]>
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" |
36bf1958 | 7 | #include "alloc.h" |
6cee5ebc | 8 | #include "bulk-checkin.h" |
b2141fc1 | 9 | #include "config.h" |
d4a4f929 | 10 | #include "date.h" |
cffbfad5 EN |
11 | #include "diff.h" |
12 | #include "diffcore.h" | |
41771fa4 | 13 | #include "hex.h" |
f6ecc62d | 14 | #include "tempfile.h" |
697cc8ef | 15 | #include "lockfile.h" |
bad68ec9 | 16 | #include "cache-tree.h" |
f35a6d3b | 17 | #include "refs.h" |
d616813d | 18 | #include "dir.h" |
87bed179 | 19 | #include "object-file.h" |
cbd53a21 | 20 | #include "object-store.h" |
6f2d7430 | 21 | #include "oid-array.h" |
041aee31 JH |
22 | #include "tree.h" |
23 | #include "commit.h" | |
39425819 | 24 | #include "blob.h" |
32a8f510 | 25 | #include "environment.h" |
f394e093 | 26 | #include "gettext.h" |
5bc07225 | 27 | #include "mem-pool.h" |
f5653856 | 28 | #include "name-hash.h" |
dabab1d6 | 29 | #include "object-name.h" |
cfc5789a | 30 | #include "resolve-undo.h" |
6cee5ebc | 31 | #include "revision.h" |
1956ecd0 | 32 | #include "run-command.h" |
6c9cd161 | 33 | #include "strbuf.h" |
74ea5c95 | 34 | #include "trace2.h" |
6c9cd161 | 35 | #include "varint.h" |
5fc2fc8f | 36 | #include "split-index.h" |
cb2a5135 | 37 | #include "symlinks.h" |
a42643aa | 38 | #include "utf8.h" |
883e248b | 39 | #include "fsmonitor.h" |
abb4bb83 | 40 | #include "thread-utils.h" |
ae9af122 | 41 | #include "progress.h" |
6e773527 | 42 | #include "sparse-index.h" |
410334ed | 43 | #include "csum-file.h" |
b2896d27 | 44 | #include "promisor-remote.h" |
dbb1c613 | 45 | #include "hook.h" |
d5ebb50d | 46 | #include "wrapper.h" |
bad68ec9 | 47 | |
b60e188c TG |
48 | /* Mask for the name length in ce_flags in the on-disk index */ |
49 | ||
50 | #define CE_NAMEMASK (0x0fff) | |
51 | ||
bad68ec9 JH |
52 | /* Index extensions. |
53 | * | |
54 | * The first letter should be 'A'..'Z' for extensions that are not | |
55 | * necessary for a correct operation (i.e. optimization data). | |
56 | * When new extensions are added that _needs_ to be understood in | |
57 | * order to correctly interpret the index file, pick character that | |
58 | * is outside the range, to cause the reader to abort. | |
59 | */ | |
60 | ||
61 | #define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) ) | |
62 | #define CACHE_EXT_TREE 0x54524545 /* "TREE" */ | |
b659b49b | 63 | #define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */ |
5fc2fc8f | 64 | #define CACHE_EXT_LINK 0x6c696e6b /* "link" */ |
83c094ad | 65 | #define CACHE_EXT_UNTRACKED 0x554E5452 /* "UNTR" */ |
883e248b | 66 | #define CACHE_EXT_FSMONITOR 0x46534D4E /* "FSMN" */ |
3b1d9e04 | 67 | #define CACHE_EXT_ENDOFINDEXENTRIES 0x454F4945 /* "EOIE" */ |
3255089a | 68 | #define CACHE_EXT_INDEXENTRYOFFSETTABLE 0x49454F54 /* "IEOT" */ |
cd42415f | 69 | #define CACHE_EXT_SPARSE_DIRECTORIES 0x73646972 /* "sdir" */ |
5fc2fc8f NTND |
70 | |
71 | /* changes that can be kept in $GIT_DIR/index (basically all extensions) */ | |
e0cf0d7d | 72 | #define EXTMASK (RESOLVE_UNDO_CHANGED | CACHE_TREE_CHANGED | \ |
c18b80a0 | 73 | CE_ENTRY_ADDED | CE_ENTRY_REMOVED | CE_ENTRY_CHANGED | \ |
883e248b | 74 | SPLIT_INDEX_ORDERED | UNTRACKED_CHANGED | FSMONITOR_CHANGED) |
e83c5163 | 75 | |
8e72d675 JM |
76 | |
77 | /* | |
78 | * This is an estimate of the pathname length in the index. We use | |
79 | * this for V4 index files to guess the un-deltafied size of the index | |
80 | * in memory because of pathname deltafication. This is not required | |
81 | * for V2/V3 index formats because their pathnames are not compressed. | |
82 | * If the initial amount of memory set aside is not sufficient, the | |
83 | * mem pool will allocate extra memory. | |
84 | */ | |
85 | #define CACHE_ENTRY_PATH_LENGTH 80 | |
86 | ||
20ec2d03 VD |
87 | enum index_search_mode { |
88 | NO_EXPAND_SPARSE = 0, | |
89 | EXPAND_SPARSE = 1 | |
90 | }; | |
91 | ||
8e72d675 JM |
92 | static inline struct cache_entry *mem_pool__ce_alloc(struct mem_pool *mem_pool, size_t len) |
93 | { | |
94 | struct cache_entry *ce; | |
95 | ce = mem_pool_alloc(mem_pool, cache_entry_size(len)); | |
96 | ce->mem_pool_allocated = 1; | |
97 | return ce; | |
98 | } | |
99 | ||
100 | static inline struct cache_entry *mem_pool__ce_calloc(struct mem_pool *mem_pool, size_t len) | |
101 | { | |
102 | struct cache_entry * ce; | |
103 | ce = mem_pool_calloc(mem_pool, 1, cache_entry_size(len)); | |
104 | ce->mem_pool_allocated = 1; | |
105 | return ce; | |
106 | } | |
107 | ||
108 | static struct mem_pool *find_mem_pool(struct index_state *istate) | |
109 | { | |
110 | struct mem_pool **pool_ptr; | |
111 | ||
112 | if (istate->split_index && istate->split_index->base) | |
113 | pool_ptr = &istate->split_index->base->ce_mem_pool; | |
114 | else | |
115 | pool_ptr = &istate->ce_mem_pool; | |
116 | ||
44c7e1a7 EN |
117 | if (!*pool_ptr) { |
118 | *pool_ptr = xmalloc(sizeof(**pool_ptr)); | |
119 | mem_pool_init(*pool_ptr, 0); | |
120 | } | |
8e72d675 JM |
121 | |
122 | return *pool_ptr; | |
123 | } | |
124 | ||
626f35c8 | 125 | static const char *alternate_index_output; |
8fd2cb40 | 126 | |
9cb76b8c JH |
127 | static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
128 | { | |
4300f844 | 129 | if (S_ISSPARSEDIR(ce->ce_mode)) |
9fadb373 | 130 | istate->sparse_index = INDEX_COLLAPSED; |
4300f844 | 131 | |
9cb76b8c | 132 | istate->cache[nr] = ce; |
96872bc2 | 133 | add_name_hash(istate, ce); |
9cb76b8c JH |
134 | } |
135 | ||
cf558704 LT |
136 | static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
137 | { | |
138 | struct cache_entry *old = istate->cache[nr]; | |
139 | ||
078a58e8 | 140 | replace_index_entry_in_base(istate, old, ce); |
2092678c | 141 | remove_name_hash(istate, old); |
a849735b | 142 | discard_cache_entry(old); |
0e267b7a | 143 | ce->ce_flags &= ~CE_HASHED; |
a22c6371 | 144 | set_index_entry(istate, nr, ce); |
078a58e8 | 145 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 146 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 147 | istate->cache_changed |= CE_ENTRY_CHANGED; |
cf558704 LT |
148 | } |
149 | ||
81dc2307 PB |
150 | void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name) |
151 | { | |
b7f9130a | 152 | struct cache_entry *old_entry = istate->cache[nr], *new_entry, *refreshed; |
81dc2307 PB |
153 | int namelen = strlen(new_name); |
154 | ||
a849735b | 155 | new_entry = make_empty_cache_entry(istate, namelen); |
285c2e25 BW |
156 | copy_cache_entry(new_entry, old_entry); |
157 | new_entry->ce_flags &= ~CE_HASHED; | |
158 | new_entry->ce_namelen = namelen; | |
159 | new_entry->index = 0; | |
160 | memcpy(new_entry->name, new_name, namelen + 1); | |
81dc2307 | 161 | |
285c2e25 BW |
162 | cache_tree_invalidate_path(istate, old_entry->name); |
163 | untracked_cache_remove_from_index(istate, old_entry->name); | |
81dc2307 | 164 | remove_index_entry_at(istate, nr); |
b7f9130a VD |
165 | |
166 | /* | |
167 | * Refresh the new index entry. Using 'refresh_cache_entry' ensures | |
168 | * we only update stat info if the entry is otherwise up-to-date (i.e., | |
169 | * the contents/mode haven't changed). This ensures that we reflect the | |
170 | * 'ctime' of the rename in the index without (incorrectly) updating | |
171 | * the cached stat info to reflect unstaged changes on disk. | |
172 | */ | |
173 | refreshed = refresh_cache_entry(istate, new_entry, CE_MATCH_REFRESH); | |
174 | if (refreshed && refreshed != new_entry) { | |
175 | add_index_entry(istate, refreshed, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); | |
176 | discard_cache_entry(new_entry); | |
177 | } else | |
178 | add_index_entry(istate, new_entry, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); | |
81dc2307 PB |
179 | } |
180 | ||
415e96c8 JH |
181 | /* |
182 | * This only updates the "non-critical" parts of the directory | |
183 | * cache, ie the parts that aren't tracked by GIT, and only used | |
184 | * to validate the cache. | |
185 | */ | |
d4c0a3ac | 186 | void fill_stat_cache_info(struct index_state *istate, struct cache_entry *ce, struct stat *st) |
415e96c8 | 187 | { |
c21d39d7 | 188 | fill_stat_data(&ce->ce_stat_data, st); |
5f73076c JH |
189 | |
190 | if (assume_unchanged) | |
7a51ed66 | 191 | ce->ce_flags |= CE_VALID; |
eadb5831 | 192 | |
883e248b | 193 | if (S_ISREG(st->st_mode)) { |
eadb5831 | 194 | ce_mark_uptodate(ce); |
b5a81697 | 195 | mark_fsmonitor_valid(istate, ce); |
883e248b | 196 | } |
415e96c8 JH |
197 | } |
198 | ||
58bf2a4c NTND |
199 | static int ce_compare_data(struct index_state *istate, |
200 | const struct cache_entry *ce, | |
201 | struct stat *st) | |
29e4d363 JH |
202 | { |
203 | int match = -1; | |
1b8ac5ea | 204 | int fd = git_open_cloexec(ce->name, O_RDONLY); |
29e4d363 JH |
205 | |
206 | if (fd >= 0) { | |
bebfecb9 | 207 | struct object_id oid; |
58bf2a4c | 208 | if (!index_fd(istate, &oid, fd, st, OBJ_BLOB, ce->name, 0)) |
6a29d7b7 | 209 | match = !oideq(&oid, &ce->oid); |
7f8508e8 | 210 | /* index_fd() closed the file descriptor already */ |
29e4d363 JH |
211 | } |
212 | return match; | |
213 | } | |
214 | ||
21a6b9fa | 215 | static int ce_compare_link(const struct cache_entry *ce, size_t expected_size) |
29e4d363 JH |
216 | { |
217 | int match = -1; | |
29e4d363 JH |
218 | void *buffer; |
219 | unsigned long size; | |
21666f1a | 220 | enum object_type type; |
a60272b3 | 221 | struct strbuf sb = STRBUF_INIT; |
29e4d363 | 222 | |
a60272b3 | 223 | if (strbuf_readlink(&sb, ce->name, expected_size)) |
29e4d363 | 224 | return -1; |
a60272b3 | 225 | |
bc726bd0 | 226 | buffer = repo_read_object_file(the_repository, &ce->oid, &type, &size); |
a60272b3 LT |
227 | if (buffer) { |
228 | if (size == sb.len) | |
229 | match = memcmp(buffer, sb.buf, size); | |
230 | free(buffer); | |
29e4d363 | 231 | } |
a60272b3 | 232 | strbuf_release(&sb); |
29e4d363 JH |
233 | return match; |
234 | } | |
235 | ||
21a6b9fa | 236 | static int ce_compare_gitlink(const struct cache_entry *ce) |
f35a6d3b | 237 | { |
1053fe82 | 238 | struct object_id oid; |
f35a6d3b LT |
239 | |
240 | /* | |
241 | * We don't actually require that the .git directory | |
302b9282 | 242 | * under GITLINK directory be a valid git directory. It |
f35a6d3b LT |
243 | * might even be missing (in case nobody populated that |
244 | * sub-project). | |
245 | * | |
246 | * If so, we consider it always to match. | |
247 | */ | |
a98e6101 | 248 | if (resolve_gitlink_ref(ce->name, "HEAD", &oid) < 0) |
f35a6d3b | 249 | return 0; |
6a29d7b7 | 250 | return !oideq(&oid, &ce->oid); |
f35a6d3b LT |
251 | } |
252 | ||
58bf2a4c NTND |
253 | static int ce_modified_check_fs(struct index_state *istate, |
254 | const struct cache_entry *ce, | |
255 | struct stat *st) | |
29e4d363 JH |
256 | { |
257 | switch (st->st_mode & S_IFMT) { | |
258 | case S_IFREG: | |
58bf2a4c | 259 | if (ce_compare_data(istate, ce, st)) |
29e4d363 JH |
260 | return DATA_CHANGED; |
261 | break; | |
262 | case S_IFLNK: | |
dc49cd76 | 263 | if (ce_compare_link(ce, xsize_t(st->st_size))) |
29e4d363 JH |
264 | return DATA_CHANGED; |
265 | break; | |
a8ee75bc | 266 | case S_IFDIR: |
7a51ed66 | 267 | if (S_ISGITLINK(ce->ce_mode)) |
c70115b4 | 268 | return ce_compare_gitlink(ce) ? DATA_CHANGED : 0; |
1cf01a34 | 269 | /* else fallthrough */ |
29e4d363 JH |
270 | default: |
271 | return TYPE_CHANGED; | |
272 | } | |
273 | return 0; | |
274 | } | |
275 | ||
21a6b9fa | 276 | static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st) |
734aab75 LT |
277 | { |
278 | unsigned int changed = 0; | |
279 | ||
7a51ed66 LT |
280 | if (ce->ce_flags & CE_REMOVE) |
281 | return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED; | |
282 | ||
283 | switch (ce->ce_mode & S_IFMT) { | |
8ae0a8c5 KS |
284 | case S_IFREG: |
285 | changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0; | |
3e09cdfd JH |
286 | /* We consider only the owner x bit to be relevant for |
287 | * "mode changes" | |
288 | */ | |
289 | if (trust_executable_bit && | |
7a51ed66 | 290 | (0100 & (ce->ce_mode ^ st->st_mode))) |
ffbe1add | 291 | changed |= MODE_CHANGED; |
8ae0a8c5 KS |
292 | break; |
293 | case S_IFLNK: | |
78a8d641 JS |
294 | if (!S_ISLNK(st->st_mode) && |
295 | (has_symlinks || !S_ISREG(st->st_mode))) | |
296 | changed |= TYPE_CHANGED; | |
8ae0a8c5 | 297 | break; |
302b9282 | 298 | case S_IFGITLINK: |
c70115b4 | 299 | /* We ignore most of the st_xxx fields for gitlinks */ |
f35a6d3b LT |
300 | if (!S_ISDIR(st->st_mode)) |
301 | changed |= TYPE_CHANGED; | |
302 | else if (ce_compare_gitlink(ce)) | |
303 | changed |= DATA_CHANGED; | |
a8ee75bc | 304 | return changed; |
8ae0a8c5 | 305 | default: |
391408e5 | 306 | BUG("unsupported ce_mode: %o", ce->ce_mode); |
8ae0a8c5 | 307 | } |
2cb45e95 | 308 | |
c21d39d7 | 309 | changed |= match_stat_data(&ce->ce_stat_data, st); |
b0391890 | 310 | |
f49c2c22 | 311 | /* Racily smudged entry? */ |
c21d39d7 | 312 | if (!ce->ce_stat_data.sd_size) { |
99d1a986 | 313 | if (!is_empty_blob_sha1(ce->oid.hash)) |
f49c2c22 LT |
314 | changed |= DATA_CHANGED; |
315 | } | |
316 | ||
407c8eb0 JH |
317 | return changed; |
318 | } | |
319 | ||
2bb4cda1 NTND |
320 | static int is_racy_stat(const struct index_state *istate, |
321 | const struct stat_data *sd) | |
6d91da6d | 322 | { |
2bb4cda1 | 323 | return (istate->timestamp.sec && |
fba2f38a KB |
324 | #ifdef USE_NSEC |
325 | /* nanosecond timestamped files can also be racy! */ | |
2bb4cda1 NTND |
326 | (istate->timestamp.sec < sd->sd_mtime.sec || |
327 | (istate->timestamp.sec == sd->sd_mtime.sec && | |
328 | istate->timestamp.nsec <= sd->sd_mtime.nsec)) | |
fba2f38a | 329 | #else |
2bb4cda1 | 330 | istate->timestamp.sec <= sd->sd_mtime.sec |
fba2f38a | 331 | #endif |
2bb4cda1 NTND |
332 | ); |
333 | } | |
334 | ||
5581a019 | 335 | int is_racy_timestamp(const struct index_state *istate, |
2bb4cda1 NTND |
336 | const struct cache_entry *ce) |
337 | { | |
338 | return (!S_ISGITLINK(ce->ce_mode) && | |
339 | is_racy_stat(istate, &ce->ce_stat_data)); | |
6d91da6d JH |
340 | } |
341 | ||
ed4efab1 NTND |
342 | int match_stat_data_racy(const struct index_state *istate, |
343 | const struct stat_data *sd, struct stat *st) | |
344 | { | |
345 | if (is_racy_stat(istate, sd)) | |
346 | return MTIME_CHANGED; | |
347 | return match_stat_data(sd, st); | |
6d91da6d JH |
348 | } |
349 | ||
883e248b | 350 | int ie_match_stat(struct index_state *istate, |
21a6b9fa | 351 | const struct cache_entry *ce, struct stat *st, |
4bd5b7da | 352 | unsigned int options) |
407c8eb0 | 353 | { |
5f73076c | 354 | unsigned int changed; |
4bd5b7da | 355 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 356 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
4bd5b7da | 357 | int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY; |
883e248b | 358 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
5f73076c | 359 | |
883e248b BP |
360 | if (!ignore_fsmonitor) |
361 | refresh_fsmonitor(istate); | |
5f73076c JH |
362 | /* |
363 | * If it's marked as always valid in the index, it's | |
364 | * valid whatever the checked-out copy says. | |
56cac48c NTND |
365 | * |
366 | * skip-worktree has the same effect with higher precedence | |
5f73076c | 367 | */ |
56cac48c NTND |
368 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) |
369 | return 0; | |
7a51ed66 | 370 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) |
5f73076c | 371 | return 0; |
883e248b BP |
372 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) |
373 | return 0; | |
5f73076c | 374 | |
331fcb59 JH |
375 | /* |
376 | * Intent-to-add entries have not been added, so the index entry | |
377 | * by definition never matches what is in the work tree until it | |
378 | * actually gets added. | |
379 | */ | |
895ff3b2 | 380 | if (ce_intent_to_add(ce)) |
331fcb59 JH |
381 | return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED; |
382 | ||
5f73076c | 383 | changed = ce_match_stat_basic(ce, st); |
407c8eb0 | 384 | |
29e4d363 JH |
385 | /* |
386 | * Within 1 second of this sequence: | |
387 | * echo xyzzy >file && git-update-index --add file | |
388 | * running this command: | |
389 | * echo frotz >file | |
390 | * would give a falsely clean cache entry. The mtime and | |
391 | * length match the cache, and other stat fields do not change. | |
392 | * | |
393 | * We could detect this at update-index time (the cache entry | |
394 | * being registered/updated records the same time as "now") | |
395 | * and delay the return from git-update-index, but that would | |
396 | * effectively mean we can make at most one commit per second, | |
397 | * which is not acceptable. Instead, we check cache entries | |
398 | * whose mtime are the same as the index file timestamp more | |
5f73076c | 399 | * carefully than others. |
29e4d363 | 400 | */ |
6d91da6d | 401 | if (!changed && is_racy_timestamp(istate, ce)) { |
42f77406 JH |
402 | if (assume_racy_is_modified) |
403 | changed |= DATA_CHANGED; | |
404 | else | |
58bf2a4c | 405 | changed |= ce_modified_check_fs(istate, ce, st); |
42f77406 | 406 | } |
b0391890 | 407 | |
29e4d363 | 408 | return changed; |
b0391890 JH |
409 | } |
410 | ||
883e248b | 411 | int ie_modified(struct index_state *istate, |
21a6b9fa RS |
412 | const struct cache_entry *ce, |
413 | struct stat *st, unsigned int options) | |
b0391890 | 414 | { |
29e4d363 | 415 | int changed, changed_fs; |
4bd5b7da JH |
416 | |
417 | changed = ie_match_stat(istate, ce, st, options); | |
b0391890 JH |
418 | if (!changed) |
419 | return 0; | |
b0391890 JH |
420 | /* |
421 | * If the mode or type has changed, there's no point in trying | |
422 | * to refresh the entry - it's not going to match | |
423 | */ | |
424 | if (changed & (MODE_CHANGED | TYPE_CHANGED)) | |
425 | return changed; | |
426 | ||
c70115b4 JH |
427 | /* |
428 | * Immediately after read-tree or update-index --cacheinfo, | |
429 | * the length field is zero, as we have never even read the | |
430 | * lstat(2) information once, and we cannot trust DATA_CHANGED | |
431 | * returned by ie_match_stat() which in turn was returned by | |
432 | * ce_match_stat_basic() to signal that the filesize of the | |
433 | * blob changed. We have to actually go to the filesystem to | |
434 | * see if the contents match, and if so, should answer "unchanged". | |
435 | * | |
436 | * The logic does not apply to gitlinks, as ce_match_stat_basic() | |
437 | * already has checked the actual HEAD from the filesystem in the | |
438 | * subproject. If ie_match_stat() already said it is different, | |
439 | * then we know it is. | |
b0391890 | 440 | */ |
c70115b4 | 441 | if ((changed & DATA_CHANGED) && |
c21d39d7 | 442 | (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0)) |
b0391890 JH |
443 | return changed; |
444 | ||
58bf2a4c | 445 | changed_fs = ce_modified_check_fs(istate, ce, st); |
29e4d363 JH |
446 | if (changed_fs) |
447 | return changed | changed_fs; | |
b0391890 JH |
448 | return 0; |
449 | } | |
450 | ||
aabc5617 EN |
451 | static int cache_name_stage_compare(const char *name1, int len1, int stage1, |
452 | const char *name2, int len2, int stage2) | |
958ba6c9 | 453 | { |
958ba6c9 LT |
454 | int cmp; |
455 | ||
ccdd4a0f | 456 | cmp = name_compare(name1, len1, name2, len2); |
0ab9e1e8 LT |
457 | if (cmp) |
458 | return cmp; | |
0ab9e1e8 | 459 | |
b60e188c | 460 | if (stage1 < stage2) |
eb38c22f | 461 | return -1; |
b60e188c | 462 | if (stage1 > stage2) |
eb38c22f | 463 | return 1; |
ccdd4a0f JM |
464 | return 0; |
465 | } | |
466 | ||
aabc5617 | 467 | int cmp_cache_name_compare(const void *a_, const void *b_) |
ccdd4a0f | 468 | { |
aabc5617 | 469 | const struct cache_entry *ce1, *ce2; |
ccdd4a0f | 470 | |
aabc5617 EN |
471 | ce1 = *((const struct cache_entry **)a_); |
472 | ce2 = *((const struct cache_entry **)b_); | |
473 | return cache_name_stage_compare(ce1->name, ce1->ce_namelen, ce_stage(ce1), | |
474 | ce2->name, ce2->ce_namelen, ce_stage(ce2)); | |
eb38c22f LT |
475 | } |
476 | ||
20ec2d03 VD |
477 | static int index_name_stage_pos(struct index_state *istate, |
478 | const char *name, int namelen, | |
479 | int stage, | |
480 | enum index_search_mode search_mode) | |
eb38c22f LT |
481 | { |
482 | int first, last; | |
483 | ||
484 | first = 0; | |
4aab5b46 | 485 | last = istate->cache_nr; |
eb38c22f | 486 | while (last > first) { |
568a05c5 | 487 | int next = first + ((last - first) >> 1); |
4aab5b46 | 488 | struct cache_entry *ce = istate->cache[next]; |
b60e188c | 489 | int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce)); |
eb38c22f | 490 | if (!cmp) |
76e7f4ec | 491 | return next; |
eb38c22f LT |
492 | if (cmp < 0) { |
493 | last = next; | |
494 | continue; | |
495 | } | |
496 | first = next+1; | |
497 | } | |
95e0321c | 498 | |
20ec2d03 | 499 | if (search_mode == EXPAND_SPARSE && istate->sparse_index && |
95e0321c DS |
500 | first > 0) { |
501 | /* Note: first <= istate->cache_nr */ | |
502 | struct cache_entry *ce = istate->cache[first - 1]; | |
503 | ||
504 | /* | |
505 | * If we are in a sparse-index _and_ the entry before the | |
506 | * insertion position is a sparse-directory entry that is | |
507 | * an ancestor of 'name', then we need to expand the index | |
508 | * and search again. This will only trigger once, because | |
509 | * thereafter the index is fully expanded. | |
510 | */ | |
511 | if (S_ISSPARSEDIR(ce->ce_mode) && | |
512 | ce_namelen(ce) < namelen && | |
513 | !strncmp(name, ce->name, ce_namelen(ce))) { | |
514 | ensure_full_index(istate); | |
20ec2d03 | 515 | return index_name_stage_pos(istate, name, namelen, stage, search_mode); |
95e0321c DS |
516 | } |
517 | } | |
518 | ||
76e7f4ec | 519 | return -first-1; |
eb38c22f LT |
520 | } |
521 | ||
847a9e5d | 522 | int index_name_pos(struct index_state *istate, const char *name, int namelen) |
b60e188c | 523 | { |
20ec2d03 VD |
524 | return index_name_stage_pos(istate, name, namelen, 0, EXPAND_SPARSE); |
525 | } | |
526 | ||
9553aa0f VD |
527 | int index_name_pos_sparse(struct index_state *istate, const char *name, int namelen) |
528 | { | |
529 | return index_name_stage_pos(istate, name, namelen, 0, NO_EXPAND_SPARSE); | |
530 | } | |
531 | ||
20ec2d03 VD |
532 | int index_entry_exists(struct index_state *istate, const char *name, int namelen) |
533 | { | |
534 | return index_name_stage_pos(istate, name, namelen, 0, NO_EXPAND_SPARSE) >= 0; | |
b60e188c TG |
535 | } |
536 | ||
4aab5b46 | 537 | int remove_index_entry_at(struct index_state *istate, int pos) |
7b937ca3 | 538 | { |
cf558704 LT |
539 | struct cache_entry *ce = istate->cache[pos]; |
540 | ||
cfc5789a | 541 | record_resolve_undo(istate, ce); |
2092678c | 542 | remove_name_hash(istate, ce); |
045113a5 | 543 | save_or_free_index_entry(istate, ce); |
e636a7b4 | 544 | istate->cache_changed |= CE_ENTRY_REMOVED; |
4aab5b46 JH |
545 | istate->cache_nr--; |
546 | if (pos >= istate->cache_nr) | |
7b937ca3 | 547 | return 0; |
f331ab9d RS |
548 | MOVE_ARRAY(istate->cache + pos, istate->cache + pos + 1, |
549 | istate->cache_nr - pos); | |
7b937ca3 LT |
550 | return 1; |
551 | } | |
552 | ||
36419c8e | 553 | /* |
98e023de | 554 | * Remove all cache entries marked for removal, that is where |
36419c8e KB |
555 | * CE_REMOVE is set in ce_flags. This is much more effective than |
556 | * calling remove_index_entry_at() for each entry to be removed. | |
557 | */ | |
6fdc2057 | 558 | void remove_marked_cache_entries(struct index_state *istate, int invalidate) |
36419c8e KB |
559 | { |
560 | struct cache_entry **ce_array = istate->cache; | |
561 | unsigned int i, j; | |
562 | ||
563 | for (i = j = 0; i < istate->cache_nr; i++) { | |
5699d17e | 564 | if (ce_array[i]->ce_flags & CE_REMOVE) { |
6fdc2057 TG |
565 | if (invalidate) { |
566 | cache_tree_invalidate_path(istate, | |
567 | ce_array[i]->name); | |
568 | untracked_cache_remove_from_index(istate, | |
569 | ce_array[i]->name); | |
570 | } | |
2092678c | 571 | remove_name_hash(istate, ce_array[i]); |
045113a5 | 572 | save_or_free_index_entry(istate, ce_array[i]); |
5699d17e | 573 | } |
36419c8e KB |
574 | else |
575 | ce_array[j++] = ce_array[i]; | |
576 | } | |
ad837d9e NTND |
577 | if (j == istate->cache_nr) |
578 | return; | |
e636a7b4 | 579 | istate->cache_changed |= CE_ENTRY_REMOVED; |
36419c8e KB |
580 | istate->cache_nr = j; |
581 | } | |
582 | ||
4aab5b46 | 583 | int remove_file_from_index(struct index_state *istate, const char *path) |
197ee8c9 | 584 | { |
4aab5b46 | 585 | int pos = index_name_pos(istate, path, strlen(path)); |
c4e3cca1 JH |
586 | if (pos < 0) |
587 | pos = -pos-1; | |
a5400efe | 588 | cache_tree_invalidate_path(istate, path); |
e931371a | 589 | untracked_cache_remove_from_index(istate, path); |
4aab5b46 JH |
590 | while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path)) |
591 | remove_index_entry_at(istate, pos); | |
197ee8c9 LT |
592 | return 0; |
593 | } | |
594 | ||
20314271 JS |
595 | static int compare_name(struct cache_entry *ce, const char *path, int namelen) |
596 | { | |
597 | return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen); | |
598 | } | |
599 | ||
600 | static int index_name_pos_also_unmerged(struct index_state *istate, | |
601 | const char *path, int namelen) | |
602 | { | |
603 | int pos = index_name_pos(istate, path, namelen); | |
604 | struct cache_entry *ce; | |
605 | ||
606 | if (pos >= 0) | |
607 | return pos; | |
608 | ||
609 | /* maybe unmerged? */ | |
610 | pos = -1 - pos; | |
611 | if (pos >= istate->cache_nr || | |
612 | compare_name((ce = istate->cache[pos]), path, namelen)) | |
613 | return -1; | |
614 | ||
615 | /* order of preference: stage 2, 1, 3 */ | |
616 | if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr && | |
617 | ce_stage((ce = istate->cache[pos + 1])) == 2 && | |
618 | !compare_name(ce, path, namelen)) | |
619 | pos++; | |
620 | return pos; | |
621 | } | |
622 | ||
1102952b LT |
623 | static int different_name(struct cache_entry *ce, struct cache_entry *alias) |
624 | { | |
625 | int len = ce_namelen(ce); | |
626 | return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len); | |
627 | } | |
628 | ||
629 | /* | |
630 | * If we add a filename that aliases in the cache, we will use the | |
631 | * name that we already have - but we don't want to update the same | |
632 | * alias twice, because that implies that there were actually two | |
633 | * different files with aliasing names! | |
634 | * | |
635 | * So we use the CE_ADDED flag to verify that the alias was an old | |
636 | * one before we accept it as | |
637 | */ | |
045113a5 NTND |
638 | static struct cache_entry *create_alias_ce(struct index_state *istate, |
639 | struct cache_entry *ce, | |
640 | struct cache_entry *alias) | |
1102952b LT |
641 | { |
642 | int len; | |
285c2e25 | 643 | struct cache_entry *new_entry; |
1102952b LT |
644 | |
645 | if (alias->ce_flags & CE_ADDED) | |
9d0a9e90 NTND |
646 | die(_("will not add file alias '%s' ('%s' already exists in index)"), |
647 | ce->name, alias->name); | |
1102952b LT |
648 | |
649 | /* Ok, create the new entry using the name of the existing alias */ | |
650 | len = ce_namelen(alias); | |
a849735b | 651 | new_entry = make_empty_cache_entry(istate, len); |
285c2e25 BW |
652 | memcpy(new_entry->name, alias->name, len); |
653 | copy_cache_entry(new_entry, ce); | |
045113a5 | 654 | save_or_free_index_entry(istate, ce); |
285c2e25 | 655 | return new_entry; |
1102952b LT |
656 | } |
657 | ||
b4b313f9 | 658 | void set_object_name_for_intent_to_add_entry(struct cache_entry *ce) |
39425819 | 659 | { |
a09c985e | 660 | struct object_id oid; |
c80d226a | 661 | if (write_object_file("", 0, OBJ_BLOB, &oid)) |
9d0a9e90 | 662 | die(_("cannot create an empty blob in the object database")); |
a09c985e | 663 | oidcpy(&ce->oid, &oid); |
39425819 JH |
664 | } |
665 | ||
610d55af | 666 | int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags) |
11be42a4 | 667 | { |
a849735b | 668 | int namelen, was_same; |
d177cab0 | 669 | mode_t st_mode = st->st_mode; |
9472935d | 670 | struct cache_entry *ce, *alias = NULL; |
56cac48c | 671 | unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY; |
38ed1d89 JH |
672 | int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND); |
673 | int pretend = flags & ADD_CACHE_PRETEND; | |
39425819 JH |
674 | int intent_only = flags & ADD_CACHE_INTENT; |
675 | int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE| | |
676 | (intent_only ? ADD_CACHE_NEW_ONLY : 0)); | |
e578d031 | 677 | unsigned hash_flags = pretend ? 0 : HASH_WRITE_OBJECT; |
f937bc2f | 678 | struct object_id oid; |
9472935d | 679 | |
9e5da3d0 JK |
680 | if (flags & ADD_CACHE_RENORMALIZE) |
681 | hash_flags |= HASH_RENORMALIZE; | |
11be42a4 | 682 | |
d177cab0 | 683 | if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode)) |
9d0a9e90 | 684 | return error(_("%s: can only add regular files, symbolic links or git-directories"), path); |
11be42a4 JS |
685 | |
686 | namelen = strlen(path); | |
d177cab0 | 687 | if (S_ISDIR(st_mode)) { |
f937bc2f KM |
688 | if (resolve_gitlink_ref(path, "HEAD", &oid) < 0) |
689 | return error(_("'%s' does not have a commit checked out"), path); | |
09595258 LT |
690 | while (namelen && path[namelen-1] == '/') |
691 | namelen--; | |
692 | } | |
a849735b | 693 | ce = make_empty_cache_entry(istate, namelen); |
11be42a4 | 694 | memcpy(ce->name, path, namelen); |
b60e188c | 695 | ce->ce_namelen = namelen; |
39425819 | 696 | if (!intent_only) |
d4c0a3ac | 697 | fill_stat_cache_info(istate, ce, st); |
388b2acd JH |
698 | else |
699 | ce->ce_flags |= CE_INTENT_TO_ADD; | |
11be42a4 | 700 | |
610d55af TG |
701 | |
702 | if (trust_executable_bit && has_symlinks) { | |
d177cab0 | 703 | ce->ce_mode = create_ce_mode(st_mode); |
610d55af | 704 | } else { |
78a8d641 JS |
705 | /* If there is an existing entry, pick the mode bits and type |
706 | * from it, otherwise assume unexecutable regular file. | |
11be42a4 | 707 | */ |
185c975f | 708 | struct cache_entry *ent; |
20314271 | 709 | int pos = index_name_pos_also_unmerged(istate, path, namelen); |
185c975f | 710 | |
4aab5b46 | 711 | ent = (0 <= pos) ? istate->cache[pos] : NULL; |
d177cab0 | 712 | ce->ce_mode = ce_mode_from_stat(ent, st_mode); |
11be42a4 JS |
713 | } |
714 | ||
dc1ae704 JJ |
715 | /* When core.ignorecase=true, determine if a directory of the same name but differing |
716 | * case already exists within the Git repository. If it does, ensure the directory | |
717 | * case of the file being added to the repository matches (is folded into) the existing | |
718 | * entry's directory case. | |
719 | */ | |
720 | if (ignore_case) { | |
41284eb0 | 721 | adjust_dirname_case(istate, ce->name); |
dc1ae704 | 722 | } |
e2c2a375 | 723 | if (!(flags & ADD_CACHE_RENORMALIZE)) { |
9472935d TB |
724 | alias = index_file_exists(istate, ce->name, |
725 | ce_namelen(ce), ignore_case); | |
726 | if (alias && | |
727 | !ce_stage(alias) && | |
728 | !ie_match_stat(istate, alias, st, ce_option)) { | |
729 | /* Nothing changed, really */ | |
730 | if (!S_ISGITLINK(alias->ce_mode)) | |
731 | ce_mark_uptodate(alias); | |
732 | alias->ce_flags |= CE_ADDED; | |
dc1ae704 | 733 | |
a849735b | 734 | discard_cache_entry(ce); |
9472935d TB |
735 | return 0; |
736 | } | |
0781b8a9 | 737 | } |
39425819 | 738 | if (!intent_only) { |
1c418243 | 739 | if (index_path(istate, &ce->oid, path, st, hash_flags)) { |
a849735b | 740 | discard_cache_entry(ce); |
9d0a9e90 | 741 | return error(_("unable to index file '%s'"), path); |
2d9426b0 | 742 | } |
39425819 | 743 | } else |
b4b313f9 | 744 | set_object_name_for_intent_to_add_entry(ce); |
39425819 | 745 | |
1102952b | 746 | if (ignore_case && alias && different_name(ce, alias)) |
045113a5 | 747 | ce = create_alias_ce(istate, ce, alias); |
1102952b | 748 | ce->ce_flags |= CE_ADDED; |
38ed1d89 | 749 | |
3bf0dd1f | 750 | /* It was suspected to be racily clean, but it turns out to be Ok */ |
38ed1d89 JH |
751 | was_same = (alias && |
752 | !ce_stage(alias) && | |
4a7e27e9 | 753 | oideq(&alias->oid, &ce->oid) && |
38ed1d89 JH |
754 | ce->ce_mode == alias->ce_mode); |
755 | ||
756 | if (pretend) | |
a849735b | 757 | discard_cache_entry(ce); |
067178ed | 758 | else if (add_index_entry(istate, ce, add_option)) { |
a849735b | 759 | discard_cache_entry(ce); |
9d0a9e90 | 760 | return error(_("unable to add '%s' to index"), path); |
067178ed | 761 | } |
38ed1d89 | 762 | if (verbose && !was_same) |
11be42a4 | 763 | printf("add '%s'\n", path); |
11be42a4 JS |
764 | return 0; |
765 | } | |
766 | ||
610d55af | 767 | int add_file_to_index(struct index_state *istate, const char *path, int flags) |
d177cab0 LT |
768 | { |
769 | struct stat st; | |
770 | if (lstat(path, &st)) | |
9d0a9e90 | 771 | die_errno(_("unable to stat '%s'"), path); |
610d55af | 772 | return add_to_index(istate, path, &st, flags); |
d177cab0 LT |
773 | } |
774 | ||
a849735b JM |
775 | struct cache_entry *make_empty_cache_entry(struct index_state *istate, size_t len) |
776 | { | |
8e72d675 | 777 | return mem_pool__ce_calloc(find_mem_pool(istate), len); |
a849735b JM |
778 | } |
779 | ||
96168827 MT |
780 | struct cache_entry *make_empty_transient_cache_entry(size_t len, |
781 | struct mem_pool *ce_mem_pool) | |
a849735b | 782 | { |
96168827 MT |
783 | if (ce_mem_pool) |
784 | return mem_pool__ce_calloc(ce_mem_pool, len); | |
a849735b JM |
785 | return xcalloc(1, cache_entry_size(len)); |
786 | } | |
787 | ||
2a1ae649 RS |
788 | enum verify_path_result { |
789 | PATH_OK, | |
790 | PATH_INVALID, | |
791 | PATH_DIR_WITH_SEP, | |
792 | }; | |
793 | ||
794 | static enum verify_path_result verify_path_internal(const char *, unsigned); | |
795 | ||
796 | int verify_path(const char *path, unsigned mode) | |
797 | { | |
c8ad9d04 | 798 | return verify_path_internal(path, mode) == PATH_OK; |
2a1ae649 RS |
799 | } |
800 | ||
a849735b JM |
801 | struct cache_entry *make_cache_entry(struct index_state *istate, |
802 | unsigned int mode, | |
825ed4d9 JM |
803 | const struct object_id *oid, |
804 | const char *path, | |
805 | int stage, | |
806 | unsigned int refresh_options) | |
6640f881 | 807 | { |
bc1c2caa | 808 | struct cache_entry *ce, *ret; |
a849735b | 809 | int len; |
6640f881 | 810 | |
c8ad9d04 | 811 | if (verify_path_internal(path, mode) == PATH_INVALID) { |
9d0a9e90 | 812 | error(_("invalid path '%s'"), path); |
6640f881 | 813 | return NULL; |
7e7abea9 | 814 | } |
6640f881 CR |
815 | |
816 | len = strlen(path); | |
a849735b | 817 | ce = make_empty_cache_entry(istate, len); |
6640f881 | 818 | |
825ed4d9 | 819 | oidcpy(&ce->oid, oid); |
6640f881 | 820 | memcpy(ce->name, path, len); |
b60e188c TG |
821 | ce->ce_flags = create_ce_flags(stage); |
822 | ce->ce_namelen = len; | |
6640f881 CR |
823 | ce->ce_mode = create_ce_mode(mode); |
824 | ||
d7b665c3 | 825 | ret = refresh_cache_entry(istate, ce, refresh_options); |
915e44c6 | 826 | if (ret != ce) |
a849735b | 827 | discard_cache_entry(ce); |
915e44c6 | 828 | return ret; |
6640f881 CR |
829 | } |
830 | ||
96168827 MT |
831 | struct cache_entry *make_transient_cache_entry(unsigned int mode, |
832 | const struct object_id *oid, | |
833 | const char *path, | |
834 | int stage, | |
835 | struct mem_pool *ce_mem_pool) | |
a849735b JM |
836 | { |
837 | struct cache_entry *ce; | |
838 | int len; | |
839 | ||
840 | if (!verify_path(path, mode)) { | |
9d0a9e90 | 841 | error(_("invalid path '%s'"), path); |
a849735b JM |
842 | return NULL; |
843 | } | |
844 | ||
845 | len = strlen(path); | |
96168827 | 846 | ce = make_empty_transient_cache_entry(len, ce_mem_pool); |
a849735b JM |
847 | |
848 | oidcpy(&ce->oid, oid); | |
849 | memcpy(ce->name, path, len); | |
850 | ce->ce_flags = create_ce_flags(stage); | |
851 | ce->ce_namelen = len; | |
852 | ce->ce_mode = create_ce_mode(mode); | |
853 | ||
854 | return ce; | |
855 | } | |
856 | ||
d9d70966 TG |
857 | /* |
858 | * Chmod an index entry with either +x or -x. | |
859 | * | |
860 | * Returns -1 if the chmod for the particular cache entry failed (if it's | |
861 | * not a regular file), -2 if an invalid flip argument is passed in, 0 | |
862 | * otherwise. | |
863 | */ | |
864 | int chmod_index_entry(struct index_state *istate, struct cache_entry *ce, | |
865 | char flip) | |
866 | { | |
867 | if (!S_ISREG(ce->ce_mode)) | |
868 | return -1; | |
869 | switch (flip) { | |
870 | case '+': | |
871 | ce->ce_mode |= 0111; | |
872 | break; | |
873 | case '-': | |
874 | ce->ce_mode &= ~0111; | |
875 | break; | |
876 | default: | |
877 | return -2; | |
878 | } | |
879 | cache_tree_invalidate_path(istate, ce->name); | |
880 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
883e248b | 881 | mark_fsmonitor_invalid(istate, ce); |
d9d70966 TG |
882 | istate->cache_changed |= CE_ENTRY_CHANGED; |
883 | ||
884 | return 0; | |
885 | } | |
886 | ||
9c5e6c80 | 887 | int ce_same_name(const struct cache_entry *a, const struct cache_entry *b) |
7b937ca3 LT |
888 | { |
889 | int len = ce_namelen(a); | |
890 | return ce_namelen(b) == len && !memcmp(a->name, b->name, len); | |
891 | } | |
892 | ||
8dcf39c4 LT |
893 | /* |
894 | * We fundamentally don't like some paths: we don't want | |
895 | * dot or dot-dot anywhere, and for obvious reasons don't | |
896 | * want to recurse into ".git" either. | |
897 | * | |
898 | * Also, we don't want double slashes or slashes at the | |
899 | * end that can make pathnames ambiguous. | |
900 | */ | |
10ecfa76 | 901 | static int verify_dotfile(const char *rest, unsigned mode) |
8dcf39c4 LT |
902 | { |
903 | /* | |
904 | * The first character was '.', but that | |
905 | * has already been discarded, we now test | |
906 | * the rest. | |
907 | */ | |
e0f530ff | 908 | |
8dcf39c4 | 909 | /* "." is not allowed */ |
e0f530ff | 910 | if (*rest == '\0' || is_dir_sep(*rest)) |
8dcf39c4 LT |
911 | return 0; |
912 | ||
e0f530ff | 913 | switch (*rest) { |
8dcf39c4 | 914 | /* |
641084b6 JK |
915 | * ".git" followed by NUL or slash is bad. Note that we match |
916 | * case-insensitively here, even if ignore_case is not set. | |
917 | * This outlaws ".GIT" everywhere out of an abundance of caution, | |
918 | * since there's really no good reason to allow it. | |
10ecfa76 JK |
919 | * |
920 | * Once we've seen ".git", we can also find ".gitmodules", etc (also | |
921 | * case-insensitively). | |
8dcf39c4 LT |
922 | */ |
923 | case 'g': | |
cc2fc7c2 JK |
924 | case 'G': |
925 | if (rest[1] != 'i' && rest[1] != 'I') | |
8dcf39c4 | 926 | break; |
cc2fc7c2 | 927 | if (rest[2] != 't' && rest[2] != 'T') |
8dcf39c4 | 928 | break; |
e19e5e66 JK |
929 | if (rest[3] == '\0' || is_dir_sep(rest[3])) |
930 | return 0; | |
10ecfa76 JK |
931 | if (S_ISLNK(mode)) { |
932 | rest += 3; | |
933 | if (skip_iprefix(rest, "modules", &rest) && | |
934 | (*rest == '\0' || is_dir_sep(*rest))) | |
935 | return 0; | |
936 | } | |
e19e5e66 | 937 | break; |
8dcf39c4 | 938 | case '.': |
e0f530ff | 939 | if (rest[1] == '\0' || is_dir_sep(rest[1])) |
8dcf39c4 LT |
940 | return 0; |
941 | } | |
942 | return 1; | |
943 | } | |
944 | ||
2a1ae649 RS |
945 | static enum verify_path_result verify_path_internal(const char *path, |
946 | unsigned mode) | |
8dcf39c4 | 947 | { |
49e268e2 | 948 | char c = 0; |
8dcf39c4 | 949 | |
56948cb6 | 950 | if (has_dos_drive_prefix(path)) |
2a1ae649 | 951 | return PATH_INVALID; |
56948cb6 | 952 | |
d2c84dad | 953 | if (!is_valid_path(path)) |
2a1ae649 | 954 | return PATH_INVALID; |
d2c84dad | 955 | |
8dcf39c4 LT |
956 | goto inside; |
957 | for (;;) { | |
958 | if (!c) | |
2a1ae649 | 959 | return PATH_OK; |
56948cb6 | 960 | if (is_dir_sep(c)) { |
8dcf39c4 | 961 | inside: |
10ecfa76 | 962 | if (protect_hfs) { |
49e268e2 | 963 | |
10ecfa76 | 964 | if (is_hfs_dotgit(path)) |
2a1ae649 | 965 | return PATH_INVALID; |
10ecfa76 JK |
966 | if (S_ISLNK(mode)) { |
967 | if (is_hfs_dotgitmodules(path)) | |
2a1ae649 | 968 | return PATH_INVALID; |
10ecfa76 JK |
969 | } |
970 | } | |
971 | if (protect_ntfs) { | |
bccc37fd | 972 | #if defined GIT_WINDOWS_NATIVE || defined __CYGWIN__ |
49e268e2 | 973 | if (c == '\\') |
2a1ae649 | 974 | return PATH_INVALID; |
49e268e2 | 975 | #endif |
10ecfa76 | 976 | if (is_ntfs_dotgit(path)) |
2a1ae649 | 977 | return PATH_INVALID; |
10ecfa76 JK |
978 | if (S_ISLNK(mode)) { |
979 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 980 | return PATH_INVALID; |
10ecfa76 JK |
981 | } |
982 | } | |
983 | ||
8dcf39c4 | 984 | c = *path++; |
10ecfa76 | 985 | if ((c == '.' && !verify_dotfile(path, mode)) || |
6e773527 | 986 | is_dir_sep(c)) |
2a1ae649 | 987 | return PATH_INVALID; |
6e773527 DS |
988 | /* |
989 | * allow terminating directory separators for | |
990 | * sparse directory entries. | |
991 | */ | |
992 | if (c == '\0') | |
2a1ae649 RS |
993 | return S_ISDIR(mode) ? PATH_DIR_WITH_SEP : |
994 | PATH_INVALID; | |
288a74bc JS |
995 | } else if (c == '\\' && protect_ntfs) { |
996 | if (is_ntfs_dotgit(path)) | |
2a1ae649 | 997 | return PATH_INVALID; |
288a74bc JS |
998 | if (S_ISLNK(mode)) { |
999 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 1000 | return PATH_INVALID; |
288a74bc | 1001 | } |
8dcf39c4 | 1002 | } |
288a74bc | 1003 | |
8dcf39c4 LT |
1004 | c = *path++; |
1005 | } | |
1006 | } | |
1007 | ||
12676608 LT |
1008 | /* |
1009 | * Do we have another file that has the beginning components being a | |
1010 | * proper superset of the name we're trying to add? | |
0f1e4f04 | 1011 | */ |
4aab5b46 JH |
1012 | static int has_file_name(struct index_state *istate, |
1013 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
0f1e4f04 | 1014 | { |
12676608 LT |
1015 | int retval = 0; |
1016 | int len = ce_namelen(ce); | |
b155725d | 1017 | int stage = ce_stage(ce); |
12676608 | 1018 | const char *name = ce->name; |
0f1e4f04 | 1019 | |
4aab5b46 JH |
1020 | while (pos < istate->cache_nr) { |
1021 | struct cache_entry *p = istate->cache[pos++]; | |
0f1e4f04 | 1022 | |
12676608 | 1023 | if (len >= ce_namelen(p)) |
0f1e4f04 | 1024 | break; |
12676608 LT |
1025 | if (memcmp(name, p->name, len)) |
1026 | break; | |
b155725d JH |
1027 | if (ce_stage(p) != stage) |
1028 | continue; | |
12676608 LT |
1029 | if (p->name[len] != '/') |
1030 | continue; | |
7a51ed66 | 1031 | if (p->ce_flags & CE_REMOVE) |
21cd8d00 | 1032 | continue; |
12676608 LT |
1033 | retval = -1; |
1034 | if (!ok_to_replace) | |
1035 | break; | |
4aab5b46 | 1036 | remove_index_entry_at(istate, --pos); |
0f1e4f04 | 1037 | } |
12676608 LT |
1038 | return retval; |
1039 | } | |
0f1e4f04 | 1040 | |
a6db3fbb JH |
1041 | |
1042 | /* | |
1043 | * Like strcmp(), but also return the offset of the first change. | |
1044 | * If strings are equal, return the length. | |
1045 | */ | |
1046 | int strcmp_offset(const char *s1, const char *s2, size_t *first_change) | |
1047 | { | |
1048 | size_t k; | |
1049 | ||
1050 | if (!first_change) | |
1051 | return strcmp(s1, s2); | |
1052 | ||
1053 | for (k = 0; s1[k] == s2[k]; k++) | |
1054 | if (s1[k] == '\0') | |
1055 | break; | |
1056 | ||
1057 | *first_change = k; | |
1058 | return (unsigned char)s1[k] - (unsigned char)s2[k]; | |
1059 | } | |
1060 | ||
12676608 LT |
1061 | /* |
1062 | * Do we have another file with a pathname that is a proper | |
1063 | * subset of the name we're trying to add? | |
06b6d81b JH |
1064 | * |
1065 | * That is, is there another file in the index with a path | |
1066 | * that matches a sub-directory in the given entry? | |
12676608 | 1067 | */ |
4aab5b46 JH |
1068 | static int has_dir_name(struct index_state *istate, |
1069 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
12676608 LT |
1070 | { |
1071 | int retval = 0; | |
b155725d | 1072 | int stage = ce_stage(ce); |
12676608 LT |
1073 | const char *name = ce->name; |
1074 | const char *slash = name + ce_namelen(ce); | |
06b6d81b JH |
1075 | size_t len_eq_last; |
1076 | int cmp_last = 0; | |
1077 | ||
1078 | /* | |
1079 | * We are frequently called during an iteration on a sorted | |
1080 | * list of pathnames and while building a new index. Therefore, | |
1081 | * there is a high probability that this entry will eventually | |
1082 | * be appended to the index, rather than inserted in the middle. | |
1083 | * If we can confirm that, we can avoid binary searches on the | |
1084 | * components of the pathname. | |
1085 | * | |
1086 | * Compare the entry's full path with the last path in the index. | |
1087 | */ | |
1088 | if (istate->cache_nr > 0) { | |
1089 | cmp_last = strcmp_offset(name, | |
1090 | istate->cache[istate->cache_nr - 1]->name, | |
1091 | &len_eq_last); | |
1092 | if (cmp_last > 0) { | |
1093 | if (len_eq_last == 0) { | |
1094 | /* | |
1095 | * The entry sorts AFTER the last one in the | |
1096 | * index and their paths have no common prefix, | |
1097 | * so there cannot be a F/D conflict. | |
1098 | */ | |
1099 | return retval; | |
1100 | } else { | |
1101 | /* | |
1102 | * The entry sorts AFTER the last one in the | |
1103 | * index, but has a common prefix. Fall through | |
1104 | * to the loop below to disect the entry's path | |
1105 | * and see where the difference is. | |
1106 | */ | |
1107 | } | |
1108 | } else if (cmp_last == 0) { | |
1109 | /* | |
1110 | * The entry exactly matches the last one in the | |
1111 | * index, but because of multiple stage and CE_REMOVE | |
1112 | * items, we fall through and let the regular search | |
1113 | * code handle it. | |
1114 | */ | |
1115 | } | |
1116 | } | |
0f1e4f04 | 1117 | |
12676608 | 1118 | for (;;) { |
b986df5c | 1119 | size_t len; |
0f1e4f04 | 1120 | |
12676608 LT |
1121 | for (;;) { |
1122 | if (*--slash == '/') | |
1123 | break; | |
1124 | if (slash <= ce->name) | |
1125 | return retval; | |
1126 | } | |
1127 | len = slash - name; | |
0f1e4f04 | 1128 | |
b986df5c JH |
1129 | if (cmp_last > 0) { |
1130 | /* | |
1131 | * (len + 1) is a directory boundary (including | |
1132 | * the trailing slash). And since the loop is | |
1133 | * decrementing "slash", the first iteration is | |
1134 | * the longest directory prefix; subsequent | |
1135 | * iterations consider parent directories. | |
1136 | */ | |
1137 | ||
1138 | if (len + 1 <= len_eq_last) { | |
1139 | /* | |
1140 | * The directory prefix (including the trailing | |
1141 | * slash) also appears as a prefix in the last | |
1142 | * entry, so the remainder cannot collide (because | |
1143 | * strcmp said the whole path was greater). | |
1144 | * | |
1145 | * EQ: last: xxx/A | |
1146 | * this: xxx/B | |
1147 | * | |
1148 | * LT: last: xxx/file_A | |
1149 | * this: xxx/file_B | |
1150 | */ | |
1151 | return retval; | |
1152 | } | |
1153 | ||
1154 | if (len > len_eq_last) { | |
1155 | /* | |
1156 | * This part of the directory prefix (excluding | |
1157 | * the trailing slash) is longer than the known | |
1158 | * equal portions, so this sub-directory cannot | |
1159 | * collide with a file. | |
1160 | * | |
1161 | * GT: last: xxxA | |
1162 | * this: xxxB/file | |
1163 | */ | |
1164 | return retval; | |
1165 | } | |
1166 | ||
b986df5c JH |
1167 | /* |
1168 | * This is a possible collision. Fall through and | |
1169 | * let the regular search code handle it. | |
1170 | * | |
1171 | * last: xxx | |
1172 | * this: xxx/file | |
1173 | */ | |
1174 | } | |
1175 | ||
20ec2d03 | 1176 | pos = index_name_stage_pos(istate, name, len, stage, EXPAND_SPARSE); |
12676608 | 1177 | if (pos >= 0) { |
21cd8d00 JH |
1178 | /* |
1179 | * Found one, but not so fast. This could | |
1180 | * be a marker that says "I was here, but | |
1181 | * I am being removed". Such an entry is | |
1182 | * not a part of the resulting tree, and | |
1183 | * it is Ok to have a directory at the same | |
1184 | * path. | |
1185 | */ | |
077c48df | 1186 | if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) { |
21cd8d00 JH |
1187 | retval = -1; |
1188 | if (!ok_to_replace) | |
1189 | break; | |
4aab5b46 | 1190 | remove_index_entry_at(istate, pos); |
21cd8d00 JH |
1191 | continue; |
1192 | } | |
12676608 | 1193 | } |
21cd8d00 JH |
1194 | else |
1195 | pos = -pos-1; | |
12676608 LT |
1196 | |
1197 | /* | |
1198 | * Trivial optimization: if we find an entry that | |
1199 | * already matches the sub-directory, then we know | |
b155725d | 1200 | * we're ok, and we can exit. |
12676608 | 1201 | */ |
4aab5b46 JH |
1202 | while (pos < istate->cache_nr) { |
1203 | struct cache_entry *p = istate->cache[pos]; | |
b155725d JH |
1204 | if ((ce_namelen(p) <= len) || |
1205 | (p->name[len] != '/') || | |
1206 | memcmp(p->name, name, len)) | |
1207 | break; /* not our subdirectory */ | |
077c48df JH |
1208 | if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE)) |
1209 | /* | |
1210 | * p is at the same stage as our entry, and | |
b155725d JH |
1211 | * is a subdirectory of what we are looking |
1212 | * at, so we cannot have conflicts at our | |
1213 | * level or anything shorter. | |
1214 | */ | |
1215 | return retval; | |
1216 | pos++; | |
192268c1 | 1217 | } |
0f1e4f04 | 1218 | } |
12676608 LT |
1219 | return retval; |
1220 | } | |
1221 | ||
1222 | /* We may be in a situation where we already have path/file and path | |
1223 | * is being added, or we already have path and path/file is being | |
1224 | * added. Either one would result in a nonsense tree that has path | |
1225 | * twice when git-write-tree tries to write it out. Prevent it. | |
a6080a0a | 1226 | * |
12676608 LT |
1227 | * If ok-to-replace is specified, we remove the conflicting entries |
1228 | * from the cache so the caller should recompute the insert position. | |
1229 | * When this happens, we return non-zero. | |
1230 | */ | |
4aab5b46 JH |
1231 | static int check_file_directory_conflict(struct index_state *istate, |
1232 | const struct cache_entry *ce, | |
1233 | int pos, int ok_to_replace) | |
12676608 | 1234 | { |
21cd8d00 JH |
1235 | int retval; |
1236 | ||
1237 | /* | |
1238 | * When ce is an "I am going away" entry, we allow it to be added | |
1239 | */ | |
7a51ed66 | 1240 | if (ce->ce_flags & CE_REMOVE) |
21cd8d00 JH |
1241 | return 0; |
1242 | ||
12676608 LT |
1243 | /* |
1244 | * We check if the path is a sub-path of a subsequent pathname | |
1245 | * first, since removing those will not change the position | |
21cd8d00 | 1246 | * in the array. |
12676608 | 1247 | */ |
4aab5b46 | 1248 | retval = has_file_name(istate, ce, pos, ok_to_replace); |
21cd8d00 | 1249 | |
12676608 LT |
1250 | /* |
1251 | * Then check if the path might have a clashing sub-directory | |
1252 | * before it. | |
1253 | */ | |
4aab5b46 | 1254 | return retval + has_dir_name(istate, ce, pos, ok_to_replace); |
0f1e4f04 JH |
1255 | } |
1256 | ||
af3785dc | 1257 | static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option) |
197ee8c9 LT |
1258 | { |
1259 | int pos; | |
192268c1 JH |
1260 | int ok_to_add = option & ADD_CACHE_OK_TO_ADD; |
1261 | int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE; | |
b155725d | 1262 | int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK; |
39425819 | 1263 | int new_only = option & ADD_CACHE_NEW_ONLY; |
5f73076c | 1264 | |
e5494631 JH |
1265 | /* |
1266 | * If this entry's path sorts after the last entry in the index, | |
1267 | * we can avoid searching for it. | |
1268 | */ | |
1269 | if (istate->cache_nr > 0 && | |
1270 | strcmp(ce->name, istate->cache[istate->cache_nr - 1]->name) > 0) | |
c097b95a | 1271 | pos = index_pos_to_insert_pos(istate->cache_nr); |
e5494631 | 1272 | else |
20ec2d03 | 1273 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce), EXPAND_SPARSE); |
197ee8c9 | 1274 | |
c35e9f5e VD |
1275 | /* |
1276 | * Cache tree path should be invalidated only after index_name_stage_pos, | |
1277 | * in case it expands a sparse index. | |
1278 | */ | |
1279 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) | |
1280 | cache_tree_invalidate_path(istate, ce->name); | |
1281 | ||
3e09cdfd | 1282 | /* existing match? Just replace it. */ |
76e7f4ec | 1283 | if (pos >= 0) { |
39425819 JH |
1284 | if (!new_only) |
1285 | replace_index_entry(istate, pos, ce); | |
197ee8c9 LT |
1286 | return 0; |
1287 | } | |
76e7f4ec | 1288 | pos = -pos-1; |
197ee8c9 | 1289 | |
ffcc9ba7 NTND |
1290 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1291 | untracked_cache_add_to_index(istate, ce->name); | |
e931371a | 1292 | |
7b937ca3 LT |
1293 | /* |
1294 | * Inserting a merged entry ("stage 0") into the index | |
1295 | * will always replace all non-merged entries.. | |
1296 | */ | |
4aab5b46 JH |
1297 | if (pos < istate->cache_nr && ce_stage(ce) == 0) { |
1298 | while (ce_same_name(istate->cache[pos], ce)) { | |
7b937ca3 | 1299 | ok_to_add = 1; |
4aab5b46 | 1300 | if (!remove_index_entry_at(istate, pos)) |
7b937ca3 LT |
1301 | break; |
1302 | } | |
1303 | } | |
1304 | ||
121481ab LT |
1305 | if (!ok_to_add) |
1306 | return -1; | |
c8ad9d04 | 1307 | if (verify_path_internal(ce->name, ce->ce_mode) == PATH_INVALID) |
9d0a9e90 | 1308 | return error(_("invalid path '%s'"), ce->name); |
121481ab | 1309 | |
3e09cdfd | 1310 | if (!skip_df_check && |
4aab5b46 | 1311 | check_file_directory_conflict(istate, ce, pos, ok_to_replace)) { |
192268c1 | 1312 | if (!ok_to_replace) |
9d0a9e90 | 1313 | return error(_("'%s' appears as both a file and as a directory"), |
4aab5b46 | 1314 | ce->name); |
20ec2d03 | 1315 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce), EXPAND_SPARSE); |
192268c1 JH |
1316 | pos = -pos-1; |
1317 | } | |
af3785dc JH |
1318 | return pos + 1; |
1319 | } | |
1320 | ||
1321 | int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option) | |
1322 | { | |
1323 | int pos; | |
1324 | ||
1325 | if (option & ADD_CACHE_JUST_APPEND) | |
1326 | pos = istate->cache_nr; | |
1327 | else { | |
1328 | int ret; | |
1329 | ret = add_index_entry_with_check(istate, ce, option); | |
1330 | if (ret <= 0) | |
1331 | return ret; | |
1332 | pos = ret - 1; | |
1333 | } | |
0f1e4f04 | 1334 | |
197ee8c9 | 1335 | /* Make sure the array is big enough .. */ |
999f5660 | 1336 | ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc); |
197ee8c9 LT |
1337 | |
1338 | /* Add it in.. */ | |
4aab5b46 | 1339 | istate->cache_nr++; |
af3785dc | 1340 | if (istate->cache_nr > pos + 1) |
f919ffeb SG |
1341 | MOVE_ARRAY(istate->cache + pos + 1, istate->cache + pos, |
1342 | istate->cache_nr - pos - 1); | |
cf558704 | 1343 | set_index_entry(istate, pos, ce); |
e636a7b4 | 1344 | istate->cache_changed |= CE_ENTRY_ADDED; |
197ee8c9 LT |
1345 | return 0; |
1346 | } | |
1347 | ||
405e5b2f LT |
1348 | /* |
1349 | * "refresh" does not calculate a new sha1 file or bring the | |
1350 | * cache up-to-date for mode/content changes. But what it | |
1351 | * _does_ do is to "re-match" the stat information of a file | |
1352 | * with the cache, so that you can refresh the cache for a | |
1353 | * file that hasn't been changed but where the stat entry is | |
1354 | * out of date. | |
1355 | * | |
1356 | * For example, you'd want to do this after doing a "git-read-tree", | |
1357 | * to link up the stat cache details with the proper files. | |
1358 | */ | |
4aab5b46 | 1359 | static struct cache_entry *refresh_cache_ent(struct index_state *istate, |
4bd5b7da | 1360 | struct cache_entry *ce, |
d05e6970 | 1361 | unsigned int options, int *err, |
a98e0f2d | 1362 | int *changed_ret, |
15268d12 JH |
1363 | int *t2_did_lstat, |
1364 | int *t2_did_scan) | |
405e5b2f LT |
1365 | { |
1366 | struct stat st; | |
1367 | struct cache_entry *updated; | |
a849735b | 1368 | int changed; |
25762726 | 1369 | int refresh = options & CE_MATCH_REFRESH; |
4bd5b7da | 1370 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 1371 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
2e2e7ec1 | 1372 | int ignore_missing = options & CE_MATCH_IGNORE_MISSING; |
883e248b | 1373 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
405e5b2f | 1374 | |
25762726 | 1375 | if (!refresh || ce_uptodate(ce)) |
eadb5831 JH |
1376 | return ce; |
1377 | ||
883e248b BP |
1378 | if (!ignore_fsmonitor) |
1379 | refresh_fsmonitor(istate); | |
aa9349d4 | 1380 | /* |
56cac48c NTND |
1381 | * CE_VALID or CE_SKIP_WORKTREE means the user promised us |
1382 | * that the change to the work tree does not matter and told | |
1383 | * us not to worry. | |
aa9349d4 | 1384 | */ |
56cac48c NTND |
1385 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) { |
1386 | ce_mark_uptodate(ce); | |
1387 | return ce; | |
1388 | } | |
aa9349d4 MSO |
1389 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) { |
1390 | ce_mark_uptodate(ce); | |
1391 | return ce; | |
1392 | } | |
883e248b BP |
1393 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) { |
1394 | ce_mark_uptodate(ce); | |
1395 | return ce; | |
1396 | } | |
aa9349d4 | 1397 | |
ccad42d4 RS |
1398 | if (has_symlink_leading_path(ce->name, ce_namelen(ce))) { |
1399 | if (ignore_missing) | |
1400 | return ce; | |
1401 | if (err) | |
1402 | *err = ENOENT; | |
1403 | return NULL; | |
1404 | } | |
1405 | ||
a98e0f2d JH |
1406 | if (t2_did_lstat) |
1407 | *t2_did_lstat = 1; | |
8fd2cb40 | 1408 | if (lstat(ce->name, &st) < 0) { |
2e2e7ec1 BK |
1409 | if (ignore_missing && errno == ENOENT) |
1410 | return ce; | |
ec0cc704 JH |
1411 | if (err) |
1412 | *err = errno; | |
8fd2cb40 JS |
1413 | return NULL; |
1414 | } | |
405e5b2f | 1415 | |
4bd5b7da | 1416 | changed = ie_match_stat(istate, ce, &st, options); |
d05e6970 JK |
1417 | if (changed_ret) |
1418 | *changed_ret = changed; | |
405e5b2f | 1419 | if (!changed) { |
4bd5b7da JH |
1420 | /* |
1421 | * The path is unchanged. If we were told to ignore | |
1422 | * valid bit, then we did the actual stat check and | |
1423 | * found that the entry is unmodified. If the entry | |
1424 | * is not marked VALID, this is the place to mark it | |
1425 | * valid again, under "assume unchanged" mode. | |
1426 | */ | |
1427 | if (ignore_valid && assume_unchanged && | |
7a51ed66 | 1428 | !(ce->ce_flags & CE_VALID)) |
405e5b2f | 1429 | ; /* mark this one VALID again */ |
eadb5831 JH |
1430 | else { |
1431 | /* | |
1432 | * We do not mark the index itself "modified" | |
1433 | * because CE_UPTODATE flag is in-core only; | |
1434 | * we are not going to write this change out. | |
1435 | */ | |
883e248b | 1436 | if (!S_ISGITLINK(ce->ce_mode)) { |
125fd984 | 1437 | ce_mark_uptodate(ce); |
b5a81697 | 1438 | mark_fsmonitor_valid(istate, ce); |
883e248b | 1439 | } |
8fd2cb40 | 1440 | return ce; |
eadb5831 | 1441 | } |
405e5b2f LT |
1442 | } |
1443 | ||
15268d12 JH |
1444 | if (t2_did_scan) |
1445 | *t2_did_scan = 1; | |
4bd5b7da | 1446 | if (ie_modified(istate, ce, &st, options)) { |
ec0cc704 JH |
1447 | if (err) |
1448 | *err = EINVAL; | |
8fd2cb40 JS |
1449 | return NULL; |
1450 | } | |
405e5b2f | 1451 | |
a849735b | 1452 | updated = make_empty_cache_entry(istate, ce_namelen(ce)); |
0e267b7a BP |
1453 | copy_cache_entry(updated, ce); |
1454 | memcpy(updated->name, ce->name, ce->ce_namelen + 1); | |
d4c0a3ac | 1455 | fill_stat_cache_info(istate, updated, &st); |
4bd5b7da JH |
1456 | /* |
1457 | * If ignore_valid is not set, we should leave CE_VALID bit | |
1458 | * alone. Otherwise, paths marked with --no-assume-unchanged | |
1459 | * (i.e. things to be edited) will reacquire CE_VALID bit | |
1460 | * automatically, which is not really what we want. | |
405e5b2f | 1461 | */ |
4bd5b7da | 1462 | if (!ignore_valid && assume_unchanged && |
7a51ed66 LT |
1463 | !(ce->ce_flags & CE_VALID)) |
1464 | updated->ce_flags &= ~CE_VALID; | |
405e5b2f | 1465 | |
e636a7b4 | 1466 | /* istate->cache_changed is updated in the caller */ |
405e5b2f LT |
1467 | return updated; |
1468 | } | |
1469 | ||
3deffc52 | 1470 | static void show_file(const char * fmt, const char * name, int in_porcelain, |
046613c5 | 1471 | int * first, const char *header_msg) |
3deffc52 MM |
1472 | { |
1473 | if (in_porcelain && *first && header_msg) { | |
1474 | printf("%s\n", header_msg); | |
cd2b8ae9 | 1475 | *first = 0; |
3deffc52 MM |
1476 | } |
1477 | printf(fmt, name); | |
1478 | } | |
1479 | ||
22184497 TG |
1480 | int repo_refresh_and_write_index(struct repository *repo, |
1481 | unsigned int refresh_flags, | |
1482 | unsigned int write_flags, | |
1483 | int gentle, | |
1484 | const struct pathspec *pathspec, | |
1485 | char *seen, const char *header_msg) | |
1486 | { | |
1487 | struct lock_file lock_file = LOCK_INIT; | |
1488 | int fd, ret = 0; | |
1489 | ||
1490 | fd = repo_hold_locked_index(repo, &lock_file, 0); | |
1491 | if (!gentle && fd < 0) | |
1492 | return -1; | |
1493 | if (refresh_index(repo->index, refresh_flags, pathspec, seen, header_msg)) | |
1494 | ret = 1; | |
1495 | if (0 <= fd && write_locked_index(repo->index, &lock_file, COMMIT_LOCK | write_flags)) | |
1496 | ret = -1; | |
1497 | return ret; | |
1498 | } | |
1499 | ||
1500 | ||
9b2d6149 NTND |
1501 | int refresh_index(struct index_state *istate, unsigned int flags, |
1502 | const struct pathspec *pathspec, | |
046613c5 | 1503 | char *seen, const char *header_msg) |
405e5b2f LT |
1504 | { |
1505 | int i; | |
1506 | int has_errors = 0; | |
1507 | int really = (flags & REFRESH_REALLY) != 0; | |
1508 | int allow_unmerged = (flags & REFRESH_UNMERGED) != 0; | |
1509 | int quiet = (flags & REFRESH_QUIET) != 0; | |
1510 | int not_new = (flags & REFRESH_IGNORE_MISSING) != 0; | |
5fdeacb0 | 1511 | int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0; |
b243012c | 1512 | int ignore_skip_worktree = (flags & REFRESH_IGNORE_SKIP_WORKTREE) != 0; |
3deffc52 MM |
1513 | int first = 1; |
1514 | int in_porcelain = (flags & REFRESH_IN_PORCELAIN); | |
25762726 BK |
1515 | unsigned int options = (CE_MATCH_REFRESH | |
1516 | (really ? CE_MATCH_IGNORE_VALID : 0) | | |
2e2e7ec1 | 1517 | (not_new ? CE_MATCH_IGNORE_MISSING : 0)); |
4bd4e730 | 1518 | const char *modified_fmt; |
73b7eae6 JK |
1519 | const char *deleted_fmt; |
1520 | const char *typechange_fmt; | |
1521 | const char *added_fmt; | |
4bd4e730 | 1522 | const char *unmerged_fmt; |
ae9af122 | 1523 | struct progress *progress = NULL; |
a98e0f2d | 1524 | int t2_sum_lstat = 0; |
15268d12 | 1525 | int t2_sum_scan = 0; |
ae9af122 NTND |
1526 | |
1527 | if (flags & REFRESH_PROGRESS && isatty(2)) | |
1528 | progress = start_delayed_progress(_("Refresh index"), | |
1529 | istate->cache_nr); | |
405e5b2f | 1530 | |
c46c406a | 1531 | trace_performance_enter(); |
a71806a7 NTND |
1532 | modified_fmt = in_porcelain ? "M\t%s\n" : "%s: needs update\n"; |
1533 | deleted_fmt = in_porcelain ? "D\t%s\n" : "%s: needs update\n"; | |
1534 | typechange_fmt = in_porcelain ? "T\t%s\n" : "%s: needs update\n"; | |
1535 | added_fmt = in_porcelain ? "A\t%s\n" : "%s: needs update\n"; | |
1536 | unmerged_fmt = in_porcelain ? "U\t%s\n" : "%s: needs merge\n"; | |
99ce720c BP |
1537 | /* |
1538 | * Use the multi-threaded preload_index() to refresh most of the | |
1539 | * cache entries quickly then in the single threaded loop below, | |
1540 | * we only have to do the special cases that are left. | |
1541 | */ | |
1542 | preload_index(istate, pathspec, 0); | |
a98e0f2d | 1543 | trace2_region_enter("index", "refresh", NULL); |
d76723ee | 1544 | |
4aab5b46 | 1545 | for (i = 0; i < istate->cache_nr; i++) { |
285c2e25 | 1546 | struct cache_entry *ce, *new_entry; |
ec0cc704 | 1547 | int cache_errno = 0; |
73b7eae6 | 1548 | int changed = 0; |
3d1f148c | 1549 | int filtered = 0; |
a98e0f2d | 1550 | int t2_did_lstat = 0; |
15268d12 | 1551 | int t2_did_scan = 0; |
ec0cc704 | 1552 | |
4aab5b46 | 1553 | ce = istate->cache[i]; |
5fdeacb0 JS |
1554 | if (ignore_submodules && S_ISGITLINK(ce->ce_mode)) |
1555 | continue; | |
b243012c MT |
1556 | if (ignore_skip_worktree && ce_skip_worktree(ce)) |
1557 | continue; | |
5fdeacb0 | 1558 | |
d76723ee DS |
1559 | /* |
1560 | * If this entry is a sparse directory, then there isn't | |
1561 | * any stat() information to update. Ignore the entry. | |
1562 | */ | |
1563 | if (S_ISSPARSEDIR(ce->ce_mode)) | |
1564 | continue; | |
1565 | ||
d7b665c3 | 1566 | if (pathspec && !ce_path_match(istate, ce, pathspec, seen)) |
3d1f148c JH |
1567 | filtered = 1; |
1568 | ||
405e5b2f | 1569 | if (ce_stage(ce)) { |
4aab5b46 JH |
1570 | while ((i < istate->cache_nr) && |
1571 | ! strcmp(istate->cache[i]->name, ce->name)) | |
405e5b2f LT |
1572 | i++; |
1573 | i--; | |
1574 | if (allow_unmerged) | |
1575 | continue; | |
3d1f148c JH |
1576 | if (!filtered) |
1577 | show_file(unmerged_fmt, ce->name, in_porcelain, | |
1578 | &first, header_msg); | |
405e5b2f LT |
1579 | has_errors = 1; |
1580 | continue; | |
1581 | } | |
1582 | ||
3d1f148c | 1583 | if (filtered) |
d616813d AJ |
1584 | continue; |
1585 | ||
a98e0f2d JH |
1586 | new_entry = refresh_cache_ent(istate, ce, options, |
1587 | &cache_errno, &changed, | |
15268d12 | 1588 | &t2_did_lstat, &t2_did_scan); |
a98e0f2d | 1589 | t2_sum_lstat += t2_did_lstat; |
15268d12 | 1590 | t2_sum_scan += t2_did_scan; |
285c2e25 | 1591 | if (new_entry == ce) |
405e5b2f | 1592 | continue; |
b7b793d1 | 1593 | display_progress(progress, i); |
285c2e25 | 1594 | if (!new_entry) { |
73b7eae6 JK |
1595 | const char *fmt; |
1596 | ||
8fd2cb40 | 1597 | if (really && cache_errno == EINVAL) { |
405e5b2f LT |
1598 | /* If we are doing --really-refresh that |
1599 | * means the index is not valid anymore. | |
1600 | */ | |
7a51ed66 | 1601 | ce->ce_flags &= ~CE_VALID; |
078a58e8 | 1602 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 1603 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 1604 | istate->cache_changed |= CE_ENTRY_CHANGED; |
405e5b2f LT |
1605 | } |
1606 | if (quiet) | |
1607 | continue; | |
73b7eae6 JK |
1608 | |
1609 | if (cache_errno == ENOENT) | |
1610 | fmt = deleted_fmt; | |
895ff3b2 | 1611 | else if (ce_intent_to_add(ce)) |
73b7eae6 JK |
1612 | fmt = added_fmt; /* must be before other checks */ |
1613 | else if (changed & TYPE_CHANGED) | |
1614 | fmt = typechange_fmt; | |
1615 | else | |
1616 | fmt = modified_fmt; | |
1617 | show_file(fmt, | |
1618 | ce->name, in_porcelain, &first, header_msg); | |
405e5b2f LT |
1619 | has_errors = 1; |
1620 | continue; | |
1621 | } | |
cf558704 | 1622 | |
285c2e25 | 1623 | replace_index_entry(istate, i, new_entry); |
405e5b2f | 1624 | } |
a98e0f2d | 1625 | trace2_data_intmax("index", NULL, "refresh/sum_lstat", t2_sum_lstat); |
15268d12 | 1626 | trace2_data_intmax("index", NULL, "refresh/sum_scan", t2_sum_scan); |
a98e0f2d | 1627 | trace2_region_leave("index", "refresh", NULL); |
b7b793d1 ÆAB |
1628 | display_progress(progress, istate->cache_nr); |
1629 | stop_progress(&progress); | |
c46c406a | 1630 | trace_performance_leave("refresh index"); |
405e5b2f LT |
1631 | return has_errors; |
1632 | } | |
1633 | ||
768d7965 JM |
1634 | struct cache_entry *refresh_cache_entry(struct index_state *istate, |
1635 | struct cache_entry *ce, | |
1636 | unsigned int options) | |
ec0cc704 | 1637 | { |
15268d12 | 1638 | return refresh_cache_ent(istate, ce, options, NULL, NULL, NULL, NULL); |
ec0cc704 JH |
1639 | } |
1640 | ||
db3b313c JH |
1641 | |
1642 | /***************************************************************** | |
1643 | * Index File I/O | |
1644 | *****************************************************************/ | |
1645 | ||
9d227781 JH |
1646 | #define INDEX_FORMAT_DEFAULT 3 |
1647 | ||
7211b9e7 | 1648 | static unsigned int get_index_format_default(struct repository *r) |
136347d7 TG |
1649 | { |
1650 | char *envversion = getenv("GIT_INDEX_VERSION"); | |
3c09d684 TG |
1651 | char *endp; |
1652 | unsigned int version = INDEX_FORMAT_DEFAULT; | |
1653 | ||
136347d7 | 1654 | if (!envversion) { |
7211b9e7 DS |
1655 | prepare_repo_settings(r); |
1656 | ||
1657 | if (r->settings.index_version >= 0) | |
1658 | version = r->settings.index_version; | |
3c09d684 TG |
1659 | if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { |
1660 | warning(_("index.version set, but the value is invalid.\n" | |
136347d7 | 1661 | "Using version %i"), INDEX_FORMAT_DEFAULT); |
3c09d684 | 1662 | return INDEX_FORMAT_DEFAULT; |
136347d7 TG |
1663 | } |
1664 | return version; | |
1665 | } | |
3c09d684 TG |
1666 | |
1667 | version = strtoul(envversion, &endp, 10); | |
1668 | if (*endp || | |
1669 | version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { | |
1670 | warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n" | |
1671 | "Using version %i"), INDEX_FORMAT_DEFAULT); | |
1672 | version = INDEX_FORMAT_DEFAULT; | |
1673 | } | |
1674 | return version; | |
136347d7 TG |
1675 | } |
1676 | ||
db3b313c JH |
1677 | /* |
1678 | * dev/ino/uid/gid/size are also just tracked to the low 32 bits | |
1679 | * Again - this is just a (very strong in practice) heuristic that | |
1680 | * the inode hasn't changed. | |
1681 | * | |
1682 | * We save the fields in big-endian order to allow using the | |
1683 | * index file over NFS transparently. | |
1684 | */ | |
1685 | struct ondisk_cache_entry { | |
1686 | struct cache_time ctime; | |
1687 | struct cache_time mtime; | |
7800c1eb TG |
1688 | uint32_t dev; |
1689 | uint32_t ino; | |
1690 | uint32_t mode; | |
1691 | uint32_t uid; | |
1692 | uint32_t gid; | |
1693 | uint32_t size; | |
575fa8a3 | 1694 | /* |
1695 | * unsigned char hash[hashsz]; | |
1696 | * uint16_t flags; | |
1697 | * if (flags & CE_EXTENDED) | |
1698 | * uint16_t flags2; | |
1699 | */ | |
1700 | unsigned char data[GIT_MAX_RAWSZ + 2 * sizeof(uint16_t)]; | |
1701 | char name[FLEX_ARRAY]; | |
db3b313c JH |
1702 | }; |
1703 | ||
6c9cd161 | 1704 | /* These are only used for v3 or lower */ |
ce012deb | 1705 | #define align_padding_size(size, len) ((size + (len) + 8) & ~7) - (size + len) |
575fa8a3 | 1706 | #define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,data) + (len) + 8) & ~7) |
db3b313c | 1707 | #define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len) |
575fa8a3 | 1708 | #define ondisk_data_size(flags, len) (the_hash_algo->rawsz + \ |
1709 | ((flags & CE_EXTENDED) ? 2 : 1) * sizeof(uint16_t) + len) | |
1710 | #define ondisk_data_size_max(len) (ondisk_data_size(CE_EXTENDED, len)) | |
1711 | #define ondisk_ce_size(ce) (ondisk_cache_entry_size(ondisk_data_size((ce)->ce_flags, ce_namelen(ce)))) | |
db3b313c | 1712 | |
a33fc72f JH |
1713 | /* Allow fsck to force verification of the index checksum. */ |
1714 | int verify_index_checksum; | |
1715 | ||
00ec50e5 BP |
1716 | /* Allow fsck to force verification of the cache entry order. */ |
1717 | int verify_ce_order; | |
1718 | ||
371ed0de | 1719 | static int verify_hdr(const struct cache_header *hdr, unsigned long size) |
e83c5163 | 1720 | { |
aab61359 | 1721 | git_hash_ctx c; |
1722 | unsigned char hash[GIT_MAX_RAWSZ]; | |
0136bac9 | 1723 | int hdr_version; |
ee1f0c24 DS |
1724 | unsigned char *start, *end; |
1725 | struct object_id oid; | |
e83c5163 | 1726 | |
ccc4feb5 | 1727 | if (hdr->hdr_signature != htonl(CACHE_SIGNATURE)) |
9d0a9e90 | 1728 | return error(_("bad signature 0x%08x"), hdr->hdr_signature); |
0136bac9 | 1729 | hdr_version = ntohl(hdr->hdr_version); |
b82a7b5b | 1730 | if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version) |
9d0a9e90 | 1731 | return error(_("bad index version %d"), hdr_version); |
a33fc72f JH |
1732 | |
1733 | if (!verify_index_checksum) | |
1734 | return 0; | |
1735 | ||
ee1f0c24 DS |
1736 | end = (unsigned char *)hdr + size; |
1737 | start = end - the_hash_algo->rawsz; | |
1738 | oidread(&oid, start); | |
1739 | if (oideq(&oid, null_oid())) | |
1740 | return 0; | |
1741 | ||
aab61359 | 1742 | the_hash_algo->init_fn(&c); |
1743 | the_hash_algo->update_fn(&c, hdr, size - the_hash_algo->rawsz); | |
1744 | the_hash_algo->final_fn(hash, &c); | |
ee1f0c24 | 1745 | if (!hasheq(hash, start)) |
9d0a9e90 | 1746 | return error(_("bad index file sha1 signature")); |
e83c5163 LT |
1747 | return 0; |
1748 | } | |
1749 | ||
4aab5b46 | 1750 | static int read_index_extension(struct index_state *istate, |
371ed0de | 1751 | const char *ext, const char *data, unsigned long sz) |
bad68ec9 JH |
1752 | { |
1753 | switch (CACHE_EXT(ext)) { | |
1754 | case CACHE_EXT_TREE: | |
4aab5b46 | 1755 | istate->cache_tree = cache_tree_read(data, sz); |
bad68ec9 | 1756 | break; |
cfc5789a JH |
1757 | case CACHE_EXT_RESOLVE_UNDO: |
1758 | istate->resolve_undo = resolve_undo_read(data, sz); | |
1759 | break; | |
5fc2fc8f NTND |
1760 | case CACHE_EXT_LINK: |
1761 | if (read_link_extension(istate, data, sz)) | |
1762 | return -1; | |
1763 | break; | |
f9e6c649 NTND |
1764 | case CACHE_EXT_UNTRACKED: |
1765 | istate->untracked = read_untracked_extension(data, sz); | |
1766 | break; | |
883e248b BP |
1767 | case CACHE_EXT_FSMONITOR: |
1768 | read_fsmonitor_extension(istate, data, sz); | |
1769 | break; | |
3b1d9e04 | 1770 | case CACHE_EXT_ENDOFINDEXENTRIES: |
3255089a | 1771 | case CACHE_EXT_INDEXENTRYOFFSETTABLE: |
3b1d9e04 BP |
1772 | /* already handled in do_read_index() */ |
1773 | break; | |
cd42415f DS |
1774 | case CACHE_EXT_SPARSE_DIRECTORIES: |
1775 | /* no content, only an indicator */ | |
9fadb373 | 1776 | istate->sparse_index = INDEX_COLLAPSED; |
cd42415f | 1777 | break; |
bad68ec9 JH |
1778 | default: |
1779 | if (*ext < 'A' || 'Z' < *ext) | |
9d0a9e90 | 1780 | return error(_("index uses %.4s extension, which we do not understand"), |
bad68ec9 | 1781 | ext); |
9d0a9e90 | 1782 | fprintf_ln(stderr, _("ignoring %.4s extension"), ext); |
bad68ec9 JH |
1783 | break; |
1784 | } | |
1785 | return 0; | |
1786 | } | |
1787 | ||
4a6ed30f VD |
1788 | /* |
1789 | * Parses the contents of the cache entry contained within the 'ondisk' buffer | |
1790 | * into a new incore 'cache_entry'. | |
1791 | * | |
1792 | * Note that 'char *ondisk' may not be aligned to a 4-byte address interval in | |
1793 | * index v4, so we cannot cast it to 'struct ondisk_cache_entry *' and access | |
1794 | * its members. Instead, we use the byte offsets of members within the struct to | |
1795 | * identify where 'get_be16()', 'get_be32()', and 'oidread()' (which can all | |
1796 | * read from an unaligned memory buffer) should read from the 'ondisk' buffer | |
1797 | * into the corresponding incore 'cache_entry' members. | |
1798 | */ | |
77ff1127 BP |
1799 | static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool, |
1800 | unsigned int version, | |
4a6ed30f | 1801 | const char *ondisk, |
6c9cd161 | 1802 | unsigned long *ent_size, |
252d079c | 1803 | const struct cache_entry *previous_ce) |
7a51ed66 | 1804 | { |
debed2a6 | 1805 | struct cache_entry *ce; |
7fec10b7 | 1806 | size_t len; |
06aaaa0b | 1807 | const char *name; |
575fa8a3 | 1808 | const unsigned hashsz = the_hash_algo->rawsz; |
4a6ed30f | 1809 | const char *flagsp = ondisk + offsetof(struct ondisk_cache_entry, data) + hashsz; |
debed2a6 | 1810 | unsigned int flags; |
f5c4a9af | 1811 | size_t copy_len = 0; |
252d079c NTND |
1812 | /* |
1813 | * Adjacent cache entries tend to share the leading paths, so it makes | |
1814 | * sense to only store the differences in later entries. In the v4 | |
1815 | * on-disk format of the index, each on-disk cache entry stores the | |
1816 | * number of bytes to be stripped from the end of the previous name, | |
1817 | * and the bytes to append to the result, to come up with its name. | |
1818 | */ | |
77ff1127 | 1819 | int expand_name_field = version == 4; |
7fec10b7 | 1820 | |
7a51ed66 | 1821 | /* On-disk flags are just 16 bits */ |
575fa8a3 | 1822 | flags = get_be16(flagsp); |
debed2a6 | 1823 | len = flags & CE_NAMEMASK; |
7fec10b7 | 1824 | |
debed2a6 | 1825 | if (flags & CE_EXTENDED) { |
06aaaa0b | 1826 | int extended_flags; |
4a6ed30f | 1827 | extended_flags = get_be16(flagsp + sizeof(uint16_t)) << 16; |
06aaaa0b NTND |
1828 | /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */ |
1829 | if (extended_flags & ~CE_EXTENDED_FLAGS) | |
9d0a9e90 | 1830 | die(_("unknown index entry format 0x%08x"), extended_flags); |
debed2a6 | 1831 | flags |= extended_flags; |
4a6ed30f | 1832 | name = (const char *)(flagsp + 2 * sizeof(uint16_t)); |
06aaaa0b NTND |
1833 | } |
1834 | else | |
4a6ed30f | 1835 | name = (const char *)(flagsp + sizeof(uint16_t)); |
06aaaa0b | 1836 | |
252d079c NTND |
1837 | if (expand_name_field) { |
1838 | const unsigned char *cp = (const unsigned char *)name; | |
1839 | size_t strip_len, previous_len; | |
6c9cd161 | 1840 | |
15beaaa3 | 1841 | /* If we're at the beginning of a block, ignore the previous name */ |
252d079c | 1842 | strip_len = decode_varint(&cp); |
77ff1127 BP |
1843 | if (previous_ce) { |
1844 | previous_len = previous_ce->ce_namelen; | |
1845 | if (previous_len < strip_len) | |
252d079c | 1846 | die(_("malformed name field in the index, near path '%s'"), |
77ff1127 BP |
1847 | previous_ce->name); |
1848 | copy_len = previous_len - strip_len; | |
252d079c | 1849 | } |
252d079c NTND |
1850 | name = (const char *)cp; |
1851 | } | |
6c9cd161 | 1852 | |
252d079c NTND |
1853 | if (len == CE_NAMEMASK) { |
1854 | len = strlen(name); | |
1855 | if (expand_name_field) | |
1856 | len += copy_len; | |
1857 | } | |
6c9cd161 | 1858 | |
77ff1127 | 1859 | ce = mem_pool__ce_alloc(ce_mem_pool, len); |
252d079c | 1860 | |
4a6ed30f VD |
1861 | /* |
1862 | * NEEDSWORK: using 'offsetof()' is cumbersome and should be replaced | |
1863 | * with something more akin to 'load_bitmap_entries_v1()'s use of | |
1864 | * 'read_be16'/'read_be32'. For consistency with the corresponding | |
1865 | * ondisk entry write function ('copy_cache_entry_to_ondisk()'), this | |
1866 | * should be done at the same time as removing references to | |
1867 | * 'ondisk_cache_entry' there. | |
1868 | */ | |
1869 | ce->ce_stat_data.sd_ctime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime) | |
1870 | + offsetof(struct cache_time, sec)); | |
1871 | ce->ce_stat_data.sd_mtime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime) | |
1872 | + offsetof(struct cache_time, sec)); | |
1873 | ce->ce_stat_data.sd_ctime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime) | |
1874 | + offsetof(struct cache_time, nsec)); | |
1875 | ce->ce_stat_data.sd_mtime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime) | |
1876 | + offsetof(struct cache_time, nsec)); | |
1877 | ce->ce_stat_data.sd_dev = get_be32(ondisk + offsetof(struct ondisk_cache_entry, dev)); | |
1878 | ce->ce_stat_data.sd_ino = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ino)); | |
1879 | ce->ce_mode = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mode)); | |
1880 | ce->ce_stat_data.sd_uid = get_be32(ondisk + offsetof(struct ondisk_cache_entry, uid)); | |
1881 | ce->ce_stat_data.sd_gid = get_be32(ondisk + offsetof(struct ondisk_cache_entry, gid)); | |
1882 | ce->ce_stat_data.sd_size = get_be32(ondisk + offsetof(struct ondisk_cache_entry, size)); | |
252d079c NTND |
1883 | ce->ce_flags = flags & ~CE_NAMEMASK; |
1884 | ce->ce_namelen = len; | |
1885 | ce->index = 0; | |
4a6ed30f | 1886 | oidread(&ce->oid, (const unsigned char *)ondisk + offsetof(struct ondisk_cache_entry, data)); |
6c9cd161 | 1887 | |
252d079c NTND |
1888 | if (expand_name_field) { |
1889 | if (copy_len) | |
1890 | memcpy(ce->name, previous_ce->name, copy_len); | |
1891 | memcpy(ce->name + copy_len, name, len + 1 - copy_len); | |
1892 | *ent_size = (name - ((char *)ondisk)) + len + 1 - copy_len; | |
1893 | } else { | |
1894 | memcpy(ce->name, name, len + 1); | |
1895 | *ent_size = ondisk_ce_size(ce); | |
6c9cd161 | 1896 | } |
debed2a6 | 1897 | return ce; |
cf558704 LT |
1898 | } |
1899 | ||
03f15a79 | 1900 | static void check_ce_order(struct index_state *istate) |
15999d0b | 1901 | { |
03f15a79 TG |
1902 | unsigned int i; |
1903 | ||
00ec50e5 BP |
1904 | if (!verify_ce_order) |
1905 | return; | |
1906 | ||
03f15a79 TG |
1907 | for (i = 1; i < istate->cache_nr; i++) { |
1908 | struct cache_entry *ce = istate->cache[i - 1]; | |
1909 | struct cache_entry *next_ce = istate->cache[i]; | |
1910 | int name_compare = strcmp(ce->name, next_ce->name); | |
1911 | ||
1912 | if (0 < name_compare) | |
9d0a9e90 | 1913 | die(_("unordered stage entries in index")); |
03f15a79 TG |
1914 | if (!name_compare) { |
1915 | if (!ce_stage(ce)) | |
9d0a9e90 | 1916 | die(_("multiple stage entries for merged file '%s'"), |
03f15a79 TG |
1917 | ce->name); |
1918 | if (ce_stage(ce) > ce_stage(next_ce)) | |
9d0a9e90 | 1919 | die(_("unordered stage entries for '%s'"), |
03f15a79 TG |
1920 | ce->name); |
1921 | } | |
15999d0b JSP |
1922 | } |
1923 | } | |
1924 | ||
435ec090 CC |
1925 | static void tweak_untracked_cache(struct index_state *istate) |
1926 | { | |
ad0fb659 DS |
1927 | struct repository *r = the_repository; |
1928 | ||
1929 | prepare_repo_settings(r); | |
1930 | ||
f1bee828 ÆAB |
1931 | switch (r->settings.core_untracked_cache) { |
1932 | case UNTRACKED_CACHE_REMOVE: | |
435ec090 | 1933 | remove_untracked_cache(istate); |
f1bee828 ÆAB |
1934 | break; |
1935 | case UNTRACKED_CACHE_WRITE: | |
ad0fb659 | 1936 | add_untracked_cache(istate); |
f1bee828 ÆAB |
1937 | break; |
1938 | case UNTRACKED_CACHE_KEEP: | |
3050b6df ÆAB |
1939 | /* |
1940 | * Either an explicit "core.untrackedCache=keep", the | |
1941 | * default if "core.untrackedCache" isn't configured, | |
1942 | * or a fallback on an unknown "core.untrackedCache" | |
1943 | * value. | |
1944 | */ | |
f1bee828 | 1945 | break; |
f1bee828 | 1946 | } |
435ec090 CC |
1947 | } |
1948 | ||
43925312 CC |
1949 | static void tweak_split_index(struct index_state *istate) |
1950 | { | |
1951 | switch (git_config_get_split_index()) { | |
1952 | case -1: /* unset: do nothing */ | |
1953 | break; | |
1954 | case 0: /* false */ | |
1955 | remove_split_index(istate); | |
1956 | break; | |
1957 | case 1: /* true */ | |
1958 | add_split_index(istate); | |
1959 | break; | |
1960 | default: /* unknown value: do nothing */ | |
1961 | break; | |
1962 | } | |
1963 | } | |
1964 | ||
435ec090 CC |
1965 | static void post_read_index_from(struct index_state *istate) |
1966 | { | |
1967 | check_ce_order(istate); | |
1968 | tweak_untracked_cache(istate); | |
43925312 | 1969 | tweak_split_index(istate); |
883e248b | 1970 | tweak_fsmonitor(istate); |
435ec090 CC |
1971 | } |
1972 | ||
8e72d675 JM |
1973 | static size_t estimate_cache_size_from_compressed(unsigned int entries) |
1974 | { | |
1975 | return entries * (sizeof(struct cache_entry) + CACHE_ENTRY_PATH_LENGTH); | |
1976 | } | |
1977 | ||
1978 | static size_t estimate_cache_size(size_t ondisk_size, unsigned int entries) | |
1979 | { | |
1980 | long per_entry = sizeof(struct cache_entry) - sizeof(struct ondisk_cache_entry); | |
1981 | ||
1982 | /* | |
1983 | * Account for potential alignment differences. | |
1984 | */ | |
c097b95a | 1985 | per_entry += align_padding_size(per_entry, 0); |
8e72d675 JM |
1986 | return ondisk_size + entries * per_entry; |
1987 | } | |
1988 | ||
3255089a BP |
1989 | struct index_entry_offset |
1990 | { | |
1991 | /* starting byte offset into index file, count of index entries in this block */ | |
1992 | int offset, nr; | |
1993 | }; | |
1994 | ||
1995 | struct index_entry_offset_table | |
1996 | { | |
1997 | int nr; | |
1998 | struct index_entry_offset entries[FLEX_ARRAY]; | |
1999 | }; | |
2000 | ||
3255089a BP |
2001 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset); |
2002 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot); | |
3255089a | 2003 | |
3b1d9e04 BP |
2004 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size); |
2005 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset); | |
2006 | ||
abb4bb83 BP |
2007 | struct load_index_extensions |
2008 | { | |
abb4bb83 | 2009 | pthread_t pthread; |
abb4bb83 BP |
2010 | struct index_state *istate; |
2011 | const char *mmap; | |
2012 | size_t mmap_size; | |
2013 | unsigned long src_offset; | |
2014 | }; | |
2015 | ||
2016 | static void *load_index_extensions(void *_data) | |
2017 | { | |
2018 | struct load_index_extensions *p = _data; | |
2019 | unsigned long src_offset = p->src_offset; | |
2020 | ||
2021 | while (src_offset <= p->mmap_size - the_hash_algo->rawsz - 8) { | |
2022 | /* After an array of active_nr index entries, | |
2023 | * there can be arbitrary number of extended | |
2024 | * sections, each of which is prefixed with | |
2025 | * extension name (4-byte) and section length | |
2026 | * in 4-byte network byte order. | |
2027 | */ | |
2028 | uint32_t extsize = get_be32(p->mmap + src_offset + 4); | |
2029 | if (read_index_extension(p->istate, | |
2030 | p->mmap + src_offset, | |
2031 | p->mmap + src_offset + 8, | |
2032 | extsize) < 0) { | |
2033 | munmap((void *)p->mmap, p->mmap_size); | |
2034 | die(_("index file corrupt")); | |
2035 | } | |
2036 | src_offset += 8; | |
2037 | src_offset += extsize; | |
2038 | } | |
2039 | ||
2040 | return NULL; | |
2041 | } | |
2042 | ||
77ff1127 BP |
2043 | /* |
2044 | * A helper function that will load the specified range of cache entries | |
2045 | * from the memory mapped file and add them to the given index. | |
2046 | */ | |
2047 | static unsigned long load_cache_entry_block(struct index_state *istate, | |
2048 | struct mem_pool *ce_mem_pool, int offset, int nr, const char *mmap, | |
2049 | unsigned long start_offset, const struct cache_entry *previous_ce) | |
2050 | { | |
2051 | int i; | |
2052 | unsigned long src_offset = start_offset; | |
2053 | ||
2054 | for (i = offset; i < offset + nr; i++) { | |
77ff1127 BP |
2055 | struct cache_entry *ce; |
2056 | unsigned long consumed; | |
2057 | ||
4a6ed30f VD |
2058 | ce = create_from_disk(ce_mem_pool, istate->version, |
2059 | mmap + src_offset, | |
2060 | &consumed, previous_ce); | |
77ff1127 BP |
2061 | set_index_entry(istate, i, ce); |
2062 | ||
2063 | src_offset += consumed; | |
2064 | previous_ce = ce; | |
2065 | } | |
2066 | return src_offset - start_offset; | |
2067 | } | |
2068 | ||
2069 | static unsigned long load_all_cache_entries(struct index_state *istate, | |
2070 | const char *mmap, size_t mmap_size, unsigned long src_offset) | |
2071 | { | |
2072 | unsigned long consumed; | |
2073 | ||
44c7e1a7 | 2074 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2075 | if (istate->version == 4) { |
44c7e1a7 | 2076 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2077 | estimate_cache_size_from_compressed(istate->cache_nr)); |
2078 | } else { | |
44c7e1a7 | 2079 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2080 | estimate_cache_size(mmap_size, istate->cache_nr)); |
2081 | } | |
2082 | ||
2083 | consumed = load_cache_entry_block(istate, istate->ce_mem_pool, | |
2084 | 0, istate->cache_nr, mmap, src_offset, NULL); | |
2085 | return consumed; | |
2086 | } | |
2087 | ||
3255089a BP |
2088 | /* |
2089 | * Mostly randomly chosen maximum thread counts: we | |
2090 | * cap the parallelism to online_cpus() threads, and we want | |
2091 | * to have at least 10000 cache entries per thread for it to | |
2092 | * be worth starting a thread. | |
2093 | */ | |
2094 | ||
2095 | #define THREAD_COST (10000) | |
2096 | ||
77ff1127 BP |
2097 | struct load_cache_entries_thread_data |
2098 | { | |
2099 | pthread_t pthread; | |
2100 | struct index_state *istate; | |
2101 | struct mem_pool *ce_mem_pool; | |
2102 | int offset; | |
2103 | const char *mmap; | |
2104 | struct index_entry_offset_table *ieot; | |
2105 | int ieot_start; /* starting index into the ieot array */ | |
2106 | int ieot_blocks; /* count of ieot entries to process */ | |
2107 | unsigned long consumed; /* return # of bytes in index file processed */ | |
2108 | }; | |
2109 | ||
2110 | /* | |
2111 | * A thread proc to run the load_cache_entries() computation | |
2112 | * across multiple background threads. | |
2113 | */ | |
2114 | static void *load_cache_entries_thread(void *_data) | |
2115 | { | |
2116 | struct load_cache_entries_thread_data *p = _data; | |
2117 | int i; | |
2118 | ||
2119 | /* iterate across all ieot blocks assigned to this thread */ | |
2120 | for (i = p->ieot_start; i < p->ieot_start + p->ieot_blocks; i++) { | |
2121 | p->consumed += load_cache_entry_block(p->istate, p->ce_mem_pool, | |
2122 | p->offset, p->ieot->entries[i].nr, p->mmap, p->ieot->entries[i].offset, NULL); | |
2123 | p->offset += p->ieot->entries[i].nr; | |
2124 | } | |
2125 | return NULL; | |
2126 | } | |
2127 | ||
2128 | static unsigned long load_cache_entries_threaded(struct index_state *istate, const char *mmap, size_t mmap_size, | |
7bd9631b | 2129 | int nr_threads, struct index_entry_offset_table *ieot) |
77ff1127 BP |
2130 | { |
2131 | int i, offset, ieot_blocks, ieot_start, err; | |
2132 | struct load_cache_entries_thread_data *data; | |
2133 | unsigned long consumed = 0; | |
2134 | ||
2135 | /* a little sanity checking */ | |
2136 | if (istate->name_hash_initialized) | |
2137 | BUG("the name hash isn't thread safe"); | |
2138 | ||
44c7e1a7 EN |
2139 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
2140 | mem_pool_init(istate->ce_mem_pool, 0); | |
77ff1127 BP |
2141 | |
2142 | /* ensure we have no more threads than we have blocks to process */ | |
2143 | if (nr_threads > ieot->nr) | |
2144 | nr_threads = ieot->nr; | |
ca56dadb | 2145 | CALLOC_ARRAY(data, nr_threads); |
77ff1127 BP |
2146 | |
2147 | offset = ieot_start = 0; | |
2148 | ieot_blocks = DIV_ROUND_UP(ieot->nr, nr_threads); | |
2149 | for (i = 0; i < nr_threads; i++) { | |
2150 | struct load_cache_entries_thread_data *p = &data[i]; | |
2151 | int nr, j; | |
2152 | ||
2153 | if (ieot_start + ieot_blocks > ieot->nr) | |
2154 | ieot_blocks = ieot->nr - ieot_start; | |
2155 | ||
2156 | p->istate = istate; | |
2157 | p->offset = offset; | |
2158 | p->mmap = mmap; | |
2159 | p->ieot = ieot; | |
2160 | p->ieot_start = ieot_start; | |
2161 | p->ieot_blocks = ieot_blocks; | |
2162 | ||
2163 | /* create a mem_pool for each thread */ | |
2164 | nr = 0; | |
2165 | for (j = p->ieot_start; j < p->ieot_start + p->ieot_blocks; j++) | |
2166 | nr += p->ieot->entries[j].nr; | |
bcd2c5ee | 2167 | p->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2168 | if (istate->version == 4) { |
44c7e1a7 | 2169 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2170 | estimate_cache_size_from_compressed(nr)); |
2171 | } else { | |
44c7e1a7 | 2172 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2173 | estimate_cache_size(mmap_size, nr)); |
2174 | } | |
2175 | ||
2176 | err = pthread_create(&p->pthread, NULL, load_cache_entries_thread, p); | |
2177 | if (err) | |
2178 | die(_("unable to create load_cache_entries thread: %s"), strerror(err)); | |
2179 | ||
2180 | /* increment by the number of cache entries in the ieot block being processed */ | |
2181 | for (j = 0; j < ieot_blocks; j++) | |
2182 | offset += ieot->entries[ieot_start + j].nr; | |
2183 | ieot_start += ieot_blocks; | |
2184 | } | |
2185 | ||
2186 | for (i = 0; i < nr_threads; i++) { | |
2187 | struct load_cache_entries_thread_data *p = &data[i]; | |
2188 | ||
2189 | err = pthread_join(p->pthread, NULL); | |
2190 | if (err) | |
2191 | die(_("unable to join load_cache_entries thread: %s"), strerror(err)); | |
2192 | mem_pool_combine(istate->ce_mem_pool, p->ce_mem_pool); | |
2193 | consumed += p->consumed; | |
2194 | } | |
2195 | ||
2196 | free(data); | |
2197 | ||
2198 | return consumed; | |
2199 | } | |
77ff1127 | 2200 | |
491df5f6 VD |
2201 | static void set_new_index_sparsity(struct index_state *istate) |
2202 | { | |
2203 | /* | |
2204 | * If the index's repo exists, mark it sparse according to | |
2205 | * repo settings. | |
2206 | */ | |
5bdf6d4a ÆAB |
2207 | prepare_repo_settings(istate->repo); |
2208 | if (!istate->repo->settings.command_requires_full_index && | |
2209 | is_sparse_index_allowed(istate, 0)) | |
2210 | istate->sparse_index = 1; | |
491df5f6 VD |
2211 | } |
2212 | ||
8fd2cb40 | 2213 | /* remember to discard_cache() before reading a different cache! */ |
3e52f70b | 2214 | int do_read_index(struct index_state *istate, const char *path, int must_exist) |
e83c5163 | 2215 | { |
77ff1127 | 2216 | int fd; |
e83c5163 | 2217 | struct stat st; |
debed2a6 | 2218 | unsigned long src_offset; |
371ed0de BP |
2219 | const struct cache_header *hdr; |
2220 | const char *mmap; | |
7a51ed66 | 2221 | size_t mmap_size; |
abb4bb83 BP |
2222 | struct load_index_extensions p; |
2223 | size_t extension_offset = 0; | |
77ff1127 BP |
2224 | int nr_threads, cpus; |
2225 | struct index_entry_offset_table *ieot = NULL; | |
e83c5163 | 2226 | |
913e0e99 | 2227 | if (istate->initialized) |
4aab5b46 | 2228 | return istate->cache_nr; |
5d1a5c02 | 2229 | |
fba2f38a KB |
2230 | istate->timestamp.sec = 0; |
2231 | istate->timestamp.nsec = 0; | |
8fd2cb40 | 2232 | fd = open(path, O_RDONLY); |
5d1a5c02 | 2233 | if (fd < 0) { |
491df5f6 VD |
2234 | if (!must_exist && errno == ENOENT) { |
2235 | set_new_index_sparsity(istate); | |
5d1a5c02 | 2236 | return 0; |
491df5f6 | 2237 | } |
9d0a9e90 | 2238 | die_errno(_("%s: index file open failed"), path); |
5d1a5c02 | 2239 | } |
e83c5163 | 2240 | |
3511a377 | 2241 | if (fstat(fd, &st)) |
9d0a9e90 | 2242 | die_errno(_("%s: cannot stat the open index"), path); |
3511a377 | 2243 | |
7a51ed66 | 2244 | mmap_size = xsize_t(st.st_size); |
aab61359 | 2245 | if (mmap_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
9d0a9e90 | 2246 | die(_("%s: index file smaller than expected"), path); |
3511a377 | 2247 | |
02638d1e | 2248 | mmap = xmmap_gently(NULL, mmap_size, PROT_READ, MAP_PRIVATE, fd, 0); |
7a51ed66 | 2249 | if (mmap == MAP_FAILED) |
dc059294 EW |
2250 | die_errno(_("%s: unable to map index file%s"), path, |
2251 | mmap_os_err()); | |
57d84f8d | 2252 | close(fd); |
e83c5163 | 2253 | |
371ed0de | 2254 | hdr = (const struct cache_header *)mmap; |
7a51ed66 | 2255 | if (verify_hdr(hdr, mmap_size) < 0) |
e83c5163 LT |
2256 | goto unmap; |
2257 | ||
92e2cab9 | 2258 | oidread(&istate->oid, (const unsigned char *)hdr + mmap_size - the_hash_algo->rawsz); |
9d227781 | 2259 | istate->version = ntohl(hdr->hdr_version); |
4aab5b46 JH |
2260 | istate->cache_nr = ntohl(hdr->hdr_entries); |
2261 | istate->cache_alloc = alloc_nr(istate->cache_nr); | |
ca56dadb | 2262 | CALLOC_ARRAY(istate->cache, istate->cache_alloc); |
913e0e99 | 2263 | istate->initialized = 1; |
7a51ed66 | 2264 | |
abb4bb83 BP |
2265 | p.istate = istate; |
2266 | p.mmap = mmap; | |
2267 | p.mmap_size = mmap_size; | |
6c9cd161 | 2268 | |
7a51ed66 | 2269 | src_offset = sizeof(*hdr); |
4aab5b46 | 2270 | |
2a9dedef JN |
2271 | if (git_config_get_index_threads(&nr_threads)) |
2272 | nr_threads = 1; | |
7a51ed66 | 2273 | |
77ff1127 BP |
2274 | /* TODO: does creating more threads than cores help? */ |
2275 | if (!nr_threads) { | |
2276 | nr_threads = istate->cache_nr / THREAD_COST; | |
2277 | cpus = online_cpus(); | |
2278 | if (nr_threads > cpus) | |
2279 | nr_threads = cpus; | |
e83c5163 | 2280 | } |
abb4bb83 | 2281 | |
88168b9b NTND |
2282 | if (!HAVE_THREADS) |
2283 | nr_threads = 1; | |
2284 | ||
abb4bb83 BP |
2285 | if (nr_threads > 1) { |
2286 | extension_offset = read_eoie_extension(mmap, mmap_size); | |
2287 | if (extension_offset) { | |
2288 | int err; | |
2289 | ||
2290 | p.src_offset = extension_offset; | |
2291 | err = pthread_create(&p.pthread, NULL, load_index_extensions, &p); | |
2292 | if (err) | |
2293 | die(_("unable to create load_index_extensions thread: %s"), strerror(err)); | |
2294 | ||
2295 | nr_threads--; | |
2296 | } | |
2297 | } | |
abb4bb83 | 2298 | |
77ff1127 BP |
2299 | /* |
2300 | * Locate and read the index entry offset table so that we can use it | |
2301 | * to multi-thread the reading of the cache entries. | |
2302 | */ | |
2303 | if (extension_offset && nr_threads > 1) | |
2304 | ieot = read_ieot_extension(mmap, mmap_size, extension_offset); | |
2305 | ||
2306 | if (ieot) { | |
7bd9631b | 2307 | src_offset += load_cache_entries_threaded(istate, mmap, mmap_size, nr_threads, ieot); |
77ff1127 | 2308 | free(ieot); |
8e72d675 | 2309 | } else { |
77ff1127 | 2310 | src_offset += load_all_cache_entries(istate, mmap, mmap_size, src_offset); |
8e72d675 | 2311 | } |
6c9cd161 | 2312 | |
fba2f38a | 2313 | istate->timestamp.sec = st.st_mtime; |
c06ff490 | 2314 | istate->timestamp.nsec = ST_MTIME_NSEC(st); |
fba2f38a | 2315 | |
abb4bb83 | 2316 | /* if we created a thread, join it otherwise load the extensions on the primary thread */ |
abb4bb83 BP |
2317 | if (extension_offset) { |
2318 | int ret = pthread_join(p.pthread, NULL); | |
2319 | if (ret) | |
2320 | die(_("unable to join load_index_extensions thread: %s"), strerror(ret)); | |
88168b9b | 2321 | } else { |
abb4bb83 BP |
2322 | p.src_offset = src_offset; |
2323 | load_index_extensions(&p); | |
bad68ec9 | 2324 | } |
371ed0de | 2325 | munmap((void *)mmap, mmap_size); |
42fee7a3 JH |
2326 | |
2327 | /* | |
2328 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2329 | * that is associated with the given "istate". | |
2330 | */ | |
2331 | trace2_data_intmax("index", the_repository, "read/version", | |
2332 | istate->version); | |
2333 | trace2_data_intmax("index", the_repository, "read/cache_nr", | |
2334 | istate->cache_nr); | |
2335 | ||
7ca4fc88 VD |
2336 | /* |
2337 | * If the command explicitly requires a full index, force it | |
2338 | * to be full. Otherwise, correct the sparsity based on repository | |
2339 | * settings and other properties of the index (if necessary). | |
2340 | */ | |
4300f844 DS |
2341 | prepare_repo_settings(istate->repo); |
2342 | if (istate->repo->settings.command_requires_full_index) | |
2343 | ensure_full_index(istate); | |
7ca4fc88 VD |
2344 | else |
2345 | ensure_correct_sparsity(istate); | |
4300f844 | 2346 | |
4aab5b46 | 2347 | return istate->cache_nr; |
e83c5163 LT |
2348 | |
2349 | unmap: | |
371ed0de | 2350 | munmap((void *)mmap, mmap_size); |
9d0a9e90 | 2351 | die(_("index file corrupt")); |
e83c5163 LT |
2352 | } |
2353 | ||
0d59ffb4 CC |
2354 | /* |
2355 | * Signal that the shared index is used by updating its mtime. | |
2356 | * | |
2357 | * This way, shared index can be removed if they have not been used | |
2358 | * for some time. | |
2359 | */ | |
a125a223 | 2360 | static void freshen_shared_index(const char *shared_index, int warn) |
0d59ffb4 | 2361 | { |
0d59ffb4 | 2362 | if (!check_and_freshen_file(shared_index, 1) && warn) |
9d0a9e90 | 2363 | warning(_("could not freshen shared index '%s'"), shared_index); |
0d59ffb4 CC |
2364 | } |
2365 | ||
a125a223 TG |
2366 | int read_index_from(struct index_state *istate, const char *path, |
2367 | const char *gitdir) | |
5fc2fc8f NTND |
2368 | { |
2369 | struct split_index *split_index; | |
2370 | int ret; | |
2182abd9 | 2371 | char *base_oid_hex; |
a125a223 | 2372 | char *base_path; |
5fc2fc8f NTND |
2373 | |
2374 | /* istate->initialized covers both .git/index and .git/sharedindex.xxx */ | |
2375 | if (istate->initialized) | |
2376 | return istate->cache_nr; | |
2377 | ||
42fee7a3 JH |
2378 | /* |
2379 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2380 | * that is associated with the given "istate". | |
2381 | */ | |
2382 | trace2_region_enter_printf("index", "do_read_index", the_repository, | |
2383 | "%s", path); | |
c46c406a | 2384 | trace_performance_enter(); |
5fc2fc8f | 2385 | ret = do_read_index(istate, path, 0); |
c46c406a | 2386 | trace_performance_leave("read cache %s", path); |
42fee7a3 JH |
2387 | trace2_region_leave_printf("index", "do_read_index", the_repository, |
2388 | "%s", path); | |
435ec090 | 2389 | |
5fc2fc8f | 2390 | split_index = istate->split_index; |
2182abd9 | 2391 | if (!split_index || is_null_oid(&split_index->base_oid)) { |
435ec090 | 2392 | post_read_index_from(istate); |
5fc2fc8f | 2393 | return ret; |
03f15a79 | 2394 | } |
5fc2fc8f | 2395 | |
c46c406a | 2396 | trace_performance_enter(); |
5fc2fc8f | 2397 | if (split_index->base) |
2f6b1eb7 | 2398 | release_index(split_index->base); |
5fc2fc8f | 2399 | else |
2f6b1eb7 | 2400 | ALLOC_ARRAY(split_index->base, 1); |
6269f8ea | 2401 | index_state_init(split_index->base, istate->repo); |
de6ae5f9 | 2402 | |
2182abd9 | 2403 | base_oid_hex = oid_to_hex(&split_index->base_oid); |
2404 | base_path = xstrfmt("%s/sharedindex.%s", gitdir, base_oid_hex); | |
42fee7a3 JH |
2405 | trace2_region_enter_printf("index", "shared/do_read_index", |
2406 | the_repository, "%s", base_path); | |
998330ac | 2407 | ret = do_read_index(split_index->base, base_path, 0); |
42fee7a3 JH |
2408 | trace2_region_leave_printf("index", "shared/do_read_index", |
2409 | the_repository, "%s", base_path); | |
998330ac SG |
2410 | if (!ret) { |
2411 | char *path_copy = xstrdup(path); | |
652891de JS |
2412 | char *base_path2 = xstrfmt("%s/sharedindex.%s", |
2413 | dirname(path_copy), base_oid_hex); | |
998330ac SG |
2414 | free(path_copy); |
2415 | trace2_region_enter_printf("index", "shared/do_read_index", | |
2416 | the_repository, "%s", base_path2); | |
2417 | ret = do_read_index(split_index->base, base_path2, 1); | |
2418 | trace2_region_leave_printf("index", "shared/do_read_index", | |
2419 | the_repository, "%s", base_path2); | |
652891de | 2420 | free(base_path2); |
998330ac | 2421 | } |
9001dc2a | 2422 | if (!oideq(&split_index->base_oid, &split_index->base->oid)) |
9d0a9e90 | 2423 | die(_("broken index, expect %s in %s, got %s"), |
2182abd9 | 2424 | base_oid_hex, base_path, |
75691ea3 | 2425 | oid_to_hex(&split_index->base->oid)); |
de6ae5f9 | 2426 | |
a125a223 | 2427 | freshen_shared_index(base_path, 0); |
5fc2fc8f | 2428 | merge_base_index(istate); |
435ec090 | 2429 | post_read_index_from(istate); |
c46c406a | 2430 | trace_performance_leave("read cache %s", base_path); |
b42ad7d5 | 2431 | free(base_path); |
5fc2fc8f NTND |
2432 | return ret; |
2433 | } | |
2434 | ||
fa7b3c2f JH |
2435 | int is_index_unborn(struct index_state *istate) |
2436 | { | |
debed2a6 | 2437 | return (!istate->cache_nr && !istate->timestamp.sec); |
fa7b3c2f JH |
2438 | } |
2439 | ||
6269f8ea | 2440 | void index_state_init(struct index_state *istate, struct repository *r) |
2f6b1eb7 | 2441 | { |
6269f8ea | 2442 | struct index_state blank = INDEX_STATE_INIT(r); |
2f6b1eb7 ÆAB |
2443 | memcpy(istate, &blank, sizeof(*istate)); |
2444 | } | |
2445 | ||
2446 | void release_index(struct index_state *istate) | |
6d297f81 | 2447 | { |
8e72d675 JM |
2448 | /* |
2449 | * Cache entries in istate->cache[] should have been allocated | |
2450 | * from the memory pool associated with this index, or from an | |
2451 | * associated split_index. There is no need to free individual | |
8616a2d0 JM |
2452 | * cache entries. validate_cache_entries can detect when this |
2453 | * assertion does not hold. | |
8e72d675 | 2454 | */ |
8616a2d0 | 2455 | validate_cache_entries(istate); |
debed2a6 | 2456 | |
cfc5789a | 2457 | resolve_undo_clear_index(istate); |
2092678c | 2458 | free_name_hash(istate); |
4aab5b46 | 2459 | cache_tree_free(&(istate->cache_tree)); |
2f6b1eb7 ÆAB |
2460 | free(istate->fsmonitor_last_update); |
2461 | free(istate->cache); | |
5fc2fc8f | 2462 | discard_split_index(istate); |
f9e6c649 | 2463 | free_untracked_cache(istate->untracked); |
8e72d675 | 2464 | |
b5fcb1c0 ÆAB |
2465 | if (istate->sparse_checkout_patterns) { |
2466 | clear_pattern_list(istate->sparse_checkout_patterns); | |
2467 | FREE_AND_NULL(istate->sparse_checkout_patterns); | |
2468 | } | |
2469 | ||
8e72d675 | 2470 | if (istate->ce_mem_pool) { |
8616a2d0 | 2471 | mem_pool_discard(istate->ce_mem_pool, should_validate_cache_entries()); |
44c7e1a7 | 2472 | FREE_AND_NULL(istate->ce_mem_pool); |
8e72d675 | 2473 | } |
6d297f81 JS |
2474 | } |
2475 | ||
2f6b1eb7 ÆAB |
2476 | void discard_index(struct index_state *istate) |
2477 | { | |
2478 | release_index(istate); | |
6269f8ea | 2479 | index_state_init(istate, istate->repo); |
2f6b1eb7 ÆAB |
2480 | } |
2481 | ||
8616a2d0 JM |
2482 | /* |
2483 | * Validate the cache entries of this index. | |
2484 | * All cache entries associated with this index | |
2485 | * should have been allocated by the memory pool | |
2486 | * associated with this index, or by a referenced | |
2487 | * split index. | |
2488 | */ | |
2489 | void validate_cache_entries(const struct index_state *istate) | |
2490 | { | |
2491 | int i; | |
2492 | ||
2493 | if (!should_validate_cache_entries() ||!istate || !istate->initialized) | |
2494 | return; | |
2495 | ||
2496 | for (i = 0; i < istate->cache_nr; i++) { | |
2497 | if (!istate) { | |
391408e5 | 2498 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2499 | } else if (!istate->ce_mem_pool || |
2500 | !mem_pool_contains(istate->ce_mem_pool, istate->cache[i])) { | |
2501 | if (!istate->split_index || | |
2502 | !istate->split_index->base || | |
2503 | !istate->split_index->base->ce_mem_pool || | |
2504 | !mem_pool_contains(istate->split_index->base->ce_mem_pool, istate->cache[i])) { | |
391408e5 | 2505 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2506 | } |
2507 | } | |
2508 | } | |
2509 | ||
2510 | if (istate->split_index) | |
2511 | validate_cache_entries(istate->split_index->base); | |
2512 | } | |
2513 | ||
d1f128b0 | 2514 | int unmerged_index(const struct index_state *istate) |
94a5728c DB |
2515 | { |
2516 | int i; | |
2517 | for (i = 0; i < istate->cache_nr; i++) { | |
2518 | if (ce_stage(istate->cache[i])) | |
2519 | return 1; | |
2520 | } | |
2521 | return 0; | |
2522 | } | |
2523 | ||
150fe065 NTND |
2524 | int repo_index_has_changes(struct repository *repo, |
2525 | struct tree *tree, | |
2526 | struct strbuf *sb) | |
cffbfad5 | 2527 | { |
150fe065 | 2528 | struct index_state *istate = repo->index; |
e1f8694f | 2529 | struct object_id cmp; |
cffbfad5 EN |
2530 | int i; |
2531 | ||
e1f8694f EN |
2532 | if (tree) |
2533 | cmp = tree->object.oid; | |
4a93b899 | 2534 | if (tree || !repo_get_oid_tree(repo, "HEAD", &cmp)) { |
cffbfad5 EN |
2535 | struct diff_options opt; |
2536 | ||
150fe065 | 2537 | repo_diff_setup(repo, &opt); |
cffbfad5 EN |
2538 | opt.flags.exit_with_status = 1; |
2539 | if (!sb) | |
2540 | opt.flags.quick = 1; | |
8d833e93 | 2541 | diff_setup_done(&opt); |
e1f8694f | 2542 | do_diff_cache(&cmp, &opt); |
cffbfad5 EN |
2543 | diffcore_std(&opt); |
2544 | for (i = 0; sb && i < diff_queued_diff.nr; i++) { | |
2545 | if (i) | |
2546 | strbuf_addch(sb, ' '); | |
2547 | strbuf_addstr(sb, diff_queued_diff.queue[i]->two->path); | |
2548 | } | |
2549 | diff_flush(&opt); | |
2550 | return opt.flags.has_changes != 0; | |
2551 | } else { | |
0c18c059 DS |
2552 | /* TODO: audit for interaction with sparse-index. */ |
2553 | ensure_full_index(istate); | |
1b9fbefb | 2554 | for (i = 0; sb && i < istate->cache_nr; i++) { |
cffbfad5 EN |
2555 | if (i) |
2556 | strbuf_addch(sb, ' '); | |
1b9fbefb | 2557 | strbuf_addstr(sb, istate->cache[i]->name); |
cffbfad5 | 2558 | } |
1b9fbefb | 2559 | return !!istate->cache_nr; |
cffbfad5 EN |
2560 | } |
2561 | } | |
2562 | ||
410334ed DS |
2563 | static int write_index_ext_header(struct hashfile *f, |
2564 | git_hash_ctx *eoie_f, | |
2565 | unsigned int ext, | |
2566 | unsigned int sz) | |
bad68ec9 | 2567 | { |
410334ed DS |
2568 | hashwrite_be32(f, ext); |
2569 | hashwrite_be32(f, sz); | |
2570 | ||
2571 | if (eoie_f) { | |
2572 | ext = htonl(ext); | |
2573 | sz = htonl(sz); | |
2574 | the_hash_algo->update_fn(eoie_f, &ext, sizeof(ext)); | |
2575 | the_hash_algo->update_fn(eoie_f, &sz, sizeof(sz)); | |
3b1d9e04 | 2576 | } |
410334ed | 2577 | return 0; |
bad68ec9 JH |
2578 | } |
2579 | ||
58bf2a4c NTND |
2580 | static void ce_smudge_racily_clean_entry(struct index_state *istate, |
2581 | struct cache_entry *ce) | |
407c8eb0 JH |
2582 | { |
2583 | /* | |
2584 | * The only thing we care about in this function is to smudge the | |
2585 | * falsely clean entry due to touch-update-touch race, so we leave | |
2586 | * everything else as they are. We are called for entries whose | |
c21d39d7 | 2587 | * ce_stat_data.sd_mtime match the index file mtime. |
c70115b4 JH |
2588 | * |
2589 | * Note that this actually does not do much for gitlinks, for | |
2590 | * which ce_match_stat_basic() always goes to the actual | |
2591 | * contents. The caller checks with is_racy_timestamp() which | |
2592 | * always says "no" for gitlinks, so we are not called for them ;-) | |
407c8eb0 JH |
2593 | */ |
2594 | struct stat st; | |
2595 | ||
2596 | if (lstat(ce->name, &st) < 0) | |
2597 | return; | |
2598 | if (ce_match_stat_basic(ce, &st)) | |
2599 | return; | |
58bf2a4c | 2600 | if (ce_modified_check_fs(istate, ce, &st)) { |
4b3511b0 JH |
2601 | /* This is "racily clean"; smudge it. Note that this |
2602 | * is a tricky code. At first glance, it may appear | |
2603 | * that it can break with this sequence: | |
2604 | * | |
2605 | * $ echo xyzzy >frotz | |
2606 | * $ git-update-index --add frotz | |
2607 | * $ : >frotz | |
2608 | * $ sleep 3 | |
2609 | * $ echo filfre >nitfol | |
2610 | * $ git-update-index --add nitfol | |
2611 | * | |
b7e58b17 | 2612 | * but it does not. When the second update-index runs, |
4b3511b0 JH |
2613 | * it notices that the entry "frotz" has the same timestamp |
2614 | * as index, and if we were to smudge it by resetting its | |
2615 | * size to zero here, then the object name recorded | |
2616 | * in index is the 6-byte file but the cached stat information | |
2617 | * becomes zero --- which would then match what we would | |
a6080a0a | 2618 | * obtain from the filesystem next time we stat("frotz"). |
4b3511b0 JH |
2619 | * |
2620 | * However, the second update-index, before calling | |
2621 | * this function, notices that the cached size is 6 | |
2622 | * bytes and what is on the filesystem is an empty | |
2623 | * file, and never calls us, so the cached size information | |
2624 | * for "frotz" stays 6 which does not match the filesystem. | |
2625 | */ | |
c21d39d7 | 2626 | ce->ce_stat_data.sd_size = 0; |
407c8eb0 JH |
2627 | } |
2628 | } | |
2629 | ||
f136f7bf | 2630 | /* Copy miscellaneous fields but not the name */ |
ce012deb | 2631 | static void copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk, |
f136f7bf | 2632 | struct cache_entry *ce) |
7a51ed66 | 2633 | { |
b60e188c | 2634 | short flags; |
575fa8a3 | 2635 | const unsigned hashsz = the_hash_algo->rawsz; |
2636 | uint16_t *flagsp = (uint16_t *)(ondisk->data + hashsz); | |
b60e188c | 2637 | |
c21d39d7 MH |
2638 | ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec); |
2639 | ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec); | |
2640 | ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec); | |
2641 | ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec); | |
2642 | ondisk->dev = htonl(ce->ce_stat_data.sd_dev); | |
2643 | ondisk->ino = htonl(ce->ce_stat_data.sd_ino); | |
7a51ed66 | 2644 | ondisk->mode = htonl(ce->ce_mode); |
c21d39d7 MH |
2645 | ondisk->uid = htonl(ce->ce_stat_data.sd_uid); |
2646 | ondisk->gid = htonl(ce->ce_stat_data.sd_gid); | |
2647 | ondisk->size = htonl(ce->ce_stat_data.sd_size); | |
575fa8a3 | 2648 | hashcpy(ondisk->data, ce->oid.hash); |
b60e188c | 2649 | |
ce51bf09 | 2650 | flags = ce->ce_flags & ~CE_NAMEMASK; |
b60e188c | 2651 | flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce)); |
575fa8a3 | 2652 | flagsp[0] = htons(flags); |
06aaaa0b | 2653 | if (ce->ce_flags & CE_EXTENDED) { |
575fa8a3 | 2654 | flagsp[1] = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16); |
f136f7bf JH |
2655 | } |
2656 | } | |
2657 | ||
410334ed | 2658 | static int ce_write_entry(struct hashfile *f, struct cache_entry *ce, |
ce012deb | 2659 | struct strbuf *previous_name, struct ondisk_cache_entry *ondisk) |
f136f7bf | 2660 | { |
9d227781 | 2661 | int size; |
00a4b035 RJ |
2662 | unsigned int saved_namelen; |
2663 | int stripped_name = 0; | |
ce012deb | 2664 | static unsigned char padding[8] = { 0x00 }; |
f136f7bf | 2665 | |
b3c96fb1 NTND |
2666 | if (ce->ce_flags & CE_STRIP_NAME) { |
2667 | saved_namelen = ce_namelen(ce); | |
2668 | ce->ce_namelen = 0; | |
00a4b035 | 2669 | stripped_name = 1; |
b3c96fb1 NTND |
2670 | } |
2671 | ||
575fa8a3 | 2672 | size = offsetof(struct ondisk_cache_entry,data) + ondisk_data_size(ce->ce_flags, 0); |
ce012deb | 2673 | |
9d227781 | 2674 | if (!previous_name) { |
ce012deb KW |
2675 | int len = ce_namelen(ce); |
2676 | copy_cache_entry_to_ondisk(ondisk, ce); | |
410334ed DS |
2677 | hashwrite(f, ondisk, size); |
2678 | hashwrite(f, ce->name, len); | |
2679 | hashwrite(f, padding, align_padding_size(size, len)); | |
9d227781 JH |
2680 | } else { |
2681 | int common, to_remove, prefix_size; | |
2682 | unsigned char to_remove_vi[16]; | |
2683 | for (common = 0; | |
2684 | (ce->name[common] && | |
2685 | common < previous_name->len && | |
2686 | ce->name[common] == previous_name->buf[common]); | |
2687 | common++) | |
2688 | ; /* still matching */ | |
2689 | to_remove = previous_name->len - common; | |
2690 | prefix_size = encode_varint(to_remove, to_remove_vi); | |
2691 | ||
ce012deb | 2692 | copy_cache_entry_to_ondisk(ondisk, ce); |
410334ed DS |
2693 | hashwrite(f, ondisk, size); |
2694 | hashwrite(f, to_remove_vi, prefix_size); | |
2695 | hashwrite(f, ce->name + common, ce_namelen(ce) - common); | |
2696 | hashwrite(f, padding, 1); | |
9d227781 JH |
2697 | |
2698 | strbuf_splice(previous_name, common, to_remove, | |
2699 | ce->name + common, ce_namelen(ce) - common); | |
06aaaa0b | 2700 | } |
00a4b035 | 2701 | if (stripped_name) { |
b3c96fb1 NTND |
2702 | ce->ce_namelen = saved_namelen; |
2703 | ce->ce_flags &= ~CE_STRIP_NAME; | |
2704 | } | |
7a51ed66 | 2705 | |
410334ed | 2706 | return 0; |
7a51ed66 LT |
2707 | } |
2708 | ||
426ddeea YM |
2709 | /* |
2710 | * This function verifies if index_state has the correct sha1 of the | |
2711 | * index file. Don't die if we have any other failure, just return 0. | |
2712 | */ | |
2713 | static int verify_index_from(const struct index_state *istate, const char *path) | |
2714 | { | |
2715 | int fd; | |
2716 | ssize_t n; | |
2717 | struct stat st; | |
aab61359 | 2718 | unsigned char hash[GIT_MAX_RAWSZ]; |
426ddeea YM |
2719 | |
2720 | if (!istate->initialized) | |
2721 | return 0; | |
2722 | ||
2723 | fd = open(path, O_RDONLY); | |
2724 | if (fd < 0) | |
2725 | return 0; | |
2726 | ||
2727 | if (fstat(fd, &st)) | |
2728 | goto out; | |
2729 | ||
aab61359 | 2730 | if (st.st_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
426ddeea YM |
2731 | goto out; |
2732 | ||
aab61359 | 2733 | n = pread_in_full(fd, hash, the_hash_algo->rawsz, st.st_size - the_hash_algo->rawsz); |
2734 | if (n != the_hash_algo->rawsz) | |
426ddeea YM |
2735 | goto out; |
2736 | ||
67947c34 | 2737 | if (!hasheq(istate->oid.hash, hash)) |
426ddeea YM |
2738 | goto out; |
2739 | ||
2740 | close(fd); | |
2741 | return 1; | |
2742 | ||
2743 | out: | |
2744 | close(fd); | |
2745 | return 0; | |
2746 | } | |
2747 | ||
1b0d968b | 2748 | static int repo_verify_index(struct repository *repo) |
426ddeea | 2749 | { |
1b0d968b | 2750 | return verify_index_from(repo->index, repo->index_file); |
426ddeea YM |
2751 | } |
2752 | ||
2ede073f | 2753 | int has_racy_timestamp(struct index_state *istate) |
483fbe2b JH |
2754 | { |
2755 | int entries = istate->cache_nr; | |
2756 | int i; | |
2757 | ||
2758 | for (i = 0; i < entries; i++) { | |
2759 | struct cache_entry *ce = istate->cache[i]; | |
2760 | if (is_racy_timestamp(istate, ce)) | |
2761 | return 1; | |
2762 | } | |
2763 | return 0; | |
2764 | } | |
2765 | ||
1b0d968b NTND |
2766 | void repo_update_index_if_able(struct repository *repo, |
2767 | struct lock_file *lockfile) | |
ccdc4ec3 | 2768 | { |
1b0d968b NTND |
2769 | if ((repo->index->cache_changed || |
2770 | has_racy_timestamp(repo->index)) && | |
2771 | repo_verify_index(repo)) | |
2772 | write_locked_index(repo->index, lockfile, COMMIT_LOCK); | |
b74c90fb MÅ |
2773 | else |
2774 | rollback_lock_file(lockfile); | |
ccdc4ec3 JH |
2775 | } |
2776 | ||
d8465500 JN |
2777 | static int record_eoie(void) |
2778 | { | |
2779 | int val; | |
2780 | ||
2781 | if (!git_config_get_bool("index.recordendofindexentries", &val)) | |
2782 | return val; | |
2a9dedef JN |
2783 | |
2784 | /* | |
2785 | * As a convenience, the end of index entries extension | |
2786 | * used for threading is written by default if the user | |
2787 | * explicitly requested threaded index reads. | |
2788 | */ | |
2789 | return !git_config_get_index_threads(&val) && val != 1; | |
d8465500 JN |
2790 | } |
2791 | ||
42916054 JN |
2792 | static int record_ieot(void) |
2793 | { | |
2794 | int val; | |
2795 | ||
2796 | if (!git_config_get_bool("index.recordoffsettable", &val)) | |
2797 | return val; | |
2a9dedef JN |
2798 | |
2799 | /* | |
2800 | * As a convenience, the offset table used for threading is | |
2801 | * written by default if the user explicitly requested | |
2802 | * threaded index reads. | |
2803 | */ | |
2804 | return !git_config_get_index_threads(&val) && val != 1; | |
42916054 JN |
2805 | } |
2806 | ||
3b7a4475 JS |
2807 | enum write_extensions { |
2808 | WRITE_NO_EXTENSION = 0, | |
2809 | WRITE_SPLIT_INDEX_EXTENSION = 1<<0, | |
2810 | WRITE_CACHE_TREE_EXTENSION = 1<<1, | |
2811 | WRITE_RESOLVE_UNDO_EXTENSION = 1<<2, | |
2812 | WRITE_UNTRACKED_CACHE_EXTENSION = 1<<3, | |
2813 | WRITE_FSMONITOR_EXTENSION = 1<<4, | |
2814 | }; | |
2815 | #define WRITE_ALL_EXTENSIONS ((enum write_extensions)-1) | |
2816 | ||
812d6b00 MÅ |
2817 | /* |
2818 | * On success, `tempfile` is closed. If it is the temporary file | |
2819 | * of a `struct lock_file`, we will therefore effectively perform | |
2820 | * a 'close_lock_file_gently()`. Since that is an implementation | |
2821 | * detail of lockfiles, callers of `do_write_index()` should not | |
2822 | * rely on it. | |
2823 | */ | |
9f41c7a6 | 2824 | static int do_write_index(struct index_state *istate, struct tempfile *tempfile, |
3b7a4475 | 2825 | enum write_extensions write_extensions, unsigned flags) |
197ee8c9 | 2826 | { |
ca54d9ba | 2827 | uint64_t start = getnanotime(); |
410334ed DS |
2828 | struct hashfile *f; |
2829 | git_hash_ctx *eoie_c = NULL; | |
197ee8c9 | 2830 | struct cache_header hdr; |
b50386c7 | 2831 | int i, err = 0, removed, extended, hdr_version; |
4aab5b46 JH |
2832 | struct cache_entry **cache = istate->cache; |
2833 | int entries = istate->cache_nr; | |
e1afca4f | 2834 | struct stat st; |
575fa8a3 | 2835 | struct ondisk_cache_entry ondisk; |
9d227781 | 2836 | struct strbuf previous_name_buf = STRBUF_INIT, *previous_name; |
4bddd983 | 2837 | int drop_cache_tree = istate->drop_cache_tree; |
3b1d9e04 | 2838 | off_t offset; |
ba95e96d | 2839 | int csum_fsync_flag; |
77ff1127 | 2840 | int ieot_entries = 1; |
3255089a BP |
2841 | struct index_entry_offset_table *ieot = NULL; |
2842 | int nr, nr_threads; | |
6269f8ea | 2843 | struct repository *r = istate->repo; |
025a0709 | 2844 | |
410334ed DS |
2845 | f = hashfd(tempfile->fd, tempfile->filename.buf); |
2846 | ||
17194b19 DS |
2847 | prepare_repo_settings(r); |
2848 | f->skip_hash = r->settings.index_skip_hash; | |
ee1f0c24 | 2849 | |
06aaaa0b | 2850 | for (i = removed = extended = 0; i < entries; i++) { |
7a51ed66 | 2851 | if (cache[i]->ce_flags & CE_REMOVE) |
025a0709 | 2852 | removed++; |
197ee8c9 | 2853 | |
06aaaa0b NTND |
2854 | /* reduce extended entries if possible */ |
2855 | cache[i]->ce_flags &= ~CE_EXTENDED; | |
2856 | if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) { | |
2857 | extended++; | |
2858 | cache[i]->ce_flags |= CE_EXTENDED; | |
2859 | } | |
2860 | } | |
2861 | ||
e8ffd034 | 2862 | if (!istate->version) |
4a93b899 | 2863 | istate->version = get_index_format_default(r); |
9d227781 JH |
2864 | |
2865 | /* demote version 3 to version 2 when the latter suffices */ | |
2866 | if (istate->version == 3 || istate->version == 2) | |
2867 | istate->version = extended ? 3 : 2; | |
2868 | ||
2869 | hdr_version = istate->version; | |
2870 | ||
ccc4feb5 | 2871 | hdr.hdr_signature = htonl(CACHE_SIGNATURE); |
9d227781 | 2872 | hdr.hdr_version = htonl(hdr_version); |
025a0709 | 2873 | hdr.hdr_entries = htonl(entries - removed); |
197ee8c9 | 2874 | |
410334ed | 2875 | hashwrite(f, &hdr, sizeof(hdr)); |
197ee8c9 | 2876 | |
2a9dedef | 2877 | if (!HAVE_THREADS || git_config_get_index_threads(&nr_threads)) |
88168b9b | 2878 | nr_threads = 1; |
62e5ee81 | 2879 | |
42916054 | 2880 | if (nr_threads != 1 && record_ieot()) { |
3255089a BP |
2881 | int ieot_blocks, cpus; |
2882 | ||
2883 | /* | |
2884 | * ensure default number of ieot blocks maps evenly to the | |
2885 | * default number of threads that will process them leaving | |
2886 | * room for the thread to load the index extensions. | |
2887 | */ | |
2888 | if (!nr_threads) { | |
2889 | ieot_blocks = istate->cache_nr / THREAD_COST; | |
2890 | cpus = online_cpus(); | |
2891 | if (ieot_blocks > cpus - 1) | |
2892 | ieot_blocks = cpus - 1; | |
2893 | } else { | |
2894 | ieot_blocks = nr_threads; | |
77ff1127 BP |
2895 | if (ieot_blocks > istate->cache_nr) |
2896 | ieot_blocks = istate->cache_nr; | |
3255089a BP |
2897 | } |
2898 | ||
2899 | /* | |
2900 | * no reason to write out the IEOT extension if we don't | |
2901 | * have enough blocks to utilize multi-threading | |
2902 | */ | |
2903 | if (ieot_blocks > 1) { | |
2904 | ieot = xcalloc(1, sizeof(struct index_entry_offset_table) | |
2905 | + (ieot_blocks * sizeof(struct index_entry_offset))); | |
77ff1127 | 2906 | ieot_entries = DIV_ROUND_UP(entries, ieot_blocks); |
3255089a BP |
2907 | } |
2908 | } | |
3255089a | 2909 | |
410334ed DS |
2910 | offset = hashfile_total(f); |
2911 | ||
3255089a | 2912 | nr = 0; |
9d227781 | 2913 | previous_name = (hdr_version == 4) ? &previous_name_buf : NULL; |
ce012deb | 2914 | |
197ee8c9 LT |
2915 | for (i = 0; i < entries; i++) { |
2916 | struct cache_entry *ce = cache[i]; | |
7a51ed66 | 2917 | if (ce->ce_flags & CE_REMOVE) |
aa16021e | 2918 | continue; |
e06c43c7 | 2919 | if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce)) |
58bf2a4c | 2920 | ce_smudge_racily_clean_entry(istate, ce); |
99d1a986 | 2921 | if (is_null_oid(&ce->oid)) { |
83bd7437 JK |
2922 | static const char msg[] = "cache entry has null sha1: %s"; |
2923 | static int allow = -1; | |
2924 | ||
2925 | if (allow < 0) | |
2926 | allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0); | |
2927 | if (allow) | |
2928 | warning(msg, ce->name); | |
2929 | else | |
b50386c7 | 2930 | err = error(msg, ce->name); |
a96d3cc3 JK |
2931 | |
2932 | drop_cache_tree = 1; | |
83bd7437 | 2933 | } |
77ff1127 | 2934 | if (ieot && i && (i % ieot_entries == 0)) { |
3255089a BP |
2935 | ieot->entries[ieot->nr].nr = nr; |
2936 | ieot->entries[ieot->nr].offset = offset; | |
2937 | ieot->nr++; | |
2938 | /* | |
2939 | * If we have a V4 index, set the first byte to an invalid | |
2940 | * character to ensure there is nothing common with the previous | |
2941 | * entry | |
2942 | */ | |
2943 | if (previous_name) | |
2944 | previous_name->buf[0] = 0; | |
2945 | nr = 0; | |
410334ed DS |
2946 | |
2947 | offset = hashfile_total(f); | |
3255089a | 2948 | } |
410334ed | 2949 | if (ce_write_entry(f, ce, previous_name, (struct ondisk_cache_entry *)&ondisk) < 0) |
b50386c7 KW |
2950 | err = -1; |
2951 | ||
2952 | if (err) | |
2953 | break; | |
3255089a BP |
2954 | nr++; |
2955 | } | |
2956 | if (ieot && nr) { | |
2957 | ieot->entries[ieot->nr].nr = nr; | |
2958 | ieot->entries[ieot->nr].offset = offset; | |
2959 | ieot->nr++; | |
197ee8c9 | 2960 | } |
9d227781 | 2961 | strbuf_release(&previous_name_buf); |
1af1c2b6 | 2962 | |
3255089a BP |
2963 | if (err) { |
2964 | free(ieot); | |
b50386c7 | 2965 | return err; |
3255089a | 2966 | } |
b50386c7 | 2967 | |
410334ed DS |
2968 | offset = hashfile_total(f); |
2969 | ||
2970 | /* | |
2971 | * The extension headers must be hashed on their own for the | |
2972 | * EOIE extension. Create a hashfile here to compute that hash. | |
2973 | */ | |
2974 | if (offset && record_eoie()) { | |
2975 | CALLOC_ARRAY(eoie_c, 1); | |
2976 | the_hash_algo->init_fn(eoie_c); | |
3255089a | 2977 | } |
3b1d9e04 | 2978 | |
3255089a BP |
2979 | /* |
2980 | * Lets write out CACHE_EXT_INDEXENTRYOFFSETTABLE first so that we | |
2981 | * can minimize the number of extensions we have to scan through to | |
2982 | * find it during load. Write it out regardless of the | |
2983 | * strip_extensions parameter as we need it when loading the shared | |
2984 | * index. | |
2985 | */ | |
3255089a BP |
2986 | if (ieot) { |
2987 | struct strbuf sb = STRBUF_INIT; | |
2988 | ||
2989 | write_ieot_extension(&sb, ieot); | |
410334ed DS |
2990 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_INDEXENTRYOFFSETTABLE, sb.len) < 0; |
2991 | hashwrite(f, sb.buf, sb.len); | |
3255089a BP |
2992 | strbuf_release(&sb); |
2993 | free(ieot); | |
2994 | if (err) | |
2995 | return -1; | |
2996 | } | |
3255089a | 2997 | |
3b7a4475 JS |
2998 | if (write_extensions & WRITE_SPLIT_INDEX_EXTENSION && |
2999 | istate->split_index) { | |
5fc2fc8f NTND |
3000 | struct strbuf sb = STRBUF_INIT; |
3001 | ||
451b66c5 JS |
3002 | if (istate->sparse_index) |
3003 | die(_("cannot write split index for a sparse index")); | |
3004 | ||
5fc2fc8f | 3005 | err = write_link_extension(&sb, istate) < 0 || |
410334ed DS |
3006 | write_index_ext_header(f, eoie_c, CACHE_EXT_LINK, |
3007 | sb.len) < 0; | |
3008 | hashwrite(f, sb.buf, sb.len); | |
5fc2fc8f NTND |
3009 | strbuf_release(&sb); |
3010 | if (err) | |
3011 | return -1; | |
3012 | } | |
3b7a4475 JS |
3013 | if (write_extensions & WRITE_CACHE_TREE_EXTENSION && |
3014 | !drop_cache_tree && istate->cache_tree) { | |
f285a2d7 | 3015 | struct strbuf sb = STRBUF_INIT; |
1dffb8fa | 3016 | |
1dffb8fa | 3017 | cache_tree_write(&sb, istate->cache_tree); |
410334ed DS |
3018 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_TREE, sb.len) < 0; |
3019 | hashwrite(f, sb.buf, sb.len); | |
1dffb8fa PH |
3020 | strbuf_release(&sb); |
3021 | if (err) | |
bad68ec9 | 3022 | return -1; |
bad68ec9 | 3023 | } |
3b7a4475 JS |
3024 | if (write_extensions & WRITE_RESOLVE_UNDO_EXTENSION && |
3025 | istate->resolve_undo) { | |
cfc5789a JH |
3026 | struct strbuf sb = STRBUF_INIT; |
3027 | ||
3028 | resolve_undo_write(&sb, istate->resolve_undo); | |
410334ed DS |
3029 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_RESOLVE_UNDO, |
3030 | sb.len) < 0; | |
3031 | hashwrite(f, sb.buf, sb.len); | |
cfc5789a JH |
3032 | strbuf_release(&sb); |
3033 | if (err) | |
3034 | return -1; | |
3035 | } | |
3b7a4475 JS |
3036 | if (write_extensions & WRITE_UNTRACKED_CACHE_EXTENSION && |
3037 | istate->untracked) { | |
83c094ad NTND |
3038 | struct strbuf sb = STRBUF_INIT; |
3039 | ||
3040 | write_untracked_extension(&sb, istate->untracked); | |
410334ed DS |
3041 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_UNTRACKED, |
3042 | sb.len) < 0; | |
3043 | hashwrite(f, sb.buf, sb.len); | |
83c094ad NTND |
3044 | strbuf_release(&sb); |
3045 | if (err) | |
3046 | return -1; | |
3047 | } | |
3b7a4475 JS |
3048 | if (write_extensions & WRITE_FSMONITOR_EXTENSION && |
3049 | istate->fsmonitor_last_update) { | |
883e248b BP |
3050 | struct strbuf sb = STRBUF_INIT; |
3051 | ||
3052 | write_fsmonitor_extension(&sb, istate); | |
410334ed DS |
3053 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_FSMONITOR, sb.len) < 0; |
3054 | hashwrite(f, sb.buf, sb.len); | |
3b1d9e04 BP |
3055 | strbuf_release(&sb); |
3056 | if (err) | |
3057 | return -1; | |
3058 | } | |
cd42415f | 3059 | if (istate->sparse_index) { |
410334ed | 3060 | if (write_index_ext_header(f, eoie_c, CACHE_EXT_SPARSE_DIRECTORIES, 0) < 0) |
cd42415f DS |
3061 | return -1; |
3062 | } | |
3b1d9e04 BP |
3063 | |
3064 | /* | |
3065 | * CACHE_EXT_ENDOFINDEXENTRIES must be written as the last entry before the SHA1 | |
3066 | * so that it can be found and processed before all the index entries are | |
3067 | * read. Write it out regardless of the strip_extensions parameter as we need it | |
3068 | * when loading the shared index. | |
3069 | */ | |
410334ed | 3070 | if (eoie_c) { |
3b1d9e04 BP |
3071 | struct strbuf sb = STRBUF_INIT; |
3072 | ||
410334ed DS |
3073 | write_eoie_extension(&sb, eoie_c, offset); |
3074 | err = write_index_ext_header(f, NULL, CACHE_EXT_ENDOFINDEXENTRIES, sb.len) < 0; | |
3075 | hashwrite(f, sb.buf, sb.len); | |
883e248b BP |
3076 | strbuf_release(&sb); |
3077 | if (err) | |
3078 | return -1; | |
3079 | } | |
e1afca4f | 3080 | |
ba95e96d NS |
3081 | csum_fsync_flag = 0; |
3082 | if (!alternate_index_output && (flags & COMMIT_LOCK)) | |
3083 | csum_fsync_flag = CSUM_FSYNC; | |
3084 | ||
3085 | finalize_hashfile(f, istate->oid.hash, FSYNC_COMPONENT_INDEX, | |
3086 | CSUM_HASH_IN_STREAM | csum_fsync_flag); | |
3087 | ||
49bd0fc2 | 3088 | if (close_tempfile_gently(tempfile)) { |
6a8c89d0 | 3089 | error(_("could not close '%s'"), get_tempfile_path(tempfile)); |
49bd0fc2 JK |
3090 | return -1; |
3091 | } | |
6a8c89d0 | 3092 | if (stat(get_tempfile_path(tempfile), &st)) |
e1afca4f | 3093 | return -1; |
5bcf109c KB |
3094 | istate->timestamp.sec = (unsigned int)st.st_mtime; |
3095 | istate->timestamp.nsec = ST_MTIME_NSEC(st); | |
ca54d9ba | 3096 | trace_performance_since(start, "write index, changed mask = %x", istate->cache_changed); |
42fee7a3 JH |
3097 | |
3098 | /* | |
3099 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3100 | * that is associated with the given "istate". | |
3101 | */ | |
3102 | trace2_data_intmax("index", the_repository, "write/version", | |
3103 | istate->version); | |
3104 | trace2_data_intmax("index", the_repository, "write/cache_nr", | |
3105 | istate->cache_nr); | |
3106 | ||
e1afca4f | 3107 | return 0; |
197ee8c9 | 3108 | } |
e46bbcf6 | 3109 | |
626f35c8 NTND |
3110 | void set_alternate_index_output(const char *name) |
3111 | { | |
3112 | alternate_index_output = name; | |
3113 | } | |
3114 | ||
3115 | static int commit_locked_index(struct lock_file *lk) | |
3116 | { | |
751baced MH |
3117 | if (alternate_index_output) |
3118 | return commit_lock_file_to(lk, alternate_index_output); | |
3119 | else | |
626f35c8 | 3120 | return commit_lock_file(lk); |
626f35c8 NTND |
3121 | } |
3122 | ||
3b7a4475 JS |
3123 | static int do_write_locked_index(struct index_state *istate, |
3124 | struct lock_file *lock, | |
3125 | unsigned flags, | |
3126 | enum write_extensions write_extensions) | |
03b86647 | 3127 | { |
42fee7a3 | 3128 | int ret; |
9fadb373 | 3129 | int was_full = istate->sparse_index == INDEX_EXPANDED; |
6e773527 | 3130 | |
ce7a9f01 | 3131 | ret = convert_to_sparse(istate, 0); |
6e773527 DS |
3132 | |
3133 | if (ret) { | |
3134 | warning(_("failed to convert to a sparse-index")); | |
3135 | return ret; | |
3136 | } | |
42fee7a3 JH |
3137 | |
3138 | /* | |
3139 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3140 | * that is associated with the given "istate". | |
3141 | */ | |
3142 | trace2_region_enter_printf("index", "do_write_index", the_repository, | |
6a8c89d0 | 3143 | "%s", get_lock_file_path(lock)); |
3b7a4475 | 3144 | ret = do_write_index(istate, lock->tempfile, write_extensions, flags); |
42fee7a3 | 3145 | trace2_region_leave_printf("index", "do_write_index", the_repository, |
6a8c89d0 | 3146 | "%s", get_lock_file_path(lock)); |
42fee7a3 | 3147 | |
6e773527 DS |
3148 | if (was_full) |
3149 | ensure_full_index(istate); | |
3150 | ||
03b86647 NTND |
3151 | if (ret) |
3152 | return ret; | |
03b86647 | 3153 | if (flags & COMMIT_LOCK) |
1956ecd0 BP |
3154 | ret = commit_locked_index(lock); |
3155 | else | |
3156 | ret = close_lock_file_gently(lock); | |
3157 | ||
dbb1c613 | 3158 | run_hooks_l("post-index-change", |
1956ecd0 BP |
3159 | istate->updated_workdir ? "1" : "0", |
3160 | istate->updated_skipworktree ? "1" : "0", NULL); | |
3161 | istate->updated_workdir = 0; | |
3162 | istate->updated_skipworktree = 0; | |
3163 | ||
3164 | return ret; | |
03b86647 NTND |
3165 | } |
3166 | ||
5fc2fc8f NTND |
3167 | static int write_split_index(struct index_state *istate, |
3168 | struct lock_file *lock, | |
3169 | unsigned flags) | |
3170 | { | |
3171 | int ret; | |
3172 | prepare_to_write_split_index(istate); | |
3b7a4475 | 3173 | ret = do_write_locked_index(istate, lock, flags, WRITE_ALL_EXTENSIONS); |
5fc2fc8f NTND |
3174 | finish_writing_split_index(istate); |
3175 | return ret; | |
3176 | } | |
3177 | ||
b9683722 CC |
3178 | static const char *shared_index_expire = "2.weeks.ago"; |
3179 | ||
3180 | static unsigned long get_shared_index_expire_date(void) | |
3181 | { | |
3182 | static unsigned long shared_index_expire_date; | |
3183 | static int shared_index_expire_date_prepared; | |
3184 | ||
3185 | if (!shared_index_expire_date_prepared) { | |
3186 | git_config_get_expiry("splitindex.sharedindexexpire", | |
3187 | &shared_index_expire); | |
3188 | shared_index_expire_date = approxidate(shared_index_expire); | |
3189 | shared_index_expire_date_prepared = 1; | |
3190 | } | |
3191 | ||
3192 | return shared_index_expire_date; | |
3193 | } | |
3194 | ||
3195 | static int should_delete_shared_index(const char *shared_index_path) | |
3196 | { | |
3197 | struct stat st; | |
3198 | unsigned long expiration; | |
3199 | ||
3200 | /* Check timestamp */ | |
3201 | expiration = get_shared_index_expire_date(); | |
3202 | if (!expiration) | |
3203 | return 0; | |
3204 | if (stat(shared_index_path, &st)) | |
78bde923 | 3205 | return error_errno(_("could not stat '%s'"), shared_index_path); |
b9683722 CC |
3206 | if (st.st_mtime > expiration) |
3207 | return 0; | |
3208 | ||
3209 | return 1; | |
3210 | } | |
3211 | ||
3212 | static int clean_shared_index_files(const char *current_hex) | |
3213 | { | |
3214 | struct dirent *de; | |
3215 | DIR *dir = opendir(get_git_dir()); | |
3216 | ||
3217 | if (!dir) | |
3218 | return error_errno(_("unable to open git dir: %s"), get_git_dir()); | |
3219 | ||
3220 | while ((de = readdir(dir)) != NULL) { | |
3221 | const char *sha1_hex; | |
3222 | const char *shared_index_path; | |
3223 | if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex)) | |
3224 | continue; | |
3225 | if (!strcmp(sha1_hex, current_hex)) | |
3226 | continue; | |
3227 | shared_index_path = git_path("%s", de->d_name); | |
3228 | if (should_delete_shared_index(shared_index_path) > 0 && | |
3229 | unlink(shared_index_path)) | |
3230 | warning_errno(_("unable to unlink: %s"), shared_index_path); | |
3231 | } | |
3232 | closedir(dir); | |
3233 | ||
3234 | return 0; | |
3235 | } | |
3236 | ||
a0a96756 | 3237 | static int write_shared_index(struct index_state *istate, |
ba95e96d | 3238 | struct tempfile **temp, unsigned flags) |
c18b80a0 NTND |
3239 | { |
3240 | struct split_index *si = istate->split_index; | |
6e773527 | 3241 | int ret, was_full = !istate->sparse_index; |
c18b80a0 | 3242 | |
c18b80a0 | 3243 | move_cache_to_base_index(istate); |
ce7a9f01 | 3244 | convert_to_sparse(istate, 0); |
42fee7a3 JH |
3245 | |
3246 | trace2_region_enter_printf("index", "shared/do_write_index", | |
6a8c89d0 | 3247 | the_repository, "%s", get_tempfile_path(*temp)); |
3b7a4475 | 3248 | ret = do_write_index(si->base, *temp, WRITE_NO_EXTENSION, flags); |
c173542c | 3249 | trace2_region_leave_printf("index", "shared/do_write_index", |
6a8c89d0 | 3250 | the_repository, "%s", get_tempfile_path(*temp)); |
42fee7a3 | 3251 | |
6e773527 DS |
3252 | if (was_full) |
3253 | ensure_full_index(istate); | |
3254 | ||
59f9d2dd | 3255 | if (ret) |
c18b80a0 | 3256 | return ret; |
7db2d08c | 3257 | ret = adjust_shared_perm(get_tempfile_path(*temp)); |
df801f3f | 3258 | if (ret) { |
9d0a9e90 | 3259 | error(_("cannot fix permission bits on '%s'"), get_tempfile_path(*temp)); |
df801f3f CC |
3260 | return ret; |
3261 | } | |
7db2d08c | 3262 | ret = rename_tempfile(temp, |
75691ea3 | 3263 | git_path("sharedindex.%s", oid_to_hex(&si->base->oid))); |
b9683722 | 3264 | if (!ret) { |
75691ea3 | 3265 | oidcpy(&si->base_oid, &si->base->oid); |
3266 | clean_shared_index_files(oid_to_hex(&si->base->oid)); | |
b9683722 CC |
3267 | } |
3268 | ||
c18b80a0 NTND |
3269 | return ret; |
3270 | } | |
3271 | ||
e6a1dd77 CC |
3272 | static const int default_max_percent_split_change = 20; |
3273 | ||
3274 | static int too_many_not_shared_entries(struct index_state *istate) | |
3275 | { | |
3276 | int i, not_shared = 0; | |
3277 | int max_split = git_config_get_max_percent_split_change(); | |
3278 | ||
3279 | switch (max_split) { | |
3280 | case -1: | |
3281 | /* not or badly configured: use the default value */ | |
3282 | max_split = default_max_percent_split_change; | |
3283 | break; | |
3284 | case 0: | |
3285 | return 1; /* 0% means always write a new shared index */ | |
3286 | case 100: | |
3287 | return 0; /* 100% means never write a new shared index */ | |
3288 | default: | |
3289 | break; /* just use the configured value */ | |
3290 | } | |
3291 | ||
3292 | /* Count not shared entries */ | |
3293 | for (i = 0; i < istate->cache_nr; i++) { | |
3294 | struct cache_entry *ce = istate->cache[i]; | |
3295 | if (!ce->index) | |
3296 | not_shared++; | |
3297 | } | |
3298 | ||
3299 | return (int64_t)istate->cache_nr * max_split < (int64_t)not_shared * 100; | |
3300 | } | |
3301 | ||
03b86647 NTND |
3302 | int write_locked_index(struct index_state *istate, struct lock_file *lock, |
3303 | unsigned flags) | |
3304 | { | |
e8ffd034 | 3305 | int new_shared_index, ret, test_split_index_env; |
5fc2fc8f NTND |
3306 | struct split_index *si = istate->split_index; |
3307 | ||
4592e608 | 3308 | if (git_env_bool("GIT_TEST_CHECK_CACHE_TREE", 0)) |
c207e9e1 | 3309 | cache_tree_verify(the_repository, istate); |
4592e608 | 3310 | |
61000814 MÅ |
3311 | if ((flags & SKIP_IF_UNCHANGED) && !istate->cache_changed) { |
3312 | if (flags & COMMIT_LOCK) | |
3313 | rollback_lock_file(lock); | |
3314 | return 0; | |
3315 | } | |
3316 | ||
3bd28eb2 AV |
3317 | if (istate->fsmonitor_last_update) |
3318 | fill_fsmonitor_bitmap(istate); | |
3319 | ||
e8ffd034 SG |
3320 | test_split_index_env = git_env_bool("GIT_TEST_SPLIT_INDEX", 0); |
3321 | ||
3322 | if ((!si && !test_split_index_env) || | |
3323 | alternate_index_output || | |
5165dd59 | 3324 | (istate->cache_changed & ~EXTMASK)) { |
3b7a4475 JS |
3325 | ret = do_write_locked_index(istate, lock, flags, |
3326 | ~WRITE_SPLIT_INDEX_EXTENSION); | |
df60cf57 | 3327 | goto out; |
5fc2fc8f NTND |
3328 | } |
3329 | ||
e8ffd034 SG |
3330 | if (test_split_index_env) { |
3331 | if (!si) { | |
3332 | si = init_split_index(istate); | |
d6e3c181 | 3333 | istate->cache_changed |= SPLIT_INDEX_ORDERED; |
e8ffd034 SG |
3334 | } else { |
3335 | int v = si->base_oid.hash[0]; | |
3336 | if ((v & 15) < 6) | |
3337 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
3338 | } | |
d6e3c181 | 3339 | } |
e6a1dd77 CC |
3340 | if (too_many_not_shared_entries(istate)) |
3341 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
0d59ffb4 CC |
3342 | |
3343 | new_shared_index = istate->cache_changed & SPLIT_INDEX_ORDERED; | |
3344 | ||
3345 | if (new_shared_index) { | |
59f9d2dd NTND |
3346 | struct tempfile *temp; |
3347 | int saved_errno; | |
3348 | ||
c9d6c788 ÆAB |
3349 | /* Same initial permissions as the main .git/index file */ |
3350 | temp = mks_tempfile_sm(git_path("sharedindex_XXXXXX"), 0, 0666); | |
59f9d2dd | 3351 | if (!temp) { |
3b7a4475 JS |
3352 | ret = do_write_locked_index(istate, lock, flags, |
3353 | ~WRITE_SPLIT_INDEX_EXTENSION); | |
ef5b3a6c NTND |
3354 | goto out; |
3355 | } | |
ba95e96d | 3356 | ret = write_shared_index(istate, &temp, flags); |
59f9d2dd NTND |
3357 | |
3358 | saved_errno = errno; | |
3359 | if (is_tempfile_active(temp)) | |
3360 | delete_tempfile(&temp); | |
3361 | errno = saved_errno; | |
3362 | ||
c18b80a0 | 3363 | if (ret) |
df60cf57 | 3364 | goto out; |
c18b80a0 NTND |
3365 | } |
3366 | ||
0d59ffb4 CC |
3367 | ret = write_split_index(istate, lock, flags); |
3368 | ||
3369 | /* Freshen the shared index only if the split-index was written */ | |
6e37c8ed | 3370 | if (!ret && !new_shared_index && !is_null_oid(&si->base_oid)) { |
a125a223 | 3371 | const char *shared_index = git_path("sharedindex.%s", |
2182abd9 | 3372 | oid_to_hex(&si->base_oid)); |
a125a223 TG |
3373 | freshen_shared_index(shared_index, 1); |
3374 | } | |
0d59ffb4 | 3375 | |
df60cf57 MÅ |
3376 | out: |
3377 | if (flags & COMMIT_LOCK) | |
3378 | rollback_lock_file(lock); | |
0d59ffb4 | 3379 | return ret; |
03b86647 NTND |
3380 | } |
3381 | ||
e46bbcf6 MV |
3382 | /* |
3383 | * Read the index file that is potentially unmerged into given | |
ad376204 EN |
3384 | * index_state, dropping any unmerged entries to stage #0 (potentially |
3385 | * resulting in a path appearing as both a file and a directory in the | |
3386 | * index; the caller is responsible to clear out the extra entries | |
3387 | * before writing the index to a tree). Returns true if the index is | |
3388 | * unmerged. Callers who want to refuse to work from an unmerged | |
3389 | * state can call this and check its return value, instead of calling | |
3390 | * read_cache(). | |
e46bbcf6 | 3391 | */ |
e1ff0a32 | 3392 | int repo_read_index_unmerged(struct repository *repo) |
e46bbcf6 | 3393 | { |
e1ff0a32 | 3394 | struct index_state *istate; |
e46bbcf6 | 3395 | int i; |
d1a43f2a | 3396 | int unmerged = 0; |
e46bbcf6 | 3397 | |
e1ff0a32 NTND |
3398 | repo_read_index(repo); |
3399 | istate = repo->index; | |
e46bbcf6 MV |
3400 | for (i = 0; i < istate->cache_nr; i++) { |
3401 | struct cache_entry *ce = istate->cache[i]; | |
d1a43f2a | 3402 | struct cache_entry *new_ce; |
a849735b | 3403 | int len; |
d1a43f2a JH |
3404 | |
3405 | if (!ce_stage(ce)) | |
e46bbcf6 | 3406 | continue; |
d1a43f2a | 3407 | unmerged = 1; |
68c4f6a5 | 3408 | len = ce_namelen(ce); |
a849735b | 3409 | new_ce = make_empty_cache_entry(istate, len); |
d1a43f2a | 3410 | memcpy(new_ce->name, ce->name, len); |
b60e188c TG |
3411 | new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED; |
3412 | new_ce->ce_namelen = len; | |
d1a43f2a | 3413 | new_ce->ce_mode = ce->ce_mode; |
ad376204 | 3414 | if (add_index_entry(istate, new_ce, ADD_CACHE_SKIP_DFCHECK)) |
9d0a9e90 | 3415 | return error(_("%s: cannot drop to stage #0"), |
5699d17e | 3416 | new_ce->name); |
e46bbcf6 | 3417 | } |
d1a43f2a | 3418 | return unmerged; |
e46bbcf6 | 3419 | } |
041aee31 | 3420 | |
98fa4738 JK |
3421 | /* |
3422 | * Returns 1 if the path is an "other" path with respect to | |
3423 | * the index; that is, the path is not mentioned in the index at all, | |
3424 | * either as a file, a directory with some files in the index, | |
3425 | * or as an unmerged entry. | |
3426 | * | |
3427 | * We helpfully remove a trailing "/" from directories so that | |
3428 | * the output of read_directory can be used as-is. | |
3429 | */ | |
847a9e5d DS |
3430 | int index_name_is_other(struct index_state *istate, const char *name, |
3431 | int namelen) | |
98fa4738 JK |
3432 | { |
3433 | int pos; | |
3434 | if (namelen && name[namelen - 1] == '/') | |
3435 | namelen--; | |
3436 | pos = index_name_pos(istate, name, namelen); | |
3437 | if (0 <= pos) | |
3438 | return 0; /* exact match */ | |
3439 | pos = -pos - 1; | |
3440 | if (pos < istate->cache_nr) { | |
3441 | struct cache_entry *ce = istate->cache[pos]; | |
3442 | if (ce_namelen(ce) == namelen && | |
3443 | !memcmp(ce->name, name, namelen)) | |
3444 | return 0; /* Yup, this one exists unmerged */ | |
3445 | } | |
3446 | return 1; | |
3447 | } | |
29fb37b2 | 3448 | |
847a9e5d | 3449 | void *read_blob_data_from_index(struct index_state *istate, |
87542508 | 3450 | const char *path, unsigned long *size) |
29fb37b2 LF |
3451 | { |
3452 | int pos, len; | |
3453 | unsigned long sz; | |
3454 | enum object_type type; | |
3455 | void *data; | |
3456 | ||
3457 | len = strlen(path); | |
3458 | pos = index_name_pos(istate, path, len); | |
3459 | if (pos < 0) { | |
3460 | /* | |
3461 | * We might be in the middle of a merge, in which | |
3462 | * case we would read stage #2 (ours). | |
3463 | */ | |
3464 | int i; | |
3465 | for (i = -pos - 1; | |
3466 | (pos < 0 && i < istate->cache_nr && | |
3467 | !strcmp(istate->cache[i]->name, path)); | |
3468 | i++) | |
3469 | if (ce_stage(istate->cache[i]) == 2) | |
3470 | pos = i; | |
3471 | } | |
3472 | if (pos < 0) | |
3473 | return NULL; | |
bc726bd0 ÆAB |
3474 | data = repo_read_object_file(the_repository, &istate->cache[pos]->oid, |
3475 | &type, &sz); | |
29fb37b2 LF |
3476 | if (!data || type != OBJ_BLOB) { |
3477 | free(data); | |
3478 | return NULL; | |
3479 | } | |
ff366825 LF |
3480 | if (size) |
3481 | *size = sz; | |
29fb37b2 LF |
3482 | return data; |
3483 | } | |
38612532 | 3484 | |
edf3b905 DT |
3485 | void move_index_extensions(struct index_state *dst, struct index_state *src) |
3486 | { | |
3487 | dst->untracked = src->untracked; | |
3488 | src->untracked = NULL; | |
836ef2b6 NTND |
3489 | dst->cache_tree = src->cache_tree; |
3490 | src->cache_tree = NULL; | |
edf3b905 | 3491 | } |
a849735b | 3492 | |
8e72d675 JM |
3493 | struct cache_entry *dup_cache_entry(const struct cache_entry *ce, |
3494 | struct index_state *istate) | |
3495 | { | |
3496 | unsigned int size = ce_size(ce); | |
3497 | int mem_pool_allocated; | |
3498 | struct cache_entry *new_entry = make_empty_cache_entry(istate, ce_namelen(ce)); | |
3499 | mem_pool_allocated = new_entry->mem_pool_allocated; | |
3500 | ||
3501 | memcpy(new_entry, ce, size); | |
3502 | new_entry->mem_pool_allocated = mem_pool_allocated; | |
3503 | return new_entry; | |
3504 | } | |
3505 | ||
a849735b JM |
3506 | void discard_cache_entry(struct cache_entry *ce) |
3507 | { | |
8616a2d0 JM |
3508 | if (ce && should_validate_cache_entries()) |
3509 | memset(ce, 0xCD, cache_entry_size(ce->ce_namelen)); | |
3510 | ||
8e72d675 JM |
3511 | if (ce && ce->mem_pool_allocated) |
3512 | return; | |
3513 | ||
a849735b JM |
3514 | free(ce); |
3515 | } | |
8616a2d0 JM |
3516 | |
3517 | int should_validate_cache_entries(void) | |
3518 | { | |
3519 | static int validate_index_cache_entries = -1; | |
3520 | ||
3521 | if (validate_index_cache_entries < 0) { | |
3522 | if (getenv("GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES")) | |
3523 | validate_index_cache_entries = 1; | |
3524 | else | |
3525 | validate_index_cache_entries = 0; | |
3526 | } | |
3527 | ||
3528 | return validate_index_cache_entries; | |
3529 | } | |
3b1d9e04 BP |
3530 | |
3531 | #define EOIE_SIZE (4 + GIT_SHA1_RAWSZ) /* <4-byte offset> + <20-byte hash> */ | |
3532 | #define EOIE_SIZE_WITH_HEADER (4 + 4 + EOIE_SIZE) /* <4-byte signature> + <4-byte length> + EOIE_SIZE */ | |
3533 | ||
3534 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size) | |
3535 | { | |
3536 | /* | |
3537 | * The end of index entries (EOIE) extension is guaranteed to be last | |
3538 | * so that it can be found by scanning backwards from the EOF. | |
3539 | * | |
3540 | * "EOIE" | |
3541 | * <4-byte length> | |
3542 | * <4-byte offset> | |
3543 | * <20-byte hash> | |
3544 | */ | |
3545 | const char *index, *eoie; | |
3546 | uint32_t extsize; | |
3547 | size_t offset, src_offset; | |
3548 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3549 | git_hash_ctx c; | |
3550 | ||
3551 | /* ensure we have an index big enough to contain an EOIE extension */ | |
3552 | if (mmap_size < sizeof(struct cache_header) + EOIE_SIZE_WITH_HEADER + the_hash_algo->rawsz) | |
3553 | return 0; | |
3554 | ||
3555 | /* validate the extension signature */ | |
3556 | index = eoie = mmap + mmap_size - EOIE_SIZE_WITH_HEADER - the_hash_algo->rawsz; | |
3557 | if (CACHE_EXT(index) != CACHE_EXT_ENDOFINDEXENTRIES) | |
3558 | return 0; | |
3559 | index += sizeof(uint32_t); | |
3560 | ||
3561 | /* validate the extension size */ | |
3562 | extsize = get_be32(index); | |
3563 | if (extsize != EOIE_SIZE) | |
3564 | return 0; | |
3565 | index += sizeof(uint32_t); | |
3566 | ||
3567 | /* | |
3568 | * Validate the offset we're going to look for the first extension | |
3569 | * signature is after the index header and before the eoie extension. | |
3570 | */ | |
3571 | offset = get_be32(index); | |
3572 | if (mmap + offset < mmap + sizeof(struct cache_header)) | |
3573 | return 0; | |
3574 | if (mmap + offset >= eoie) | |
3575 | return 0; | |
3576 | index += sizeof(uint32_t); | |
3577 | ||
3578 | /* | |
3579 | * The hash is computed over extension types and their sizes (but not | |
3580 | * their contents). E.g. if we have "TREE" extension that is N-bytes | |
3581 | * long, "REUC" extension that is M-bytes long, followed by "EOIE", | |
3582 | * then the hash would be: | |
3583 | * | |
3584 | * SHA-1("TREE" + <binary representation of N> + | |
3585 | * "REUC" + <binary representation of M>) | |
3586 | */ | |
3587 | src_offset = offset; | |
3588 | the_hash_algo->init_fn(&c); | |
3589 | while (src_offset < mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) { | |
3590 | /* After an array of active_nr index entries, | |
3591 | * there can be arbitrary number of extended | |
3592 | * sections, each of which is prefixed with | |
3593 | * extension name (4-byte) and section length | |
3594 | * in 4-byte network byte order. | |
3595 | */ | |
3596 | uint32_t extsize; | |
3597 | memcpy(&extsize, mmap + src_offset + 4, 4); | |
3598 | extsize = ntohl(extsize); | |
3599 | ||
3600 | /* verify the extension size isn't so large it will wrap around */ | |
3601 | if (src_offset + 8 + extsize < src_offset) | |
3602 | return 0; | |
3603 | ||
3604 | the_hash_algo->update_fn(&c, mmap + src_offset, 8); | |
3605 | ||
3606 | src_offset += 8; | |
3607 | src_offset += extsize; | |
3608 | } | |
3609 | the_hash_algo->final_fn(hash, &c); | |
3610 | if (!hasheq(hash, (const unsigned char *)index)) | |
3611 | return 0; | |
3612 | ||
3613 | /* Validate that the extension offsets returned us back to the eoie extension. */ | |
3614 | if (src_offset != mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) | |
3615 | return 0; | |
3616 | ||
3617 | return offset; | |
3618 | } | |
3619 | ||
3620 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset) | |
3621 | { | |
3622 | uint32_t buffer; | |
3623 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3624 | ||
3625 | /* offset */ | |
3626 | put_be32(&buffer, offset); | |
3627 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3628 | ||
3629 | /* hash */ | |
3630 | the_hash_algo->final_fn(hash, eoie_context); | |
3631 | strbuf_add(sb, hash, the_hash_algo->rawsz); | |
3632 | } | |
3255089a | 3633 | |
3255089a BP |
3634 | #define IEOT_VERSION (1) |
3635 | ||
3636 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset) | |
3637 | { | |
ec36c42a NTND |
3638 | const char *index = NULL; |
3639 | uint32_t extsize, ext_version; | |
3640 | struct index_entry_offset_table *ieot; | |
3641 | int i, nr; | |
3642 | ||
3643 | /* find the IEOT extension */ | |
3644 | if (!offset) | |
3645 | return NULL; | |
3646 | while (offset <= mmap_size - the_hash_algo->rawsz - 8) { | |
3647 | extsize = get_be32(mmap + offset + 4); | |
3648 | if (CACHE_EXT((mmap + offset)) == CACHE_EXT_INDEXENTRYOFFSETTABLE) { | |
3649 | index = mmap + offset + 4 + 4; | |
3650 | break; | |
3651 | } | |
3652 | offset += 8; | |
3653 | offset += extsize; | |
3654 | } | |
3655 | if (!index) | |
3656 | return NULL; | |
3657 | ||
3658 | /* validate the version is IEOT_VERSION */ | |
3659 | ext_version = get_be32(index); | |
3660 | if (ext_version != IEOT_VERSION) { | |
3661 | error("invalid IEOT version %d", ext_version); | |
3662 | return NULL; | |
3663 | } | |
3664 | index += sizeof(uint32_t); | |
3665 | ||
3666 | /* extension size - version bytes / bytes per entry */ | |
3667 | nr = (extsize - sizeof(uint32_t)) / (sizeof(uint32_t) + sizeof(uint32_t)); | |
3668 | if (!nr) { | |
3669 | error("invalid number of IEOT entries %d", nr); | |
3670 | return NULL; | |
3671 | } | |
3672 | ieot = xmalloc(sizeof(struct index_entry_offset_table) | |
3673 | + (nr * sizeof(struct index_entry_offset))); | |
3674 | ieot->nr = nr; | |
3675 | for (i = 0; i < nr; i++) { | |
3676 | ieot->entries[i].offset = get_be32(index); | |
3677 | index += sizeof(uint32_t); | |
3678 | ieot->entries[i].nr = get_be32(index); | |
3679 | index += sizeof(uint32_t); | |
3680 | } | |
3681 | ||
3682 | return ieot; | |
3255089a BP |
3683 | } |
3684 | ||
3685 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot) | |
3686 | { | |
ec36c42a NTND |
3687 | uint32_t buffer; |
3688 | int i; | |
3255089a | 3689 | |
ec36c42a NTND |
3690 | /* version */ |
3691 | put_be32(&buffer, IEOT_VERSION); | |
3692 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3693 | |
ec36c42a NTND |
3694 | /* ieot */ |
3695 | for (i = 0; i < ieot->nr; i++) { | |
3255089a | 3696 | |
ec36c42a NTND |
3697 | /* offset */ |
3698 | put_be32(&buffer, ieot->entries[i].offset); | |
3699 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3700 | |
ec36c42a NTND |
3701 | /* count */ |
3702 | put_be32(&buffer, ieot->entries[i].nr); | |
3703 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3704 | } | |
3255089a | 3705 | } |
b2896d27 JT |
3706 | |
3707 | void prefetch_cache_entries(const struct index_state *istate, | |
3708 | must_prefetch_predicate must_prefetch) | |
3709 | { | |
3710 | int i; | |
3711 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
3712 | ||
3713 | for (i = 0; i < istate->cache_nr; i++) { | |
3714 | struct cache_entry *ce = istate->cache[i]; | |
3715 | ||
3716 | if (S_ISGITLINK(ce->ce_mode) || !must_prefetch(ce)) | |
3717 | continue; | |
3718 | if (!oid_object_info_extended(the_repository, &ce->oid, | |
3719 | NULL, | |
3720 | OBJECT_INFO_FOR_PREFETCH)) | |
3721 | continue; | |
3722 | oid_array_append(&to_fetch, &ce->oid); | |
3723 | } | |
3724 | promisor_remote_get_direct(the_repository, | |
3725 | to_fetch.oid, to_fetch.nr); | |
3726 | oid_array_clear(&to_fetch); | |
3727 | } | |
1a40e7be EN |
3728 | |
3729 | static int read_one_entry_opt(struct index_state *istate, | |
3730 | const struct object_id *oid, | |
3731 | struct strbuf *base, | |
3732 | const char *pathname, | |
3733 | unsigned mode, int opt) | |
3734 | { | |
3735 | int len; | |
3736 | struct cache_entry *ce; | |
3737 | ||
3738 | if (S_ISDIR(mode)) | |
3739 | return READ_TREE_RECURSIVE; | |
3740 | ||
3741 | len = strlen(pathname); | |
3742 | ce = make_empty_cache_entry(istate, base->len + len); | |
3743 | ||
3744 | ce->ce_mode = create_ce_mode(mode); | |
3745 | ce->ce_flags = create_ce_flags(1); | |
3746 | ce->ce_namelen = base->len + len; | |
3747 | memcpy(ce->name, base->buf, base->len); | |
3748 | memcpy(ce->name + base->len, pathname, len+1); | |
3749 | oidcpy(&ce->oid, oid); | |
3750 | return add_index_entry(istate, ce, opt); | |
3751 | } | |
3752 | ||
3753 | static int read_one_entry(const struct object_id *oid, struct strbuf *base, | |
3754 | const char *pathname, unsigned mode, | |
3755 | void *context) | |
3756 | { | |
3757 | struct index_state *istate = context; | |
3758 | return read_one_entry_opt(istate, oid, base, pathname, | |
3759 | mode, | |
3760 | ADD_CACHE_OK_TO_ADD|ADD_CACHE_SKIP_DFCHECK); | |
3761 | } | |
3762 | ||
3763 | /* | |
3764 | * This is used when the caller knows there is no existing entries at | |
3765 | * the stage that will conflict with the entry being added. | |
3766 | */ | |
3767 | static int read_one_entry_quick(const struct object_id *oid, struct strbuf *base, | |
3768 | const char *pathname, unsigned mode, | |
3769 | void *context) | |
3770 | { | |
3771 | struct index_state *istate = context; | |
3772 | return read_one_entry_opt(istate, oid, base, pathname, | |
3773 | mode, ADD_CACHE_JUST_APPEND); | |
3774 | } | |
3775 | ||
3776 | /* | |
3777 | * Read the tree specified with --with-tree option | |
3778 | * (typically, HEAD) into stage #1 and then | |
3779 | * squash them down to stage #0. This is used for | |
3780 | * --error-unmatch to list and check the path patterns | |
3781 | * that were given from the command line. We are not | |
3782 | * going to write this index out. | |
3783 | */ | |
3784 | void overlay_tree_on_index(struct index_state *istate, | |
3785 | const char *tree_name, const char *prefix) | |
3786 | { | |
3787 | struct tree *tree; | |
3788 | struct object_id oid; | |
3789 | struct pathspec pathspec; | |
3790 | struct cache_entry *last_stage0 = NULL; | |
3791 | int i; | |
3792 | read_tree_fn_t fn = NULL; | |
3793 | int err; | |
3794 | ||
3795 | if (repo_get_oid(the_repository, tree_name, &oid)) | |
3796 | die("tree-ish %s not found.", tree_name); | |
3797 | tree = parse_tree_indirect(&oid); | |
3798 | if (!tree) | |
3799 | die("bad tree-ish %s", tree_name); | |
3800 | ||
3801 | /* Hoist the unmerged entries up to stage #3 to make room */ | |
3802 | /* TODO: audit for interaction with sparse-index. */ | |
3803 | ensure_full_index(istate); | |
3804 | for (i = 0; i < istate->cache_nr; i++) { | |
3805 | struct cache_entry *ce = istate->cache[i]; | |
3806 | if (!ce_stage(ce)) | |
3807 | continue; | |
3808 | ce->ce_flags |= CE_STAGEMASK; | |
3809 | } | |
3810 | ||
3811 | if (prefix) { | |
3812 | static const char *(matchbuf[1]); | |
3813 | matchbuf[0] = NULL; | |
3814 | parse_pathspec(&pathspec, PATHSPEC_ALL_MAGIC, | |
3815 | PATHSPEC_PREFER_CWD, prefix, matchbuf); | |
3816 | } else | |
3817 | memset(&pathspec, 0, sizeof(pathspec)); | |
3818 | ||
3819 | /* | |
3820 | * See if we have cache entry at the stage. If so, | |
3821 | * do it the original slow way, otherwise, append and then | |
3822 | * sort at the end. | |
3823 | */ | |
3824 | for (i = 0; !fn && i < istate->cache_nr; i++) { | |
3825 | const struct cache_entry *ce = istate->cache[i]; | |
3826 | if (ce_stage(ce) == 1) | |
3827 | fn = read_one_entry; | |
3828 | } | |
3829 | ||
3830 | if (!fn) | |
3831 | fn = read_one_entry_quick; | |
3832 | err = read_tree(the_repository, tree, &pathspec, fn, istate); | |
3833 | clear_pathspec(&pathspec); | |
3834 | if (err) | |
3835 | die("unable to read tree entries %s", tree_name); | |
3836 | ||
3837 | /* | |
3838 | * Sort the cache entry -- we need to nuke the cache tree, though. | |
3839 | */ | |
3840 | if (fn == read_one_entry_quick) { | |
3841 | cache_tree_free(&istate->cache_tree); | |
3842 | QSORT(istate->cache, istate->cache_nr, cmp_cache_name_compare); | |
3843 | } | |
3844 | ||
3845 | for (i = 0; i < istate->cache_nr; i++) { | |
3846 | struct cache_entry *ce = istate->cache[i]; | |
3847 | switch (ce_stage(ce)) { | |
3848 | case 0: | |
3849 | last_stage0 = ce; | |
3850 | /* fallthru */ | |
3851 | default: | |
3852 | continue; | |
3853 | case 1: | |
3854 | /* | |
3855 | * If there is stage #0 entry for this, we do not | |
3856 | * need to show it. We use CE_UPDATE bit to mark | |
3857 | * such an entry. | |
3858 | */ | |
3859 | if (last_stage0 && | |
3860 | !strcmp(last_stage0->name, ce->name)) | |
3861 | ce->ce_flags |= CE_UPDATE; | |
3862 | } | |
3863 | } | |
3864 | } | |
6cee5ebc EN |
3865 | |
3866 | struct update_callback_data { | |
3867 | struct index_state *index; | |
3868 | int include_sparse; | |
3869 | int flags; | |
3870 | int add_errors; | |
3871 | }; | |
3872 | ||
3873 | static int fix_unmerged_status(struct diff_filepair *p, | |
3874 | struct update_callback_data *data) | |
3875 | { | |
3876 | if (p->status != DIFF_STATUS_UNMERGED) | |
3877 | return p->status; | |
3878 | if (!(data->flags & ADD_CACHE_IGNORE_REMOVAL) && !p->two->mode) | |
3879 | /* | |
3880 | * This is not an explicit add request, and the | |
3881 | * path is missing from the working tree (deleted) | |
3882 | */ | |
3883 | return DIFF_STATUS_DELETED; | |
3884 | else | |
3885 | /* | |
3886 | * Either an explicit add request, or path exists | |
3887 | * in the working tree. An attempt to explicitly | |
3888 | * add a path that does not exist in the working tree | |
3889 | * will be caught as an error by the caller immediately. | |
3890 | */ | |
3891 | return DIFF_STATUS_MODIFIED; | |
3892 | } | |
3893 | ||
3894 | static void update_callback(struct diff_queue_struct *q, | |
3895 | struct diff_options *opt UNUSED, void *cbdata) | |
3896 | { | |
3897 | int i; | |
3898 | struct update_callback_data *data = cbdata; | |
3899 | ||
3900 | for (i = 0; i < q->nr; i++) { | |
3901 | struct diff_filepair *p = q->queue[i]; | |
3902 | const char *path = p->one->path; | |
3903 | ||
3904 | if (!data->include_sparse && | |
3905 | !path_in_sparse_checkout(path, data->index)) | |
3906 | continue; | |
3907 | ||
3908 | switch (fix_unmerged_status(p, data)) { | |
3909 | default: | |
3910 | die(_("unexpected diff status %c"), p->status); | |
3911 | case DIFF_STATUS_MODIFIED: | |
3912 | case DIFF_STATUS_TYPE_CHANGED: | |
3913 | if (add_file_to_index(data->index, path, data->flags)) { | |
3914 | if (!(data->flags & ADD_CACHE_IGNORE_ERRORS)) | |
3915 | die(_("updating files failed")); | |
3916 | data->add_errors++; | |
3917 | } | |
3918 | break; | |
3919 | case DIFF_STATUS_DELETED: | |
3920 | if (data->flags & ADD_CACHE_IGNORE_REMOVAL) | |
3921 | break; | |
3922 | if (!(data->flags & ADD_CACHE_PRETEND)) | |
3923 | remove_file_from_index(data->index, path); | |
3924 | if (data->flags & (ADD_CACHE_PRETEND|ADD_CACHE_VERBOSE)) | |
3925 | printf(_("remove '%s'\n"), path); | |
3926 | break; | |
3927 | } | |
3928 | } | |
3929 | } | |
3930 | ||
3931 | int add_files_to_cache(struct repository *repo, const char *prefix, | |
3932 | const struct pathspec *pathspec, int include_sparse, | |
3933 | int flags) | |
3934 | { | |
3935 | struct update_callback_data data; | |
3936 | struct rev_info rev; | |
3937 | ||
3938 | memset(&data, 0, sizeof(data)); | |
3939 | data.index = repo->index; | |
3940 | data.include_sparse = include_sparse; | |
3941 | data.flags = flags; | |
3942 | ||
3943 | repo_init_revisions(repo, &rev, prefix); | |
3944 | setup_revisions(0, NULL, &rev, NULL); | |
3945 | if (pathspec) | |
3946 | copy_pathspec(&rev.prune_data, pathspec); | |
3947 | rev.diffopt.output_format = DIFF_FORMAT_CALLBACK; | |
3948 | rev.diffopt.format_callback = update_callback; | |
3949 | rev.diffopt.format_callback_data = &data; | |
3950 | rev.diffopt.flags.override_submodule_config = 1; | |
3951 | rev.max_count = 0; /* do not compare unmerged paths with stage #2 */ | |
3952 | ||
3953 | /* | |
3954 | * Use an ODB transaction to optimize adding multiple objects. | |
3955 | * This function is invoked from commands other than 'add', which | |
3956 | * may not have their own transaction active. | |
3957 | */ | |
3958 | begin_odb_transaction(); | |
3959 | run_diff_files(&rev, DIFF_RACY_IS_MODIFIED); | |
3960 | end_odb_transaction(); | |
3961 | ||
3962 | release_revisions(&rev); | |
3963 | return !!data.add_errors; | |
3964 | } |