entries(void) | |
3521 | { | |
3522 | static int validate_index_cache_entries = -1; | |
3523 | ||
3524 | if (validate_index_cache_entries < 0) { | |
3525 | if (getenv("GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES")) | |
3526 | validate_index_cache_entries = 1; | |
3527 | else | |
3528 | validate_index_cache_entries = 0; | |
3529 | } | |
3530 | ||
3531 | return validate_index_cache_entries; | |
3532 | } | |
3b1d9e04 BP |
3533 | |
3534 | #define EOIE_SIZE (4 + GIT_SHA1_RAWSZ) /* <4-byte offset> + <20-byte hash> */ | |
3535 | #define EOIE_SIZE_WITH_HEADER (4 + 4 + EOIE_SIZE) /* <4-byte signature> + <4-byte length> + EOIE_SIZE */ | |
3536 | ||
3537 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size) | |
3538 | { | |
3539 | /* | |
3540 | * The end of index entries (EOIE) extension is guaranteed to be last | |
3541 | * so that it can be found by scanning backwards from the EOF. | |
3542 | * | |
3543 | * "EOIE" | |
3544 | * <4-byte length> | |
3545 | * <4-byte offset> | |
3546 | * <20-byte hash> | |
3547 | */ | |
3548 | const char *index, *eoie; | |
3549 | uint32_t extsize; | |
3550 | size_t offset, src_offset; | |
3551 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3552 | git_hash_ctx c; | |
3553 | ||
3554 | /* ensure we have an index big enough to contain an EOIE extension */ | |
3555 | if (mmap_size < sizeof(struct cache_header) + EOIE_SIZE_WITH_HEADER + the_hash_algo->rawsz) | |
3556 | return 0; | |
3557 | ||
3558 | /* validate the extension signature */ | |
3559 | index = eoie = mmap + mmap_size - EOIE_SIZE_WITH_HEADER - the_hash_algo->rawsz; | |
3560 | if (CACHE_EXT(index) != CACHE_EXT_ENDOFINDEXENTRIES) | |
3561 | return 0; | |
3562 | index += sizeof(uint32_t); | |
3563 | ||
3564 | /* validate the extension size */ | |
3565 | extsize = get_be32(index); | |
3566 | if (extsize != EOIE_SIZE) | |
3567 | return 0; | |
3568 | index += sizeof(uint32_t); | |
3569 | ||
3570 | /* | |
3571 | * Validate the offset we're going to look for the first extension | |
3572 | * signature is after the index header and before the eoie extension. | |
3573 | */ | |
3574 | offset = get_be32(index); | |
3575 | if (mmap + offset < mmap + sizeof(struct cache_header)) | |
3576 | return 0; | |
3577 | if (mmap + offset >= eoie) | |
3578 | return 0; | |
3579 | index += sizeof(uint32_t); | |
3580 | ||
3581 | /* | |
3582 | * The hash is computed over extension types and their sizes (but not | |
3583 | * their contents). E.g. if we have "TREE" extension that is N-bytes | |
3584 | * long, "REUC" extension that is M-bytes long, followed by "EOIE", | |
3585 | * then the hash would be: | |
3586 | * | |
3587 | * SHA-1("TREE" + <binary representation of N> + | |
3588 | * "REUC" + <binary representation of M>) | |
3589 | */ | |
3590 | src_offset = offset; | |
3591 | the_hash_algo->init_fn(&c); | |
3592 | while (src_offset < mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) { | |
3593 | /* After an array of active_nr index entries, | |
3594 | * there can be arbitrary number of extended | |
3595 | * sections, each of which is prefixed with | |
3596 | * extension name (4-byte) and section length | |
3597 | * in 4-byte network byte order. | |
3598 | */ | |
3599 | uint32_t extsize; | |
3600 | memcpy(&extsize, mmap + src_offset + 4, 4); | |
3601 | extsize = ntohl(extsize); | |
3602 | ||
3603 | /* verify the extension size isn't so large it will wrap around */ | |
3604 | if (src_offset + 8 + extsize < src_offset) | |
3605 | return 0; | |
3606 | ||
3607 | the_hash_algo->update_fn(&c, mmap + src_offset, 8); | |
3608 | ||
3609 | src_offset += 8; | |
3610 | src_offset += extsize; | |
3611 | } | |
3612 | the_hash_algo->final_fn(hash, &c); | |
3613 | if (!hasheq(hash, (const unsigned char *)index)) | |
3614 | return 0; | |
3615 | ||
3616 | /* Validate that the extension offsets returned us back to the eoie extension. */ | |
3617 | if (src_offset != mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) | |
3618 | return 0; | |
3619 | ||
3620 | return offset; | |
3621 | } | |
3622 | ||
3623 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset) | |
3624 | { | |
3625 | uint32_t buffer; | |
3626 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3627 | ||
3628 | /* offset */ | |
3629 | put_be32(&buffer, offset); | |
3630 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3631 | ||
3632 | /* hash */ | |
3633 | the_hash_algo->final_fn(hash, eoie_context); | |
3634 | strbuf_add(sb, hash, the_hash_algo->rawsz); | |
3635 | } | |
3255089a | 3636 | |
3255089a BP |
3637 | #define IEOT_VERSION (1) |
3638 | ||
3639 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset) | |
3640 | { | |
ec36c42a NTND |
3641 | const char *index = NULL; |
3642 | uint32_t extsize, ext_version; | |
3643 | struct index_entry_offset_table *ieot; | |
3644 | int i, nr; | |
3645 | ||
3646 | /* find the IEOT extension */ | |
3647 | if (!offset) | |
3648 | return NULL; | |
3649 | while (offset <= mmap_size - the_hash_algo->rawsz - 8) { | |
3650 | extsize = get_be32(mmap + offset + 4); | |
3651 | if (CACHE_EXT((mmap + offset)) == CACHE_EXT_INDEXENTRYOFFSETTABLE) { | |
3652 | index = mmap + offset + 4 + 4; | |
3653 | break; | |
3654 | } | |
3655 | offset += 8; | |
3656 | offset += extsize; | |
3657 | } | |
3658 | if (!index) | |
3659 | return NULL; | |
3660 | ||
3661 | /* validate the version is IEOT_VERSION */ | |
3662 | ext_version = get_be32(index); | |
3663 | if (ext_version != IEOT_VERSION) { | |
3664 | error("invalid IEOT version %d", ext_version); | |
3665 | return NULL; | |
3666 | } | |
3667 | index += sizeof(uint32_t); | |
3668 | ||
3669 | /* extension size - version bytes / bytes per entry */ | |
3670 | nr = (extsize - sizeof(uint32_t)) / (sizeof(uint32_t) + sizeof(uint32_t)); | |
3671 | if (!nr) { | |
3672 | error("invalid number of IEOT entries %d", nr); | |
3673 | return NULL; | |
3674 | } | |
3675 | ieot = xmalloc(sizeof(struct index_entry_offset_table) | |
3676 | + (nr * sizeof(struct index_entry_offset))); | |
3677 | ieot->nr = nr; | |
3678 | for (i = 0; i < nr; i++) { | |
3679 | ieot->entries[i].offset = get_be32(index); | |
3680 | index += sizeof(uint32_t); | |
3681 | ieot->entries[i].nr = get_be32(index); | |
3682 | index += sizeof(uint32_t); | |
3683 | } | |
3684 | ||
3685 | return ieot; | |
3255089a BP |
3686 | } |
3687 | ||
3688 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot) | |
3689 | { | |
ec36c42a NTND |
3690 | uint32_t buffer; |
3691 | int i; | |
3255089a | 3692 | |
ec36c42a NTND |
3693 | /* version */ |
3694 | put_be32(&buffer, IEOT_VERSION); | |
3695 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3696 | |
ec36c42a NTND |
3697 | /* ieot */ |
3698 | for (i = 0; i < ieot->nr; i++) { | |
3255089a | 3699 | |
ec36c42a NTND |
3700 | /* offset */ |
3701 | put_be32(&buffer, ieot->entries[i].offset); | |
3702 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3703 | |
ec36c42a NTND |
3704 | /* count */ |
3705 | put_be32(&buffer, ieot->entries[i].nr); | |
3706 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3707 | } | |
3255089a | 3708 | } |
b2896d27 JT |
3709 | |
3710 | void prefetch_cache_entries(const struct index_state *istate, | |
3711 | must_prefetch_predicate must_prefetch) | |
3712 | { | |
3713 | int i; | |
3714 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
3715 | ||
3716 | for (i = 0; i < istate->cache_nr; i++) { | |
3717 | struct cache_entry *ce = istate->cache[i]; | |
3718 | ||
3719 | if (S_ISGITLINK(ce->ce_mode) || !must_prefetch(ce)) | |
3720 | continue; | |
3721 | if (!oid_object_info_extended(the_repository, &ce->oid, | |
3722 | NULL, | |
3723 | OBJECT_INFO_FOR_PREFETCH)) | |
3724 | continue; | |
3725 | oid_array_append(&to_fetch, &ce->oid); | |
3726 | } | |
3727 | promisor_remote_get_direct(the_repository, | |
3728 | to_fetch.oid, to_fetch.nr); | |
3729 | oid_array_clear(&to_fetch); | |
3730 | } |