]>
Commit | Line | Data |
---|---|---|
8bc9a0c7 LT |
1 | /* |
2 | * GIT - The information manager from hell | |
3 | * | |
4 | * Copyright (C) Linus Torvalds, 2005 | |
5 | */ | |
bc5c5ec0 | 6 | #include "git-compat-util.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" |
fbffdfb1 | 30 | #include "preload-index.h" |
08c46a49 | 31 | #include "read-cache.h" |
cfc5789a | 32 | #include "resolve-undo.h" |
6cee5ebc | 33 | #include "revision.h" |
1956ecd0 | 34 | #include "run-command.h" |
6c9cd161 | 35 | #include "strbuf.h" |
74ea5c95 | 36 | #include "trace2.h" |
6c9cd161 | 37 | #include "varint.h" |
5fc2fc8f | 38 | #include "split-index.h" |
cb2a5135 | 39 | #include "symlinks.h" |
a42643aa | 40 | #include "utf8.h" |
883e248b | 41 | #include "fsmonitor.h" |
abb4bb83 | 42 | #include "thread-utils.h" |
ae9af122 | 43 | #include "progress.h" |
6e773527 | 44 | #include "sparse-index.h" |
410334ed | 45 | #include "csum-file.h" |
b2896d27 | 46 | #include "promisor-remote.h" |
dbb1c613 | 47 | #include "hook.h" |
d5ebb50d | 48 | #include "wrapper.h" |
bad68ec9 | 49 | |
b60e188c TG |
50 | /* Mask for the name length in ce_flags in the on-disk index */ |
51 | ||
52 | #define CE_NAMEMASK (0x0fff) | |
53 | ||
bad68ec9 JH |
54 | /* Index extensions. |
55 | * | |
56 | * The first letter should be 'A'..'Z' for extensions that are not | |
57 | * necessary for a correct operation (i.e. optimization data). | |
58 | * When new extensions are added that _needs_ to be understood in | |
59 | * order to correctly interpret the index file, pick character that | |
60 | * is outside the range, to cause the reader to abort. | |
61 | */ | |
62 | ||
63 | #define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) ) | |
64 | #define CACHE_EXT_TREE 0x54524545 /* "TREE" */ | |
b659b49b | 65 | #define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */ |
5fc2fc8f | 66 | #define CACHE_EXT_LINK 0x6c696e6b /* "link" */ |
83c094ad | 67 | #define CACHE_EXT_UNTRACKED 0x554E5452 /* "UNTR" */ |
883e248b | 68 | #define CACHE_EXT_FSMONITOR 0x46534D4E /* "FSMN" */ |
3b1d9e04 | 69 | #define CACHE_EXT_ENDOFINDEXENTRIES 0x454F4945 /* "EOIE" */ |
3255089a | 70 | #define CACHE_EXT_INDEXENTRYOFFSETTABLE 0x49454F54 /* "IEOT" */ |
cd42415f | 71 | #define CACHE_EXT_SPARSE_DIRECTORIES 0x73646972 /* "sdir" */ |
5fc2fc8f NTND |
72 | |
73 | /* changes that can be kept in $GIT_DIR/index (basically all extensions) */ | |
e0cf0d7d | 74 | #define EXTMASK (RESOLVE_UNDO_CHANGED | CACHE_TREE_CHANGED | \ |
c18b80a0 | 75 | CE_ENTRY_ADDED | CE_ENTRY_REMOVED | CE_ENTRY_CHANGED | \ |
883e248b | 76 | SPLIT_INDEX_ORDERED | UNTRACKED_CHANGED | FSMONITOR_CHANGED) |
e83c5163 | 77 | |
8e72d675 JM |
78 | |
79 | /* | |
80 | * This is an estimate of the pathname length in the index. We use | |
81 | * this for V4 index files to guess the un-deltafied size of the index | |
82 | * in memory because of pathname deltafication. This is not required | |
83 | * for V2/V3 index formats because their pathnames are not compressed. | |
84 | * If the initial amount of memory set aside is not sufficient, the | |
85 | * mem pool will allocate extra memory. | |
86 | */ | |
87 | #define CACHE_ENTRY_PATH_LENGTH 80 | |
88 | ||
20ec2d03 VD |
89 | enum index_search_mode { |
90 | NO_EXPAND_SPARSE = 0, | |
91 | EXPAND_SPARSE = 1 | |
92 | }; | |
93 | ||
8e72d675 JM |
94 | static inline struct cache_entry *mem_pool__ce_alloc(struct mem_pool *mem_pool, size_t len) |
95 | { | |
96 | struct cache_entry *ce; | |
97 | ce = mem_pool_alloc(mem_pool, cache_entry_size(len)); | |
98 | ce->mem_pool_allocated = 1; | |
99 | return ce; | |
100 | } | |
101 | ||
102 | static inline struct cache_entry *mem_pool__ce_calloc(struct mem_pool *mem_pool, size_t len) | |
103 | { | |
104 | struct cache_entry * ce; | |
105 | ce = mem_pool_calloc(mem_pool, 1, cache_entry_size(len)); | |
106 | ce->mem_pool_allocated = 1; | |
107 | return ce; | |
108 | } | |
109 | ||
110 | static struct mem_pool *find_mem_pool(struct index_state *istate) | |
111 | { | |
112 | struct mem_pool **pool_ptr; | |
113 | ||
114 | if (istate->split_index && istate->split_index->base) | |
115 | pool_ptr = &istate->split_index->base->ce_mem_pool; | |
116 | else | |
117 | pool_ptr = &istate->ce_mem_pool; | |
118 | ||
44c7e1a7 EN |
119 | if (!*pool_ptr) { |
120 | *pool_ptr = xmalloc(sizeof(**pool_ptr)); | |
121 | mem_pool_init(*pool_ptr, 0); | |
122 | } | |
8e72d675 JM |
123 | |
124 | return *pool_ptr; | |
125 | } | |
126 | ||
626f35c8 | 127 | static const char *alternate_index_output; |
8fd2cb40 | 128 | |
9cb76b8c JH |
129 | static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
130 | { | |
4300f844 | 131 | if (S_ISSPARSEDIR(ce->ce_mode)) |
9fadb373 | 132 | istate->sparse_index = INDEX_COLLAPSED; |
4300f844 | 133 | |
9cb76b8c | 134 | istate->cache[nr] = ce; |
96872bc2 | 135 | add_name_hash(istate, ce); |
9cb76b8c JH |
136 | } |
137 | ||
cf558704 LT |
138 | static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
139 | { | |
140 | struct cache_entry *old = istate->cache[nr]; | |
141 | ||
078a58e8 | 142 | replace_index_entry_in_base(istate, old, ce); |
2092678c | 143 | remove_name_hash(istate, old); |
a849735b | 144 | discard_cache_entry(old); |
0e267b7a | 145 | ce->ce_flags &= ~CE_HASHED; |
a22c6371 | 146 | set_index_entry(istate, nr, ce); |
078a58e8 | 147 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 148 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 149 | istate->cache_changed |= CE_ENTRY_CHANGED; |
cf558704 LT |
150 | } |
151 | ||
81dc2307 PB |
152 | void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name) |
153 | { | |
b7f9130a | 154 | struct cache_entry *old_entry = istate->cache[nr], *new_entry, *refreshed; |
81dc2307 PB |
155 | int namelen = strlen(new_name); |
156 | ||
a849735b | 157 | new_entry = make_empty_cache_entry(istate, namelen); |
285c2e25 BW |
158 | copy_cache_entry(new_entry, old_entry); |
159 | new_entry->ce_flags &= ~CE_HASHED; | |
160 | new_entry->ce_namelen = namelen; | |
161 | new_entry->index = 0; | |
162 | memcpy(new_entry->name, new_name, namelen + 1); | |
81dc2307 | 163 | |
285c2e25 BW |
164 | cache_tree_invalidate_path(istate, old_entry->name); |
165 | untracked_cache_remove_from_index(istate, old_entry->name); | |
81dc2307 | 166 | remove_index_entry_at(istate, nr); |
b7f9130a VD |
167 | |
168 | /* | |
169 | * Refresh the new index entry. Using 'refresh_cache_entry' ensures | |
170 | * we only update stat info if the entry is otherwise up-to-date (i.e., | |
171 | * the contents/mode haven't changed). This ensures that we reflect the | |
172 | * 'ctime' of the rename in the index without (incorrectly) updating | |
173 | * the cached stat info to reflect unstaged changes on disk. | |
174 | */ | |
175 | refreshed = refresh_cache_entry(istate, new_entry, CE_MATCH_REFRESH); | |
176 | if (refreshed && refreshed != new_entry) { | |
177 | add_index_entry(istate, refreshed, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); | |
178 | discard_cache_entry(new_entry); | |
179 | } else | |
180 | add_index_entry(istate, new_entry, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); | |
81dc2307 PB |
181 | } |
182 | ||
415e96c8 JH |
183 | /* |
184 | * This only updates the "non-critical" parts of the directory | |
185 | * cache, ie the parts that aren't tracked by GIT, and only used | |
186 | * to validate the cache. | |
187 | */ | |
d4c0a3ac | 188 | void fill_stat_cache_info(struct index_state *istate, struct cache_entry *ce, struct stat *st) |
415e96c8 | 189 | { |
c21d39d7 | 190 | fill_stat_data(&ce->ce_stat_data, st); |
5f73076c JH |
191 | |
192 | if (assume_unchanged) | |
7a51ed66 | 193 | ce->ce_flags |= CE_VALID; |
eadb5831 | 194 | |
883e248b | 195 | if (S_ISREG(st->st_mode)) { |
eadb5831 | 196 | ce_mark_uptodate(ce); |
b5a81697 | 197 | mark_fsmonitor_valid(istate, ce); |
883e248b | 198 | } |
415e96c8 JH |
199 | } |
200 | ||
58bf2a4c NTND |
201 | static int ce_compare_data(struct index_state *istate, |
202 | const struct cache_entry *ce, | |
203 | struct stat *st) | |
29e4d363 JH |
204 | { |
205 | int match = -1; | |
1b8ac5ea | 206 | int fd = git_open_cloexec(ce->name, O_RDONLY); |
29e4d363 JH |
207 | |
208 | if (fd >= 0) { | |
bebfecb9 | 209 | struct object_id oid; |
58bf2a4c | 210 | if (!index_fd(istate, &oid, fd, st, OBJ_BLOB, ce->name, 0)) |
6a29d7b7 | 211 | match = !oideq(&oid, &ce->oid); |
7f8508e8 | 212 | /* index_fd() closed the file descriptor already */ |
29e4d363 JH |
213 | } |
214 | return match; | |
215 | } | |
216 | ||
21a6b9fa | 217 | static int ce_compare_link(const struct cache_entry *ce, size_t expected_size) |
29e4d363 JH |
218 | { |
219 | int match = -1; | |
29e4d363 JH |
220 | void *buffer; |
221 | unsigned long size; | |
21666f1a | 222 | enum object_type type; |
a60272b3 | 223 | struct strbuf sb = STRBUF_INIT; |
29e4d363 | 224 | |
a60272b3 | 225 | if (strbuf_readlink(&sb, ce->name, expected_size)) |
29e4d363 | 226 | return -1; |
a60272b3 | 227 | |
bc726bd0 | 228 | buffer = repo_read_object_file(the_repository, &ce->oid, &type, &size); |
a60272b3 LT |
229 | if (buffer) { |
230 | if (size == sb.len) | |
231 | match = memcmp(buffer, sb.buf, size); | |
232 | free(buffer); | |
29e4d363 | 233 | } |
a60272b3 | 234 | strbuf_release(&sb); |
29e4d363 JH |
235 | return match; |
236 | } | |
237 | ||
21a6b9fa | 238 | static int ce_compare_gitlink(const struct cache_entry *ce) |
f35a6d3b | 239 | { |
1053fe82 | 240 | struct object_id oid; |
f35a6d3b LT |
241 | |
242 | /* | |
243 | * We don't actually require that the .git directory | |
302b9282 | 244 | * under GITLINK directory be a valid git directory. It |
f35a6d3b LT |
245 | * might even be missing (in case nobody populated that |
246 | * sub-project). | |
247 | * | |
248 | * If so, we consider it always to match. | |
249 | */ | |
a98e6101 | 250 | if (resolve_gitlink_ref(ce->name, "HEAD", &oid) < 0) |
f35a6d3b | 251 | return 0; |
6a29d7b7 | 252 | return !oideq(&oid, &ce->oid); |
f35a6d3b LT |
253 | } |
254 | ||
58bf2a4c NTND |
255 | static int ce_modified_check_fs(struct index_state *istate, |
256 | const struct cache_entry *ce, | |
257 | struct stat *st) | |
29e4d363 JH |
258 | { |
259 | switch (st->st_mode & S_IFMT) { | |
260 | case S_IFREG: | |
58bf2a4c | 261 | if (ce_compare_data(istate, ce, st)) |
29e4d363 JH |
262 | return DATA_CHANGED; |
263 | break; | |
264 | case S_IFLNK: | |
dc49cd76 | 265 | if (ce_compare_link(ce, xsize_t(st->st_size))) |
29e4d363 JH |
266 | return DATA_CHANGED; |
267 | break; | |
a8ee75bc | 268 | case S_IFDIR: |
7a51ed66 | 269 | if (S_ISGITLINK(ce->ce_mode)) |
c70115b4 | 270 | return ce_compare_gitlink(ce) ? DATA_CHANGED : 0; |
1cf01a34 | 271 | /* else fallthrough */ |
29e4d363 JH |
272 | default: |
273 | return TYPE_CHANGED; | |
274 | } | |
275 | return 0; | |
276 | } | |
277 | ||
21a6b9fa | 278 | static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st) |
734aab75 LT |
279 | { |
280 | unsigned int changed = 0; | |
281 | ||
7a51ed66 LT |
282 | if (ce->ce_flags & CE_REMOVE) |
283 | return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED; | |
284 | ||
285 | switch (ce->ce_mode & S_IFMT) { | |
8ae0a8c5 KS |
286 | case S_IFREG: |
287 | changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0; | |
3e09cdfd JH |
288 | /* We consider only the owner x bit to be relevant for |
289 | * "mode changes" | |
290 | */ | |
291 | if (trust_executable_bit && | |
7a51ed66 | 292 | (0100 & (ce->ce_mode ^ st->st_mode))) |
ffbe1add | 293 | changed |= MODE_CHANGED; |
8ae0a8c5 KS |
294 | break; |
295 | case S_IFLNK: | |
78a8d641 JS |
296 | if (!S_ISLNK(st->st_mode) && |
297 | (has_symlinks || !S_ISREG(st->st_mode))) | |
298 | changed |= TYPE_CHANGED; | |
8ae0a8c5 | 299 | break; |
302b9282 | 300 | case S_IFGITLINK: |
c70115b4 | 301 | /* We ignore most of the st_xxx fields for gitlinks */ |
f35a6d3b LT |
302 | if (!S_ISDIR(st->st_mode)) |
303 | changed |= TYPE_CHANGED; | |
304 | else if (ce_compare_gitlink(ce)) | |
305 | changed |= DATA_CHANGED; | |
a8ee75bc | 306 | return changed; |
8ae0a8c5 | 307 | default: |
391408e5 | 308 | BUG("unsupported ce_mode: %o", ce->ce_mode); |
8ae0a8c5 | 309 | } |
2cb45e95 | 310 | |
c21d39d7 | 311 | changed |= match_stat_data(&ce->ce_stat_data, st); |
b0391890 | 312 | |
f49c2c22 | 313 | /* Racily smudged entry? */ |
c21d39d7 | 314 | if (!ce->ce_stat_data.sd_size) { |
99d1a986 | 315 | if (!is_empty_blob_sha1(ce->oid.hash)) |
f49c2c22 LT |
316 | changed |= DATA_CHANGED; |
317 | } | |
318 | ||
407c8eb0 JH |
319 | return changed; |
320 | } | |
321 | ||
2bb4cda1 NTND |
322 | static int is_racy_stat(const struct index_state *istate, |
323 | const struct stat_data *sd) | |
6d91da6d | 324 | { |
2bb4cda1 | 325 | return (istate->timestamp.sec && |
fba2f38a KB |
326 | #ifdef USE_NSEC |
327 | /* nanosecond timestamped files can also be racy! */ | |
2bb4cda1 NTND |
328 | (istate->timestamp.sec < sd->sd_mtime.sec || |
329 | (istate->timestamp.sec == sd->sd_mtime.sec && | |
330 | istate->timestamp.nsec <= sd->sd_mtime.nsec)) | |
fba2f38a | 331 | #else |
2bb4cda1 | 332 | istate->timestamp.sec <= sd->sd_mtime.sec |
fba2f38a | 333 | #endif |
2bb4cda1 NTND |
334 | ); |
335 | } | |
336 | ||
5581a019 | 337 | int is_racy_timestamp(const struct index_state *istate, |
2bb4cda1 NTND |
338 | const struct cache_entry *ce) |
339 | { | |
340 | return (!S_ISGITLINK(ce->ce_mode) && | |
341 | is_racy_stat(istate, &ce->ce_stat_data)); | |
6d91da6d JH |
342 | } |
343 | ||
ed4efab1 NTND |
344 | int match_stat_data_racy(const struct index_state *istate, |
345 | const struct stat_data *sd, struct stat *st) | |
346 | { | |
347 | if (is_racy_stat(istate, sd)) | |
348 | return MTIME_CHANGED; | |
349 | return match_stat_data(sd, st); | |
6d91da6d JH |
350 | } |
351 | ||
883e248b | 352 | int ie_match_stat(struct index_state *istate, |
21a6b9fa | 353 | const struct cache_entry *ce, struct stat *st, |
4bd5b7da | 354 | unsigned int options) |
407c8eb0 | 355 | { |
5f73076c | 356 | unsigned int changed; |
4bd5b7da | 357 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 358 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
4bd5b7da | 359 | int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY; |
883e248b | 360 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
5f73076c | 361 | |
883e248b BP |
362 | if (!ignore_fsmonitor) |
363 | refresh_fsmonitor(istate); | |
5f73076c JH |
364 | /* |
365 | * If it's marked as always valid in the index, it's | |
366 | * valid whatever the checked-out copy says. | |
56cac48c NTND |
367 | * |
368 | * skip-worktree has the same effect with higher precedence | |
5f73076c | 369 | */ |
56cac48c NTND |
370 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) |
371 | return 0; | |
7a51ed66 | 372 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) |
5f73076c | 373 | return 0; |
883e248b BP |
374 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) |
375 | return 0; | |
5f73076c | 376 | |
331fcb59 JH |
377 | /* |
378 | * Intent-to-add entries have not been added, so the index entry | |
379 | * by definition never matches what is in the work tree until it | |
380 | * actually gets added. | |
381 | */ | |
895ff3b2 | 382 | if (ce_intent_to_add(ce)) |
331fcb59 JH |
383 | return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED; |
384 | ||
5f73076c | 385 | changed = ce_match_stat_basic(ce, st); |
407c8eb0 | 386 | |
29e4d363 JH |
387 | /* |
388 | * Within 1 second of this sequence: | |
389 | * echo xyzzy >file && git-update-index --add file | |
390 | * running this command: | |
391 | * echo frotz >file | |
392 | * would give a falsely clean cache entry. The mtime and | |
393 | * length match the cache, and other stat fields do not change. | |
394 | * | |
395 | * We could detect this at update-index time (the cache entry | |
396 | * being registered/updated records the same time as "now") | |
397 | * and delay the return from git-update-index, but that would | |
398 | * effectively mean we can make at most one commit per second, | |
399 | * which is not acceptable. Instead, we check cache entries | |
400 | * whose mtime are the same as the index file timestamp more | |
5f73076c | 401 | * carefully than others. |
29e4d363 | 402 | */ |
6d91da6d | 403 | if (!changed && is_racy_timestamp(istate, ce)) { |
42f77406 JH |
404 | if (assume_racy_is_modified) |
405 | changed |= DATA_CHANGED; | |
406 | else | |
58bf2a4c | 407 | changed |= ce_modified_check_fs(istate, ce, st); |
42f77406 | 408 | } |
b0391890 | 409 | |
29e4d363 | 410 | return changed; |
b0391890 JH |
411 | } |
412 | ||
883e248b | 413 | int ie_modified(struct index_state *istate, |
21a6b9fa RS |
414 | const struct cache_entry *ce, |
415 | struct stat *st, unsigned int options) | |
b0391890 | 416 | { |
29e4d363 | 417 | int changed, changed_fs; |
4bd5b7da JH |
418 | |
419 | changed = ie_match_stat(istate, ce, st, options); | |
b0391890 JH |
420 | if (!changed) |
421 | return 0; | |
b0391890 JH |
422 | /* |
423 | * If the mode or type has changed, there's no point in trying | |
424 | * to refresh the entry - it's not going to match | |
425 | */ | |
426 | if (changed & (MODE_CHANGED | TYPE_CHANGED)) | |
427 | return changed; | |
428 | ||
c70115b4 JH |
429 | /* |
430 | * Immediately after read-tree or update-index --cacheinfo, | |
431 | * the length field is zero, as we have never even read the | |
432 | * lstat(2) information once, and we cannot trust DATA_CHANGED | |
433 | * returned by ie_match_stat() which in turn was returned by | |
434 | * ce_match_stat_basic() to signal that the filesize of the | |
435 | * blob changed. We have to actually go to the filesystem to | |
436 | * see if the contents match, and if so, should answer "unchanged". | |
437 | * | |
438 | * The logic does not apply to gitlinks, as ce_match_stat_basic() | |
439 | * already has checked the actual HEAD from the filesystem in the | |
440 | * subproject. If ie_match_stat() already said it is different, | |
441 | * then we know it is. | |
b0391890 | 442 | */ |
c70115b4 | 443 | if ((changed & DATA_CHANGED) && |
c21d39d7 | 444 | (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0)) |
b0391890 JH |
445 | return changed; |
446 | ||
58bf2a4c | 447 | changed_fs = ce_modified_check_fs(istate, ce, st); |
29e4d363 JH |
448 | if (changed_fs) |
449 | return changed | changed_fs; | |
b0391890 JH |
450 | return 0; |
451 | } | |
452 | ||
aabc5617 EN |
453 | static int cache_name_stage_compare(const char *name1, int len1, int stage1, |
454 | const char *name2, int len2, int stage2) | |
958ba6c9 | 455 | { |
958ba6c9 LT |
456 | int cmp; |
457 | ||
ccdd4a0f | 458 | cmp = name_compare(name1, len1, name2, len2); |
0ab9e1e8 LT |
459 | if (cmp) |
460 | return cmp; | |
0ab9e1e8 | 461 | |
b60e188c | 462 | if (stage1 < stage2) |
eb38c22f | 463 | return -1; |
b60e188c | 464 | if (stage1 > stage2) |
eb38c22f | 465 | return 1; |
ccdd4a0f JM |
466 | return 0; |
467 | } | |
468 | ||
aabc5617 | 469 | int cmp_cache_name_compare(const void *a_, const void *b_) |
ccdd4a0f | 470 | { |
aabc5617 | 471 | const struct cache_entry *ce1, *ce2; |
ccdd4a0f | 472 | |
aabc5617 EN |
473 | ce1 = *((const struct cache_entry **)a_); |
474 | ce2 = *((const struct cache_entry **)b_); | |
475 | return cache_name_stage_compare(ce1->name, ce1->ce_namelen, ce_stage(ce1), | |
476 | ce2->name, ce2->ce_namelen, ce_stage(ce2)); | |
eb38c22f LT |
477 | } |
478 | ||
20ec2d03 VD |
479 | static int index_name_stage_pos(struct index_state *istate, |
480 | const char *name, int namelen, | |
481 | int stage, | |
482 | enum index_search_mode search_mode) | |
eb38c22f LT |
483 | { |
484 | int first, last; | |
485 | ||
486 | first = 0; | |
4aab5b46 | 487 | last = istate->cache_nr; |
eb38c22f | 488 | while (last > first) { |
568a05c5 | 489 | int next = first + ((last - first) >> 1); |
4aab5b46 | 490 | struct cache_entry *ce = istate->cache[next]; |
b60e188c | 491 | int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce)); |
eb38c22f | 492 | if (!cmp) |
76e7f4ec | 493 | return next; |
eb38c22f LT |
494 | if (cmp < 0) { |
495 | last = next; | |
496 | continue; | |
497 | } | |
498 | first = next+1; | |
499 | } | |
95e0321c | 500 | |
20ec2d03 | 501 | if (search_mode == EXPAND_SPARSE && istate->sparse_index && |
95e0321c DS |
502 | first > 0) { |
503 | /* Note: first <= istate->cache_nr */ | |
504 | struct cache_entry *ce = istate->cache[first - 1]; | |
505 | ||
506 | /* | |
507 | * If we are in a sparse-index _and_ the entry before the | |
508 | * insertion position is a sparse-directory entry that is | |
509 | * an ancestor of 'name', then we need to expand the index | |
510 | * and search again. This will only trigger once, because | |
511 | * thereafter the index is fully expanded. | |
512 | */ | |
513 | if (S_ISSPARSEDIR(ce->ce_mode) && | |
514 | ce_namelen(ce) < namelen && | |
515 | !strncmp(name, ce->name, ce_namelen(ce))) { | |
516 | ensure_full_index(istate); | |
20ec2d03 | 517 | return index_name_stage_pos(istate, name, namelen, stage, search_mode); |
95e0321c DS |
518 | } |
519 | } | |
520 | ||
76e7f4ec | 521 | return -first-1; |
eb38c22f LT |
522 | } |
523 | ||
847a9e5d | 524 | int index_name_pos(struct index_state *istate, const char *name, int namelen) |
b60e188c | 525 | { |
20ec2d03 VD |
526 | return index_name_stage_pos(istate, name, namelen, 0, EXPAND_SPARSE); |
527 | } | |
528 | ||
9553aa0f VD |
529 | int index_name_pos_sparse(struct index_state *istate, const char *name, int namelen) |
530 | { | |
531 | return index_name_stage_pos(istate, name, namelen, 0, NO_EXPAND_SPARSE); | |
532 | } | |
533 | ||
20ec2d03 VD |
534 | int index_entry_exists(struct index_state *istate, const char *name, int namelen) |
535 | { | |
536 | return index_name_stage_pos(istate, name, namelen, 0, NO_EXPAND_SPARSE) >= 0; | |
b60e188c TG |
537 | } |
538 | ||
4aab5b46 | 539 | int remove_index_entry_at(struct index_state *istate, int pos) |
7b937ca3 | 540 | { |
cf558704 LT |
541 | struct cache_entry *ce = istate->cache[pos]; |
542 | ||
cfc5789a | 543 | record_resolve_undo(istate, ce); |
2092678c | 544 | remove_name_hash(istate, ce); |
045113a5 | 545 | save_or_free_index_entry(istate, ce); |
e636a7b4 | 546 | istate->cache_changed |= CE_ENTRY_REMOVED; |
4aab5b46 JH |
547 | istate->cache_nr--; |
548 | if (pos >= istate->cache_nr) | |
7b937ca3 | 549 | return 0; |
f331ab9d RS |
550 | MOVE_ARRAY(istate->cache + pos, istate->cache + pos + 1, |
551 | istate->cache_nr - pos); | |
7b937ca3 LT |
552 | return 1; |
553 | } | |
554 | ||
36419c8e | 555 | /* |
98e023de | 556 | * Remove all cache entries marked for removal, that is where |
36419c8e KB |
557 | * CE_REMOVE is set in ce_flags. This is much more effective than |
558 | * calling remove_index_entry_at() for each entry to be removed. | |
559 | */ | |
6fdc2057 | 560 | void remove_marked_cache_entries(struct index_state *istate, int invalidate) |
36419c8e KB |
561 | { |
562 | struct cache_entry **ce_array = istate->cache; | |
563 | unsigned int i, j; | |
564 | ||
565 | for (i = j = 0; i < istate->cache_nr; i++) { | |
5699d17e | 566 | if (ce_array[i]->ce_flags & CE_REMOVE) { |
6fdc2057 TG |
567 | if (invalidate) { |
568 | cache_tree_invalidate_path(istate, | |
569 | ce_array[i]->name); | |
570 | untracked_cache_remove_from_index(istate, | |
571 | ce_array[i]->name); | |
572 | } | |
2092678c | 573 | remove_name_hash(istate, ce_array[i]); |
045113a5 | 574 | save_or_free_index_entry(istate, ce_array[i]); |
5699d17e | 575 | } |
36419c8e KB |
576 | else |
577 | ce_array[j++] = ce_array[i]; | |
578 | } | |
ad837d9e NTND |
579 | if (j == istate->cache_nr) |
580 | return; | |
e636a7b4 | 581 | istate->cache_changed |= CE_ENTRY_REMOVED; |
36419c8e KB |
582 | istate->cache_nr = j; |
583 | } | |
584 | ||
4aab5b46 | 585 | int remove_file_from_index(struct index_state *istate, const char *path) |
197ee8c9 | 586 | { |
4aab5b46 | 587 | int pos = index_name_pos(istate, path, strlen(path)); |
c4e3cca1 JH |
588 | if (pos < 0) |
589 | pos = -pos-1; | |
a5400efe | 590 | cache_tree_invalidate_path(istate, path); |
e931371a | 591 | untracked_cache_remove_from_index(istate, path); |
4aab5b46 JH |
592 | while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path)) |
593 | remove_index_entry_at(istate, pos); | |
197ee8c9 LT |
594 | return 0; |
595 | } | |
596 | ||
20314271 JS |
597 | static int compare_name(struct cache_entry *ce, const char *path, int namelen) |
598 | { | |
599 | return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen); | |
600 | } | |
601 | ||
602 | static int index_name_pos_also_unmerged(struct index_state *istate, | |
603 | const char *path, int namelen) | |
604 | { | |
605 | int pos = index_name_pos(istate, path, namelen); | |
606 | struct cache_entry *ce; | |
607 | ||
608 | if (pos >= 0) | |
609 | return pos; | |
610 | ||
611 | /* maybe unmerged? */ | |
612 | pos = -1 - pos; | |
613 | if (pos >= istate->cache_nr || | |
614 | compare_name((ce = istate->cache[pos]), path, namelen)) | |
615 | return -1; | |
616 | ||
617 | /* order of preference: stage 2, 1, 3 */ | |
618 | if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr && | |
619 | ce_stage((ce = istate->cache[pos + 1])) == 2 && | |
620 | !compare_name(ce, path, namelen)) | |
621 | pos++; | |
622 | return pos; | |
623 | } | |
624 | ||
1102952b LT |
625 | static int different_name(struct cache_entry *ce, struct cache_entry *alias) |
626 | { | |
627 | int len = ce_namelen(ce); | |
628 | return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len); | |
629 | } | |
630 | ||
631 | /* | |
632 | * If we add a filename that aliases in the cache, we will use the | |
633 | * name that we already have - but we don't want to update the same | |
634 | * alias twice, because that implies that there were actually two | |
635 | * different files with aliasing names! | |
636 | * | |
637 | * So we use the CE_ADDED flag to verify that the alias was an old | |
638 | * one before we accept it as | |
639 | */ | |
045113a5 NTND |
640 | static struct cache_entry *create_alias_ce(struct index_state *istate, |
641 | struct cache_entry *ce, | |
642 | struct cache_entry *alias) | |
1102952b LT |
643 | { |
644 | int len; | |
285c2e25 | 645 | struct cache_entry *new_entry; |
1102952b LT |
646 | |
647 | if (alias->ce_flags & CE_ADDED) | |
9d0a9e90 NTND |
648 | die(_("will not add file alias '%s' ('%s' already exists in index)"), |
649 | ce->name, alias->name); | |
1102952b LT |
650 | |
651 | /* Ok, create the new entry using the name of the existing alias */ | |
652 | len = ce_namelen(alias); | |
a849735b | 653 | new_entry = make_empty_cache_entry(istate, len); |
285c2e25 BW |
654 | memcpy(new_entry->name, alias->name, len); |
655 | copy_cache_entry(new_entry, ce); | |
045113a5 | 656 | save_or_free_index_entry(istate, ce); |
285c2e25 | 657 | return new_entry; |
1102952b LT |
658 | } |
659 | ||
b4b313f9 | 660 | void set_object_name_for_intent_to_add_entry(struct cache_entry *ce) |
39425819 | 661 | { |
a09c985e | 662 | struct object_id oid; |
c80d226a | 663 | if (write_object_file("", 0, OBJ_BLOB, &oid)) |
9d0a9e90 | 664 | die(_("cannot create an empty blob in the object database")); |
a09c985e | 665 | oidcpy(&ce->oid, &oid); |
39425819 JH |
666 | } |
667 | ||
610d55af | 668 | int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags) |
11be42a4 | 669 | { |
a849735b | 670 | int namelen, was_same; |
d177cab0 | 671 | mode_t st_mode = st->st_mode; |
9472935d | 672 | struct cache_entry *ce, *alias = NULL; |
56cac48c | 673 | unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY; |
38ed1d89 JH |
674 | int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND); |
675 | int pretend = flags & ADD_CACHE_PRETEND; | |
39425819 JH |
676 | int intent_only = flags & ADD_CACHE_INTENT; |
677 | int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE| | |
678 | (intent_only ? ADD_CACHE_NEW_ONLY : 0)); | |
e578d031 | 679 | unsigned hash_flags = pretend ? 0 : HASH_WRITE_OBJECT; |
f937bc2f | 680 | struct object_id oid; |
9472935d | 681 | |
9e5da3d0 JK |
682 | if (flags & ADD_CACHE_RENORMALIZE) |
683 | hash_flags |= HASH_RENORMALIZE; | |
11be42a4 | 684 | |
d177cab0 | 685 | if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode)) |
9d0a9e90 | 686 | return error(_("%s: can only add regular files, symbolic links or git-directories"), path); |
11be42a4 JS |
687 | |
688 | namelen = strlen(path); | |
d177cab0 | 689 | if (S_ISDIR(st_mode)) { |
f937bc2f KM |
690 | if (resolve_gitlink_ref(path, "HEAD", &oid) < 0) |
691 | return error(_("'%s' does not have a commit checked out"), path); | |
09595258 LT |
692 | while (namelen && path[namelen-1] == '/') |
693 | namelen--; | |
694 | } | |
a849735b | 695 | ce = make_empty_cache_entry(istate, namelen); |
11be42a4 | 696 | memcpy(ce->name, path, namelen); |
b60e188c | 697 | ce->ce_namelen = namelen; |
39425819 | 698 | if (!intent_only) |
d4c0a3ac | 699 | fill_stat_cache_info(istate, ce, st); |
388b2acd JH |
700 | else |
701 | ce->ce_flags |= CE_INTENT_TO_ADD; | |
11be42a4 | 702 | |
610d55af TG |
703 | |
704 | if (trust_executable_bit && has_symlinks) { | |
d177cab0 | 705 | ce->ce_mode = create_ce_mode(st_mode); |
610d55af | 706 | } else { |
78a8d641 JS |
707 | /* If there is an existing entry, pick the mode bits and type |
708 | * from it, otherwise assume unexecutable regular file. | |
11be42a4 | 709 | */ |
185c975f | 710 | struct cache_entry *ent; |
20314271 | 711 | int pos = index_name_pos_also_unmerged(istate, path, namelen); |
185c975f | 712 | |
4aab5b46 | 713 | ent = (0 <= pos) ? istate->cache[pos] : NULL; |
d177cab0 | 714 | ce->ce_mode = ce_mode_from_stat(ent, st_mode); |
11be42a4 JS |
715 | } |
716 | ||
dc1ae704 JJ |
717 | /* When core.ignorecase=true, determine if a directory of the same name but differing |
718 | * case already exists within the Git repository. If it does, ensure the directory | |
719 | * case of the file being added to the repository matches (is folded into) the existing | |
720 | * entry's directory case. | |
721 | */ | |
722 | if (ignore_case) { | |
41284eb0 | 723 | adjust_dirname_case(istate, ce->name); |
dc1ae704 | 724 | } |
e2c2a375 | 725 | if (!(flags & ADD_CACHE_RENORMALIZE)) { |
9472935d TB |
726 | alias = index_file_exists(istate, ce->name, |
727 | ce_namelen(ce), ignore_case); | |
728 | if (alias && | |
729 | !ce_stage(alias) && | |
730 | !ie_match_stat(istate, alias, st, ce_option)) { | |
731 | /* Nothing changed, really */ | |
732 | if (!S_ISGITLINK(alias->ce_mode)) | |
733 | ce_mark_uptodate(alias); | |
734 | alias->ce_flags |= CE_ADDED; | |
dc1ae704 | 735 | |
a849735b | 736 | discard_cache_entry(ce); |
9472935d TB |
737 | return 0; |
738 | } | |
0781b8a9 | 739 | } |
39425819 | 740 | if (!intent_only) { |
1c418243 | 741 | if (index_path(istate, &ce->oid, path, st, hash_flags)) { |
a849735b | 742 | discard_cache_entry(ce); |
9d0a9e90 | 743 | return error(_("unable to index file '%s'"), path); |
2d9426b0 | 744 | } |
39425819 | 745 | } else |
b4b313f9 | 746 | set_object_name_for_intent_to_add_entry(ce); |
39425819 | 747 | |
1102952b | 748 | if (ignore_case && alias && different_name(ce, alias)) |
045113a5 | 749 | ce = create_alias_ce(istate, ce, alias); |
1102952b | 750 | ce->ce_flags |= CE_ADDED; |
38ed1d89 | 751 | |
3bf0dd1f | 752 | /* It was suspected to be racily clean, but it turns out to be Ok */ |
38ed1d89 JH |
753 | was_same = (alias && |
754 | !ce_stage(alias) && | |
4a7e27e9 | 755 | oideq(&alias->oid, &ce->oid) && |
38ed1d89 JH |
756 | ce->ce_mode == alias->ce_mode); |
757 | ||
758 | if (pretend) | |
a849735b | 759 | discard_cache_entry(ce); |
067178ed | 760 | else if (add_index_entry(istate, ce, add_option)) { |
a849735b | 761 | discard_cache_entry(ce); |
9d0a9e90 | 762 | return error(_("unable to add '%s' to index"), path); |
067178ed | 763 | } |
38ed1d89 | 764 | if (verbose && !was_same) |
11be42a4 | 765 | printf("add '%s'\n", path); |
11be42a4 JS |
766 | return 0; |
767 | } | |
768 | ||
610d55af | 769 | int add_file_to_index(struct index_state *istate, const char *path, int flags) |
d177cab0 LT |
770 | { |
771 | struct stat st; | |
772 | if (lstat(path, &st)) | |
9d0a9e90 | 773 | die_errno(_("unable to stat '%s'"), path); |
610d55af | 774 | return add_to_index(istate, path, &st, flags); |
d177cab0 LT |
775 | } |
776 | ||
a849735b JM |
777 | struct cache_entry *make_empty_cache_entry(struct index_state *istate, size_t len) |
778 | { | |
8e72d675 | 779 | return mem_pool__ce_calloc(find_mem_pool(istate), len); |
a849735b JM |
780 | } |
781 | ||
96168827 MT |
782 | struct cache_entry *make_empty_transient_cache_entry(size_t len, |
783 | struct mem_pool *ce_mem_pool) | |
a849735b | 784 | { |
96168827 MT |
785 | if (ce_mem_pool) |
786 | return mem_pool__ce_calloc(ce_mem_pool, len); | |
a849735b JM |
787 | return xcalloc(1, cache_entry_size(len)); |
788 | } | |
789 | ||
2a1ae649 RS |
790 | enum verify_path_result { |
791 | PATH_OK, | |
792 | PATH_INVALID, | |
793 | PATH_DIR_WITH_SEP, | |
794 | }; | |
795 | ||
796 | static enum verify_path_result verify_path_internal(const char *, unsigned); | |
797 | ||
798 | int verify_path(const char *path, unsigned mode) | |
799 | { | |
c8ad9d04 | 800 | return verify_path_internal(path, mode) == PATH_OK; |
2a1ae649 RS |
801 | } |
802 | ||
a849735b JM |
803 | struct cache_entry *make_cache_entry(struct index_state *istate, |
804 | unsigned int mode, | |
825ed4d9 JM |
805 | const struct object_id *oid, |
806 | const char *path, | |
807 | int stage, | |
808 | unsigned int refresh_options) | |
6640f881 | 809 | { |
bc1c2caa | 810 | struct cache_entry *ce, *ret; |
a849735b | 811 | int len; |
6640f881 | 812 | |
c8ad9d04 | 813 | if (verify_path_internal(path, mode) == PATH_INVALID) { |
9d0a9e90 | 814 | error(_("invalid path '%s'"), path); |
6640f881 | 815 | return NULL; |
7e7abea9 | 816 | } |
6640f881 CR |
817 | |
818 | len = strlen(path); | |
a849735b | 819 | ce = make_empty_cache_entry(istate, len); |
6640f881 | 820 | |
825ed4d9 | 821 | oidcpy(&ce->oid, oid); |
6640f881 | 822 | memcpy(ce->name, path, len); |
b60e188c TG |
823 | ce->ce_flags = create_ce_flags(stage); |
824 | ce->ce_namelen = len; | |
6640f881 CR |
825 | ce->ce_mode = create_ce_mode(mode); |
826 | ||
d7b665c3 | 827 | ret = refresh_cache_entry(istate, ce, refresh_options); |
915e44c6 | 828 | if (ret != ce) |
a849735b | 829 | discard_cache_entry(ce); |
915e44c6 | 830 | return ret; |
6640f881 CR |
831 | } |
832 | ||
96168827 MT |
833 | struct cache_entry *make_transient_cache_entry(unsigned int mode, |
834 | const struct object_id *oid, | |
835 | const char *path, | |
836 | int stage, | |
837 | struct mem_pool *ce_mem_pool) | |
a849735b JM |
838 | { |
839 | struct cache_entry *ce; | |
840 | int len; | |
841 | ||
842 | if (!verify_path(path, mode)) { | |
9d0a9e90 | 843 | error(_("invalid path '%s'"), path); |
a849735b JM |
844 | return NULL; |
845 | } | |
846 | ||
847 | len = strlen(path); | |
96168827 | 848 | ce = make_empty_transient_cache_entry(len, ce_mem_pool); |
a849735b JM |
849 | |
850 | oidcpy(&ce->oid, oid); | |
851 | memcpy(ce->name, path, len); | |
852 | ce->ce_flags = create_ce_flags(stage); | |
853 | ce->ce_namelen = len; | |
854 | ce->ce_mode = create_ce_mode(mode); | |
855 | ||
856 | return ce; | |
857 | } | |
858 | ||
d9d70966 TG |
859 | /* |
860 | * Chmod an index entry with either +x or -x. | |
861 | * | |
862 | * Returns -1 if the chmod for the particular cache entry failed (if it's | |
863 | * not a regular file), -2 if an invalid flip argument is passed in, 0 | |
864 | * otherwise. | |
865 | */ | |
866 | int chmod_index_entry(struct index_state *istate, struct cache_entry *ce, | |
867 | char flip) | |
868 | { | |
869 | if (!S_ISREG(ce->ce_mode)) | |
870 | return -1; | |
871 | switch (flip) { | |
872 | case '+': | |
873 | ce->ce_mode |= 0111; | |
874 | break; | |
875 | case '-': | |
876 | ce->ce_mode &= ~0111; | |
877 | break; | |
878 | default: | |
879 | return -2; | |
880 | } | |
881 | cache_tree_invalidate_path(istate, ce->name); | |
882 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
883e248b | 883 | mark_fsmonitor_invalid(istate, ce); |
d9d70966 TG |
884 | istate->cache_changed |= CE_ENTRY_CHANGED; |
885 | ||
886 | return 0; | |
887 | } | |
888 | ||
9c5e6c80 | 889 | int ce_same_name(const struct cache_entry *a, const struct cache_entry *b) |
7b937ca3 LT |
890 | { |
891 | int len = ce_namelen(a); | |
892 | return ce_namelen(b) == len && !memcmp(a->name, b->name, len); | |
893 | } | |
894 | ||
8dcf39c4 LT |
895 | /* |
896 | * We fundamentally don't like some paths: we don't want | |
897 | * dot or dot-dot anywhere, and for obvious reasons don't | |
898 | * want to recurse into ".git" either. | |
899 | * | |
900 | * Also, we don't want double slashes or slashes at the | |
901 | * end that can make pathnames ambiguous. | |
902 | */ | |
10ecfa76 | 903 | static int verify_dotfile(const char *rest, unsigned mode) |
8dcf39c4 LT |
904 | { |
905 | /* | |
906 | * The first character was '.', but that | |
907 | * has already been discarded, we now test | |
908 | * the rest. | |
909 | */ | |
e0f530ff | 910 | |
8dcf39c4 | 911 | /* "." is not allowed */ |
e0f530ff | 912 | if (*rest == '\0' || is_dir_sep(*rest)) |
8dcf39c4 LT |
913 | return 0; |
914 | ||
e0f530ff | 915 | switch (*rest) { |
8dcf39c4 | 916 | /* |
641084b6 JK |
917 | * ".git" followed by NUL or slash is bad. Note that we match |
918 | * case-insensitively here, even if ignore_case is not set. | |
919 | * This outlaws ".GIT" everywhere out of an abundance of caution, | |
920 | * since there's really no good reason to allow it. | |
10ecfa76 JK |
921 | * |
922 | * Once we've seen ".git", we can also find ".gitmodules", etc (also | |
923 | * case-insensitively). | |
8dcf39c4 LT |
924 | */ |
925 | case 'g': | |
cc2fc7c2 JK |
926 | case 'G': |
927 | if (rest[1] != 'i' && rest[1] != 'I') | |
8dcf39c4 | 928 | break; |
cc2fc7c2 | 929 | if (rest[2] != 't' && rest[2] != 'T') |
8dcf39c4 | 930 | break; |
e19e5e66 JK |
931 | if (rest[3] == '\0' || is_dir_sep(rest[3])) |
932 | return 0; | |
10ecfa76 JK |
933 | if (S_ISLNK(mode)) { |
934 | rest += 3; | |
935 | if (skip_iprefix(rest, "modules", &rest) && | |
936 | (*rest == '\0' || is_dir_sep(*rest))) | |
937 | return 0; | |
938 | } | |
e19e5e66 | 939 | break; |
8dcf39c4 | 940 | case '.': |
e0f530ff | 941 | if (rest[1] == '\0' || is_dir_sep(rest[1])) |
8dcf39c4 LT |
942 | return 0; |
943 | } | |
944 | return 1; | |
945 | } | |
946 | ||
2a1ae649 RS |
947 | static enum verify_path_result verify_path_internal(const char *path, |
948 | unsigned mode) | |
8dcf39c4 | 949 | { |
49e268e2 | 950 | char c = 0; |
8dcf39c4 | 951 | |
56948cb6 | 952 | if (has_dos_drive_prefix(path)) |
2a1ae649 | 953 | return PATH_INVALID; |
56948cb6 | 954 | |
d2c84dad | 955 | if (!is_valid_path(path)) |
2a1ae649 | 956 | return PATH_INVALID; |
d2c84dad | 957 | |
8dcf39c4 LT |
958 | goto inside; |
959 | for (;;) { | |
960 | if (!c) | |
2a1ae649 | 961 | return PATH_OK; |
56948cb6 | 962 | if (is_dir_sep(c)) { |
8dcf39c4 | 963 | inside: |
10ecfa76 | 964 | if (protect_hfs) { |
49e268e2 | 965 | |
10ecfa76 | 966 | if (is_hfs_dotgit(path)) |
2a1ae649 | 967 | return PATH_INVALID; |
10ecfa76 JK |
968 | if (S_ISLNK(mode)) { |
969 | if (is_hfs_dotgitmodules(path)) | |
2a1ae649 | 970 | return PATH_INVALID; |
10ecfa76 JK |
971 | } |
972 | } | |
973 | if (protect_ntfs) { | |
bccc37fd | 974 | #if defined GIT_WINDOWS_NATIVE || defined __CYGWIN__ |
49e268e2 | 975 | if (c == '\\') |
2a1ae649 | 976 | return PATH_INVALID; |
49e268e2 | 977 | #endif |
10ecfa76 | 978 | if (is_ntfs_dotgit(path)) |
2a1ae649 | 979 | return PATH_INVALID; |
10ecfa76 JK |
980 | if (S_ISLNK(mode)) { |
981 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 982 | return PATH_INVALID; |
10ecfa76 JK |
983 | } |
984 | } | |
985 | ||
8dcf39c4 | 986 | c = *path++; |
10ecfa76 | 987 | if ((c == '.' && !verify_dotfile(path, mode)) || |
6e773527 | 988 | is_dir_sep(c)) |
2a1ae649 | 989 | return PATH_INVALID; |
6e773527 DS |
990 | /* |
991 | * allow terminating directory separators for | |
992 | * sparse directory entries. | |
993 | */ | |
994 | if (c == '\0') | |
2a1ae649 RS |
995 | return S_ISDIR(mode) ? PATH_DIR_WITH_SEP : |
996 | PATH_INVALID; | |
288a74bc JS |
997 | } else if (c == '\\' && protect_ntfs) { |
998 | if (is_ntfs_dotgit(path)) | |
2a1ae649 | 999 | return PATH_INVALID; |
288a74bc JS |
1000 | if (S_ISLNK(mode)) { |
1001 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 1002 | return PATH_INVALID; |
288a74bc | 1003 | } |
8dcf39c4 | 1004 | } |
288a74bc | 1005 | |
8dcf39c4 LT |
1006 | c = *path++; |
1007 | } | |
1008 | } | |
1009 | ||
12676608 LT |
1010 | /* |
1011 | * Do we have another file that has the beginning components being a | |
1012 | * proper superset of the name we're trying to add? | |
0f1e4f04 | 1013 | */ |
4aab5b46 JH |
1014 | static int has_file_name(struct index_state *istate, |
1015 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
0f1e4f04 | 1016 | { |
12676608 LT |
1017 | int retval = 0; |
1018 | int len = ce_namelen(ce); | |
b155725d | 1019 | int stage = ce_stage(ce); |
12676608 | 1020 | const char *name = ce->name; |
0f1e4f04 | 1021 | |
4aab5b46 JH |
1022 | while (pos < istate->cache_nr) { |
1023 | struct cache_entry *p = istate->cache[pos++]; | |
0f1e4f04 | 1024 | |
12676608 | 1025 | if (len >= ce_namelen(p)) |
0f1e4f04 | 1026 | break; |
12676608 LT |
1027 | if (memcmp(name, p->name, len)) |
1028 | break; | |
b155725d JH |
1029 | if (ce_stage(p) != stage) |
1030 | continue; | |
12676608 LT |
1031 | if (p->name[len] != '/') |
1032 | continue; | |
7a51ed66 | 1033 | if (p->ce_flags & CE_REMOVE) |
21cd8d00 | 1034 | continue; |
12676608 LT |
1035 | retval = -1; |
1036 | if (!ok_to_replace) | |
1037 | break; | |
4aab5b46 | 1038 | remove_index_entry_at(istate, --pos); |
0f1e4f04 | 1039 | } |
12676608 LT |
1040 | return retval; |
1041 | } | |
0f1e4f04 | 1042 | |
a6db3fbb JH |
1043 | |
1044 | /* | |
1045 | * Like strcmp(), but also return the offset of the first change. | |
1046 | * If strings are equal, return the length. | |
1047 | */ | |
1048 | int strcmp_offset(const char *s1, const char *s2, size_t *first_change) | |
1049 | { | |
1050 | size_t k; | |
1051 | ||
1052 | if (!first_change) | |
1053 | return strcmp(s1, s2); | |
1054 | ||
1055 | for (k = 0; s1[k] == s2[k]; k++) | |
1056 | if (s1[k] == '\0') | |
1057 | break; | |
1058 | ||
1059 | *first_change = k; | |
1060 | return (unsigned char)s1[k] - (unsigned char)s2[k]; | |
1061 | } | |
1062 | ||
12676608 LT |
1063 | /* |
1064 | * Do we have another file with a pathname that is a proper | |
1065 | * subset of the name we're trying to add? | |
06b6d81b JH |
1066 | * |
1067 | * That is, is there another file in the index with a path | |
1068 | * that matches a sub-directory in the given entry? | |
12676608 | 1069 | */ |
4aab5b46 JH |
1070 | static int has_dir_name(struct index_state *istate, |
1071 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
12676608 LT |
1072 | { |
1073 | int retval = 0; | |
b155725d | 1074 | int stage = ce_stage(ce); |
12676608 LT |
1075 | const char *name = ce->name; |
1076 | const char *slash = name + ce_namelen(ce); | |
06b6d81b JH |
1077 | size_t len_eq_last; |
1078 | int cmp_last = 0; | |
1079 | ||
1080 | /* | |
1081 | * We are frequently called during an iteration on a sorted | |
1082 | * list of pathnames and while building a new index. Therefore, | |
1083 | * there is a high probability that this entry will eventually | |
1084 | * be appended to the index, rather than inserted in the middle. | |
1085 | * If we can confirm that, we can avoid binary searches on the | |
1086 | * components of the pathname. | |
1087 | * | |
1088 | * Compare the entry's full path with the last path in the index. | |
1089 | */ | |
1090 | if (istate->cache_nr > 0) { | |
1091 | cmp_last = strcmp_offset(name, | |
1092 | istate->cache[istate->cache_nr - 1]->name, | |
1093 | &len_eq_last); | |
1094 | if (cmp_last > 0) { | |
1095 | if (len_eq_last == 0) { | |
1096 | /* | |
1097 | * The entry sorts AFTER the last one in the | |
1098 | * index and their paths have no common prefix, | |
1099 | * so there cannot be a F/D conflict. | |
1100 | */ | |
1101 | return retval; | |
1102 | } else { | |
1103 | /* | |
1104 | * The entry sorts AFTER the last one in the | |
1105 | * index, but has a common prefix. Fall through | |
1106 | * to the loop below to disect the entry's path | |
1107 | * and see where the difference is. | |
1108 | */ | |
1109 | } | |
1110 | } else if (cmp_last == 0) { | |
1111 | /* | |
1112 | * The entry exactly matches the last one in the | |
1113 | * index, but because of multiple stage and CE_REMOVE | |
1114 | * items, we fall through and let the regular search | |
1115 | * code handle it. | |
1116 | */ | |
1117 | } | |
1118 | } | |
0f1e4f04 | 1119 | |
12676608 | 1120 | for (;;) { |
b986df5c | 1121 | size_t len; |
0f1e4f04 | 1122 | |
12676608 LT |
1123 | for (;;) { |
1124 | if (*--slash == '/') | |
1125 | break; | |
1126 | if (slash <= ce->name) | |
1127 | return retval; | |
1128 | } | |
1129 | len = slash - name; | |
0f1e4f04 | 1130 | |
b986df5c JH |
1131 | if (cmp_last > 0) { |
1132 | /* | |
1133 | * (len + 1) is a directory boundary (including | |
1134 | * the trailing slash). And since the loop is | |
1135 | * decrementing "slash", the first iteration is | |
1136 | * the longest directory prefix; subsequent | |
1137 | * iterations consider parent directories. | |
1138 | */ | |
1139 | ||
1140 | if (len + 1 <= len_eq_last) { | |
1141 | /* | |
1142 | * The directory prefix (including the trailing | |
1143 | * slash) also appears as a prefix in the last | |
1144 | * entry, so the remainder cannot collide (because | |
1145 | * strcmp said the whole path was greater). | |
1146 | * | |
1147 | * EQ: last: xxx/A | |
1148 | * this: xxx/B | |
1149 | * | |
1150 | * LT: last: xxx/file_A | |
1151 | * this: xxx/file_B | |
1152 | */ | |
1153 | return retval; | |
1154 | } | |
1155 | ||
1156 | if (len > len_eq_last) { | |
1157 | /* | |
1158 | * This part of the directory prefix (excluding | |
1159 | * the trailing slash) is longer than the known | |
1160 | * equal portions, so this sub-directory cannot | |
1161 | * collide with a file. | |
1162 | * | |
1163 | * GT: last: xxxA | |
1164 | * this: xxxB/file | |
1165 | */ | |
1166 | return retval; | |
1167 | } | |
1168 | ||
b986df5c JH |
1169 | /* |
1170 | * This is a possible collision. Fall through and | |
1171 | * let the regular search code handle it. | |
1172 | * | |
1173 | * last: xxx | |
1174 | * this: xxx/file | |
1175 | */ | |
1176 | } | |
1177 | ||
20ec2d03 | 1178 | pos = index_name_stage_pos(istate, name, len, stage, EXPAND_SPARSE); |
12676608 | 1179 | if (pos >= 0) { |
21cd8d00 JH |
1180 | /* |
1181 | * Found one, but not so fast. This could | |
1182 | * be a marker that says "I was here, but | |
1183 | * I am being removed". Such an entry is | |
1184 | * not a part of the resulting tree, and | |
1185 | * it is Ok to have a directory at the same | |
1186 | * path. | |
1187 | */ | |
077c48df | 1188 | if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) { |
21cd8d00 JH |
1189 | retval = -1; |
1190 | if (!ok_to_replace) | |
1191 | break; | |
4aab5b46 | 1192 | remove_index_entry_at(istate, pos); |
21cd8d00 JH |
1193 | continue; |
1194 | } | |
12676608 | 1195 | } |
21cd8d00 JH |
1196 | else |
1197 | pos = -pos-1; | |
12676608 LT |
1198 | |
1199 | /* | |
1200 | * Trivial optimization: if we find an entry that | |
1201 | * already matches the sub-directory, then we know | |
b155725d | 1202 | * we're ok, and we can exit. |
12676608 | 1203 | */ |
4aab5b46 JH |
1204 | while (pos < istate->cache_nr) { |
1205 | struct cache_entry *p = istate->cache[pos]; | |
b155725d JH |
1206 | if ((ce_namelen(p) <= len) || |
1207 | (p->name[len] != '/') || | |
1208 | memcmp(p->name, name, len)) | |
1209 | break; /* not our subdirectory */ | |
077c48df JH |
1210 | if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE)) |
1211 | /* | |
1212 | * p is at the same stage as our entry, and | |
b155725d JH |
1213 | * is a subdirectory of what we are looking |
1214 | * at, so we cannot have conflicts at our | |
1215 | * level or anything shorter. | |
1216 | */ | |
1217 | return retval; | |
1218 | pos++; | |
192268c1 | 1219 | } |
0f1e4f04 | 1220 | } |
12676608 LT |
1221 | return retval; |
1222 | } | |
1223 | ||
1224 | /* We may be in a situation where we already have path/file and path | |
1225 | * is being added, or we already have path and path/file is being | |
1226 | * added. Either one would result in a nonsense tree that has path | |
1227 | * twice when git-write-tree tries to write it out. Prevent it. | |
a6080a0a | 1228 | * |
12676608 LT |
1229 | * If ok-to-replace is specified, we remove the conflicting entries |
1230 | * from the cache so the caller should recompute the insert position. | |
1231 | * When this happens, we return non-zero. | |
1232 | */ | |
4aab5b46 JH |
1233 | static int check_file_directory_conflict(struct index_state *istate, |
1234 | const struct cache_entry *ce, | |
1235 | int pos, int ok_to_replace) | |
12676608 | 1236 | { |
21cd8d00 JH |
1237 | int retval; |
1238 | ||
1239 | /* | |
1240 | * When ce is an "I am going away" entry, we allow it to be added | |
1241 | */ | |
7a51ed66 | 1242 | if (ce->ce_flags & CE_REMOVE) |
21cd8d00 JH |
1243 | return 0; |
1244 | ||
12676608 LT |
1245 | /* |
1246 | * We check if the path is a sub-path of a subsequent pathname | |
1247 | * first, since removing those will not change the position | |
21cd8d00 | 1248 | * in the array. |
12676608 | 1249 | */ |
4aab5b46 | 1250 | retval = has_file_name(istate, ce, pos, ok_to_replace); |
21cd8d00 | 1251 | |
12676608 LT |
1252 | /* |
1253 | * Then check if the path might have a clashing sub-directory | |
1254 | * before it. | |
1255 | */ | |
4aab5b46 | 1256 | return retval + has_dir_name(istate, ce, pos, ok_to_replace); |
0f1e4f04 JH |
1257 | } |
1258 | ||
af3785dc | 1259 | static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option) |
197ee8c9 LT |
1260 | { |
1261 | int pos; | |
192268c1 JH |
1262 | int ok_to_add = option & ADD_CACHE_OK_TO_ADD; |
1263 | int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE; | |
b155725d | 1264 | int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK; |
39425819 | 1265 | int new_only = option & ADD_CACHE_NEW_ONLY; |
5f73076c | 1266 | |
e5494631 JH |
1267 | /* |
1268 | * If this entry's path sorts after the last entry in the index, | |
1269 | * we can avoid searching for it. | |
1270 | */ | |
1271 | if (istate->cache_nr > 0 && | |
1272 | strcmp(ce->name, istate->cache[istate->cache_nr - 1]->name) > 0) | |
c097b95a | 1273 | pos = index_pos_to_insert_pos(istate->cache_nr); |
e5494631 | 1274 | else |
20ec2d03 | 1275 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce), EXPAND_SPARSE); |
197ee8c9 | 1276 | |
c35e9f5e VD |
1277 | /* |
1278 | * Cache tree path should be invalidated only after index_name_stage_pos, | |
1279 | * in case it expands a sparse index. | |
1280 | */ | |
1281 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) | |
1282 | cache_tree_invalidate_path(istate, ce->name); | |
1283 | ||
3e09cdfd | 1284 | /* existing match? Just replace it. */ |
76e7f4ec | 1285 | if (pos >= 0) { |
39425819 JH |
1286 | if (!new_only) |
1287 | replace_index_entry(istate, pos, ce); | |
197ee8c9 LT |
1288 | return 0; |
1289 | } | |
76e7f4ec | 1290 | pos = -pos-1; |
197ee8c9 | 1291 | |
ffcc9ba7 NTND |
1292 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1293 | untracked_cache_add_to_index(istate, ce->name); | |
e931371a | 1294 | |
7b937ca3 LT |
1295 | /* |
1296 | * Inserting a merged entry ("stage 0") into the index | |
1297 | * will always replace all non-merged entries.. | |
1298 | */ | |
4aab5b46 JH |
1299 | if (pos < istate->cache_nr && ce_stage(ce) == 0) { |
1300 | while (ce_same_name(istate->cache[pos], ce)) { | |
7b937ca3 | 1301 | ok_to_add = 1; |
4aab5b46 | 1302 | if (!remove_index_entry_at(istate, pos)) |
7b937ca3 LT |
1303 | break; |
1304 | } | |
1305 | } | |
1306 | ||
121481ab LT |
1307 | if (!ok_to_add) |
1308 | return -1; | |
c8ad9d04 | 1309 | if (verify_path_internal(ce->name, ce->ce_mode) == PATH_INVALID) |
9d0a9e90 | 1310 | return error(_("invalid path '%s'"), ce->name); |
121481ab | 1311 | |
3e09cdfd | 1312 | if (!skip_df_check && |
4aab5b46 | 1313 | check_file_directory_conflict(istate, ce, pos, ok_to_replace)) { |
192268c1 | 1314 | if (!ok_to_replace) |
9d0a9e90 | 1315 | return error(_("'%s' appears as both a file and as a directory"), |
4aab5b46 | 1316 | ce->name); |
20ec2d03 | 1317 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce), EXPAND_SPARSE); |
192268c1 JH |
1318 | pos = -pos-1; |
1319 | } | |
af3785dc JH |
1320 | return pos + 1; |
1321 | } | |
1322 | ||
1323 | int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option) | |
1324 | { | |
1325 | int pos; | |
1326 | ||
1327 | if (option & ADD_CACHE_JUST_APPEND) | |
1328 | pos = istate->cache_nr; | |
1329 | else { | |
1330 | int ret; | |
1331 | ret = add_index_entry_with_check(istate, ce, option); | |
1332 | if (ret <= 0) | |
1333 | return ret; | |
1334 | pos = ret - 1; | |
1335 | } | |
0f1e4f04 | 1336 | |
197ee8c9 | 1337 | /* Make sure the array is big enough .. */ |
999f5660 | 1338 | ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc); |
197ee8c9 LT |
1339 | |
1340 | /* Add it in.. */ | |
4aab5b46 | 1341 | istate->cache_nr++; |
af3785dc | 1342 | if (istate->cache_nr > pos + 1) |
f919ffeb SG |
1343 | MOVE_ARRAY(istate->cache + pos + 1, istate->cache + pos, |
1344 | istate->cache_nr - pos - 1); | |
cf558704 | 1345 | set_index_entry(istate, pos, ce); |
e636a7b4 | 1346 | istate->cache_changed |= CE_ENTRY_ADDED; |
197ee8c9 LT |
1347 | return 0; |
1348 | } | |
1349 | ||
405e5b2f LT |
1350 | /* |
1351 | * "refresh" does not calculate a new sha1 file or bring the | |
1352 | * cache up-to-date for mode/content changes. But what it | |
1353 | * _does_ do is to "re-match" the stat information of a file | |
1354 | * with the cache, so that you can refresh the cache for a | |
1355 | * file that hasn't been changed but where the stat entry is | |
1356 | * out of date. | |
1357 | * | |
1358 | * For example, you'd want to do this after doing a "git-read-tree", | |
1359 | * to link up the stat cache details with the proper files. | |
1360 | */ | |
4aab5b46 | 1361 | static struct cache_entry *refresh_cache_ent(struct index_state *istate, |
4bd5b7da | 1362 | struct cache_entry *ce, |
d05e6970 | 1363 | unsigned int options, int *err, |
a98e0f2d | 1364 | int *changed_ret, |
15268d12 JH |
1365 | int *t2_did_lstat, |
1366 | int *t2_did_scan) | |
405e5b2f LT |
1367 | { |
1368 | struct stat st; | |
1369 | struct cache_entry *updated; | |
a849735b | 1370 | int changed; |
25762726 | 1371 | int refresh = options & CE_MATCH_REFRESH; |
4bd5b7da | 1372 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 1373 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
2e2e7ec1 | 1374 | int ignore_missing = options & CE_MATCH_IGNORE_MISSING; |
883e248b | 1375 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
405e5b2f | 1376 | |
25762726 | 1377 | if (!refresh || ce_uptodate(ce)) |
eadb5831 JH |
1378 | return ce; |
1379 | ||
883e248b BP |
1380 | if (!ignore_fsmonitor) |
1381 | refresh_fsmonitor(istate); | |
aa9349d4 | 1382 | /* |
56cac48c NTND |
1383 | * CE_VALID or CE_SKIP_WORKTREE means the user promised us |
1384 | * that the change to the work tree does not matter and told | |
1385 | * us not to worry. | |
aa9349d4 | 1386 | */ |
56cac48c NTND |
1387 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) { |
1388 | ce_mark_uptodate(ce); | |
1389 | return ce; | |
1390 | } | |
aa9349d4 MSO |
1391 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) { |
1392 | ce_mark_uptodate(ce); | |
1393 | return ce; | |
1394 | } | |
883e248b BP |
1395 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) { |
1396 | ce_mark_uptodate(ce); | |
1397 | return ce; | |
1398 | } | |
aa9349d4 | 1399 | |
ccad42d4 RS |
1400 | if (has_symlink_leading_path(ce->name, ce_namelen(ce))) { |
1401 | if (ignore_missing) | |
1402 | return ce; | |
1403 | if (err) | |
1404 | *err = ENOENT; | |
1405 | return NULL; | |
1406 | } | |
1407 | ||
a98e0f2d JH |
1408 | if (t2_did_lstat) |
1409 | *t2_did_lstat = 1; | |
8fd2cb40 | 1410 | if (lstat(ce->name, &st) < 0) { |
2e2e7ec1 BK |
1411 | if (ignore_missing && errno == ENOENT) |
1412 | return ce; | |
ec0cc704 JH |
1413 | if (err) |
1414 | *err = errno; | |
8fd2cb40 JS |
1415 | return NULL; |
1416 | } | |
405e5b2f | 1417 | |
4bd5b7da | 1418 | changed = ie_match_stat(istate, ce, &st, options); |
d05e6970 JK |
1419 | if (changed_ret) |
1420 | *changed_ret = changed; | |
405e5b2f | 1421 | if (!changed) { |
4bd5b7da JH |
1422 | /* |
1423 | * The path is unchanged. If we were told to ignore | |
1424 | * valid bit, then we did the actual stat check and | |
1425 | * found that the entry is unmodified. If the entry | |
1426 | * is not marked VALID, this is the place to mark it | |
1427 | * valid again, under "assume unchanged" mode. | |
1428 | */ | |
1429 | if (ignore_valid && assume_unchanged && | |
7a51ed66 | 1430 | !(ce->ce_flags & CE_VALID)) |
405e5b2f | 1431 | ; /* mark this one VALID again */ |
eadb5831 JH |
1432 | else { |
1433 | /* | |
1434 | * We do not mark the index itself "modified" | |
1435 | * because CE_UPTODATE flag is in-core only; | |
1436 | * we are not going to write this change out. | |
1437 | */ | |
883e248b | 1438 | if (!S_ISGITLINK(ce->ce_mode)) { |
125fd984 | 1439 | ce_mark_uptodate(ce); |
b5a81697 | 1440 | mark_fsmonitor_valid(istate, ce); |
883e248b | 1441 | } |
8fd2cb40 | 1442 | return ce; |
eadb5831 | 1443 | } |
405e5b2f LT |
1444 | } |
1445 | ||
15268d12 JH |
1446 | if (t2_did_scan) |
1447 | *t2_did_scan = 1; | |
4bd5b7da | 1448 | if (ie_modified(istate, ce, &st, options)) { |
ec0cc704 JH |
1449 | if (err) |
1450 | *err = EINVAL; | |
8fd2cb40 JS |
1451 | return NULL; |
1452 | } | |
405e5b2f | 1453 | |
a849735b | 1454 | updated = make_empty_cache_entry(istate, ce_namelen(ce)); |
0e267b7a BP |
1455 | copy_cache_entry(updated, ce); |
1456 | memcpy(updated->name, ce->name, ce->ce_namelen + 1); | |
d4c0a3ac | 1457 | fill_stat_cache_info(istate, updated, &st); |
4bd5b7da JH |
1458 | /* |
1459 | * If ignore_valid is not set, we should leave CE_VALID bit | |
1460 | * alone. Otherwise, paths marked with --no-assume-unchanged | |
1461 | * (i.e. things to be edited) will reacquire CE_VALID bit | |
1462 | * automatically, which is not really what we want. | |
405e5b2f | 1463 | */ |
4bd5b7da | 1464 | if (!ignore_valid && assume_unchanged && |
7a51ed66 LT |
1465 | !(ce->ce_flags & CE_VALID)) |
1466 | updated->ce_flags &= ~CE_VALID; | |
405e5b2f | 1467 | |
e636a7b4 | 1468 | /* istate->cache_changed is updated in the caller */ |
405e5b2f LT |
1469 | return updated; |
1470 | } | |
1471 | ||
3deffc52 | 1472 | static void show_file(const char * fmt, const char * name, int in_porcelain, |
046613c5 | 1473 | int * first, const char *header_msg) |
3deffc52 MM |
1474 | { |
1475 | if (in_porcelain && *first && header_msg) { | |
1476 | printf("%s\n", header_msg); | |
cd2b8ae9 | 1477 | *first = 0; |
3deffc52 MM |
1478 | } |
1479 | printf(fmt, name); | |
1480 | } | |
1481 | ||
22184497 TG |
1482 | int repo_refresh_and_write_index(struct repository *repo, |
1483 | unsigned int refresh_flags, | |
1484 | unsigned int write_flags, | |
1485 | int gentle, | |
1486 | const struct pathspec *pathspec, | |
1487 | char *seen, const char *header_msg) | |
1488 | { | |
1489 | struct lock_file lock_file = LOCK_INIT; | |
1490 | int fd, ret = 0; | |
1491 | ||
1492 | fd = repo_hold_locked_index(repo, &lock_file, 0); | |
1493 | if (!gentle && fd < 0) | |
1494 | return -1; | |
1495 | if (refresh_index(repo->index, refresh_flags, pathspec, seen, header_msg)) | |
1496 | ret = 1; | |
1497 | if (0 <= fd && write_locked_index(repo->index, &lock_file, COMMIT_LOCK | write_flags)) | |
1498 | ret = -1; | |
1499 | return ret; | |
1500 | } | |
1501 | ||
1502 | ||
9b2d6149 NTND |
1503 | int refresh_index(struct index_state *istate, unsigned int flags, |
1504 | const struct pathspec *pathspec, | |
046613c5 | 1505 | char *seen, const char *header_msg) |
405e5b2f LT |
1506 | { |
1507 | int i; | |
1508 | int has_errors = 0; | |
1509 | int really = (flags & REFRESH_REALLY) != 0; | |
1510 | int allow_unmerged = (flags & REFRESH_UNMERGED) != 0; | |
1511 | int quiet = (flags & REFRESH_QUIET) != 0; | |
1512 | int not_new = (flags & REFRESH_IGNORE_MISSING) != 0; | |
5fdeacb0 | 1513 | int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0; |
b243012c | 1514 | int ignore_skip_worktree = (flags & REFRESH_IGNORE_SKIP_WORKTREE) != 0; |
3deffc52 MM |
1515 | int first = 1; |
1516 | int in_porcelain = (flags & REFRESH_IN_PORCELAIN); | |
25762726 BK |
1517 | unsigned int options = (CE_MATCH_REFRESH | |
1518 | (really ? CE_MATCH_IGNORE_VALID : 0) | | |
2e2e7ec1 | 1519 | (not_new ? CE_MATCH_IGNORE_MISSING : 0)); |
4bd4e730 | 1520 | const char *modified_fmt; |
73b7eae6 JK |
1521 | const char *deleted_fmt; |
1522 | const char *typechange_fmt; | |
1523 | const char *added_fmt; | |
4bd4e730 | 1524 | const char *unmerged_fmt; |
ae9af122 | 1525 | struct progress *progress = NULL; |
a98e0f2d | 1526 | int t2_sum_lstat = 0; |
15268d12 | 1527 | int t2_sum_scan = 0; |
ae9af122 NTND |
1528 | |
1529 | if (flags & REFRESH_PROGRESS && isatty(2)) | |
1530 | progress = start_delayed_progress(_("Refresh index"), | |
1531 | istate->cache_nr); | |
405e5b2f | 1532 | |
c46c406a | 1533 | trace_performance_enter(); |
a71806a7 NTND |
1534 | modified_fmt = in_porcelain ? "M\t%s\n" : "%s: needs update\n"; |
1535 | deleted_fmt = in_porcelain ? "D\t%s\n" : "%s: needs update\n"; | |
1536 | typechange_fmt = in_porcelain ? "T\t%s\n" : "%s: needs update\n"; | |
1537 | added_fmt = in_porcelain ? "A\t%s\n" : "%s: needs update\n"; | |
1538 | unmerged_fmt = in_porcelain ? "U\t%s\n" : "%s: needs merge\n"; | |
99ce720c BP |
1539 | /* |
1540 | * Use the multi-threaded preload_index() to refresh most of the | |
1541 | * cache entries quickly then in the single threaded loop below, | |
1542 | * we only have to do the special cases that are left. | |
1543 | */ | |
1544 | preload_index(istate, pathspec, 0); | |
a98e0f2d | 1545 | trace2_region_enter("index", "refresh", NULL); |
d76723ee | 1546 | |
4aab5b46 | 1547 | for (i = 0; i < istate->cache_nr; i++) { |
285c2e25 | 1548 | struct cache_entry *ce, *new_entry; |
ec0cc704 | 1549 | int cache_errno = 0; |
73b7eae6 | 1550 | int changed = 0; |
3d1f148c | 1551 | int filtered = 0; |
a98e0f2d | 1552 | int t2_did_lstat = 0; |
15268d12 | 1553 | int t2_did_scan = 0; |
ec0cc704 | 1554 | |
4aab5b46 | 1555 | ce = istate->cache[i]; |
5fdeacb0 JS |
1556 | if (ignore_submodules && S_ISGITLINK(ce->ce_mode)) |
1557 | continue; | |
b243012c MT |
1558 | if (ignore_skip_worktree && ce_skip_worktree(ce)) |
1559 | continue; | |
5fdeacb0 | 1560 | |
d76723ee DS |
1561 | /* |
1562 | * If this entry is a sparse directory, then there isn't | |
1563 | * any stat() information to update. Ignore the entry. | |
1564 | */ | |
1565 | if (S_ISSPARSEDIR(ce->ce_mode)) | |
1566 | continue; | |
1567 | ||
d7b665c3 | 1568 | if (pathspec && !ce_path_match(istate, ce, pathspec, seen)) |
3d1f148c JH |
1569 | filtered = 1; |
1570 | ||
405e5b2f | 1571 | if (ce_stage(ce)) { |
4aab5b46 JH |
1572 | while ((i < istate->cache_nr) && |
1573 | ! strcmp(istate->cache[i]->name, ce->name)) | |
405e5b2f LT |
1574 | i++; |
1575 | i--; | |
1576 | if (allow_unmerged) | |
1577 | continue; | |
3d1f148c JH |
1578 | if (!filtered) |
1579 | show_file(unmerged_fmt, ce->name, in_porcelain, | |
1580 | &first, header_msg); | |
405e5b2f LT |
1581 | has_errors = 1; |
1582 | continue; | |
1583 | } | |
1584 | ||
3d1f148c | 1585 | if (filtered) |
d616813d AJ |
1586 | continue; |
1587 | ||
a98e0f2d JH |
1588 | new_entry = refresh_cache_ent(istate, ce, options, |
1589 | &cache_errno, &changed, | |
15268d12 | 1590 | &t2_did_lstat, &t2_did_scan); |
a98e0f2d | 1591 | t2_sum_lstat += t2_did_lstat; |
15268d12 | 1592 | t2_sum_scan += t2_did_scan; |
285c2e25 | 1593 | if (new_entry == ce) |
405e5b2f | 1594 | continue; |
b7b793d1 | 1595 | display_progress(progress, i); |
285c2e25 | 1596 | if (!new_entry) { |
73b7eae6 JK |
1597 | const char *fmt; |
1598 | ||
8fd2cb40 | 1599 | if (really && cache_errno == EINVAL) { |
405e5b2f LT |
1600 | /* If we are doing --really-refresh that |
1601 | * means the index is not valid anymore. | |
1602 | */ | |
7a51ed66 | 1603 | ce->ce_flags &= ~CE_VALID; |
078a58e8 | 1604 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 1605 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 1606 | istate->cache_changed |= CE_ENTRY_CHANGED; |
405e5b2f LT |
1607 | } |
1608 | if (quiet) | |
1609 | continue; | |
73b7eae6 JK |
1610 | |
1611 | if (cache_errno == ENOENT) | |
1612 | fmt = deleted_fmt; | |
895ff3b2 | 1613 | else if (ce_intent_to_add(ce)) |
73b7eae6 JK |
1614 | fmt = added_fmt; /* must be before other checks */ |
1615 | else if (changed & TYPE_CHANGED) | |
1616 | fmt = typechange_fmt; | |
1617 | else | |
1618 | fmt = modified_fmt; | |
1619 | show_file(fmt, | |
1620 | ce->name, in_porcelain, &first, header_msg); | |
405e5b2f LT |
1621 | has_errors = 1; |
1622 | continue; | |
1623 | } | |
cf558704 | 1624 | |
285c2e25 | 1625 | replace_index_entry(istate, i, new_entry); |
405e5b2f | 1626 | } |
a98e0f2d | 1627 | trace2_data_intmax("index", NULL, "refresh/sum_lstat", t2_sum_lstat); |
15268d12 | 1628 | trace2_data_intmax("index", NULL, "refresh/sum_scan", t2_sum_scan); |
a98e0f2d | 1629 | trace2_region_leave("index", "refresh", NULL); |
b7b793d1 ÆAB |
1630 | display_progress(progress, istate->cache_nr); |
1631 | stop_progress(&progress); | |
c46c406a | 1632 | trace_performance_leave("refresh index"); |
405e5b2f LT |
1633 | return has_errors; |
1634 | } | |
1635 | ||
768d7965 JM |
1636 | struct cache_entry *refresh_cache_entry(struct index_state *istate, |
1637 | struct cache_entry *ce, | |
1638 | unsigned int options) | |
ec0cc704 | 1639 | { |
15268d12 | 1640 | return refresh_cache_ent(istate, ce, options, NULL, NULL, NULL, NULL); |
ec0cc704 JH |
1641 | } |
1642 | ||
db3b313c JH |
1643 | |
1644 | /***************************************************************** | |
1645 | * Index File I/O | |
1646 | *****************************************************************/ | |
1647 | ||
9d227781 JH |
1648 | #define INDEX_FORMAT_DEFAULT 3 |
1649 | ||
7211b9e7 | 1650 | static unsigned int get_index_format_default(struct repository *r) |
136347d7 TG |
1651 | { |
1652 | char *envversion = getenv("GIT_INDEX_VERSION"); | |
3c09d684 TG |
1653 | char *endp; |
1654 | unsigned int version = INDEX_FORMAT_DEFAULT; | |
1655 | ||
136347d7 | 1656 | if (!envversion) { |
7211b9e7 DS |
1657 | prepare_repo_settings(r); |
1658 | ||
1659 | if (r->settings.index_version >= 0) | |
1660 | version = r->settings.index_version; | |
3c09d684 TG |
1661 | if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { |
1662 | warning(_("index.version set, but the value is invalid.\n" | |
136347d7 | 1663 | "Using version %i"), INDEX_FORMAT_DEFAULT); |
3c09d684 | 1664 | return INDEX_FORMAT_DEFAULT; |
136347d7 TG |
1665 | } |
1666 | return version; | |
1667 | } | |
3c09d684 TG |
1668 | |
1669 | version = strtoul(envversion, &endp, 10); | |
1670 | if (*endp || | |
1671 | version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { | |
1672 | warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n" | |
1673 | "Using version %i"), INDEX_FORMAT_DEFAULT); | |
1674 | version = INDEX_FORMAT_DEFAULT; | |
1675 | } | |
1676 | return version; | |
136347d7 TG |
1677 | } |
1678 | ||
db3b313c JH |
1679 | /* |
1680 | * dev/ino/uid/gid/size are also just tracked to the low 32 bits | |
1681 | * Again - this is just a (very strong in practice) heuristic that | |
1682 | * the inode hasn't changed. | |
1683 | * | |
1684 | * We save the fields in big-endian order to allow using the | |
1685 | * index file over NFS transparently. | |
1686 | */ | |
1687 | struct ondisk_cache_entry { | |
1688 | struct cache_time ctime; | |
1689 | struct cache_time mtime; | |
7800c1eb TG |
1690 | uint32_t dev; |
1691 | uint32_t ino; | |
1692 | uint32_t mode; | |
1693 | uint32_t uid; | |
1694 | uint32_t gid; | |
1695 | uint32_t size; | |
575fa8a3 | 1696 | /* |
1697 | * unsigned char hash[hashsz]; | |
1698 | * uint16_t flags; | |
1699 | * if (flags & CE_EXTENDED) | |
1700 | * uint16_t flags2; | |
1701 | */ | |
1702 | unsigned char data[GIT_MAX_RAWSZ + 2 * sizeof(uint16_t)]; | |
1703 | char name[FLEX_ARRAY]; | |
db3b313c JH |
1704 | }; |
1705 | ||
6c9cd161 | 1706 | /* These are only used for v3 or lower */ |
ce012deb | 1707 | #define align_padding_size(size, len) ((size + (len) + 8) & ~7) - (size + len) |
575fa8a3 | 1708 | #define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,data) + (len) + 8) & ~7) |
db3b313c | 1709 | #define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len) |
575fa8a3 | 1710 | #define ondisk_data_size(flags, len) (the_hash_algo->rawsz + \ |
1711 | ((flags & CE_EXTENDED) ? 2 : 1) * sizeof(uint16_t) + len) | |
1712 | #define ondisk_data_size_max(len) (ondisk_data_size(CE_EXTENDED, len)) | |
1713 | #define ondisk_ce_size(ce) (ondisk_cache_entry_size(ondisk_data_size((ce)->ce_flags, ce_namelen(ce)))) | |
db3b313c | 1714 | |
a33fc72f JH |
1715 | /* Allow fsck to force verification of the index checksum. */ |
1716 | int verify_index_checksum; | |
1717 | ||
00ec50e5 BP |
1718 | /* Allow fsck to force verification of the cache entry order. */ |
1719 | int verify_ce_order; | |
1720 | ||
371ed0de | 1721 | static int verify_hdr(const struct cache_header *hdr, unsigned long size) |
e83c5163 | 1722 | { |
aab61359 | 1723 | git_hash_ctx c; |
1724 | unsigned char hash[GIT_MAX_RAWSZ]; | |
0136bac9 | 1725 | int hdr_version; |
ee1f0c24 DS |
1726 | unsigned char *start, *end; |
1727 | struct object_id oid; | |
e83c5163 | 1728 | |
ccc4feb5 | 1729 | if (hdr->hdr_signature != htonl(CACHE_SIGNATURE)) |
9d0a9e90 | 1730 | return error(_("bad signature 0x%08x"), hdr->hdr_signature); |
0136bac9 | 1731 | hdr_version = ntohl(hdr->hdr_version); |
b82a7b5b | 1732 | if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version) |
9d0a9e90 | 1733 | return error(_("bad index version %d"), hdr_version); |
a33fc72f JH |
1734 | |
1735 | if (!verify_index_checksum) | |
1736 | return 0; | |
1737 | ||
ee1f0c24 DS |
1738 | end = (unsigned char *)hdr + size; |
1739 | start = end - the_hash_algo->rawsz; | |
1740 | oidread(&oid, start); | |
1741 | if (oideq(&oid, null_oid())) | |
1742 | return 0; | |
1743 | ||
aab61359 | 1744 | the_hash_algo->init_fn(&c); |
1745 | the_hash_algo->update_fn(&c, hdr, size - the_hash_algo->rawsz); | |
1746 | the_hash_algo->final_fn(hash, &c); | |
ee1f0c24 | 1747 | if (!hasheq(hash, start)) |
9d0a9e90 | 1748 | return error(_("bad index file sha1 signature")); |
e83c5163 LT |
1749 | return 0; |
1750 | } | |
1751 | ||
4aab5b46 | 1752 | static int read_index_extension(struct index_state *istate, |
371ed0de | 1753 | const char *ext, const char *data, unsigned long sz) |
bad68ec9 JH |
1754 | { |
1755 | switch (CACHE_EXT(ext)) { | |
1756 | case CACHE_EXT_TREE: | |
4aab5b46 | 1757 | istate->cache_tree = cache_tree_read(data, sz); |
bad68ec9 | 1758 | break; |
cfc5789a JH |
1759 | case CACHE_EXT_RESOLVE_UNDO: |
1760 | istate->resolve_undo = resolve_undo_read(data, sz); | |
1761 | break; | |
5fc2fc8f NTND |
1762 | case CACHE_EXT_LINK: |
1763 | if (read_link_extension(istate, data, sz)) | |
1764 | return -1; | |
1765 | break; | |
f9e6c649 NTND |
1766 | case CACHE_EXT_UNTRACKED: |
1767 | istate->untracked = read_untracked_extension(data, sz); | |
1768 | break; | |
883e248b BP |
1769 | case CACHE_EXT_FSMONITOR: |
1770 | read_fsmonitor_extension(istate, data, sz); | |
1771 | break; | |
3b1d9e04 | 1772 | case CACHE_EXT_ENDOFINDEXENTRIES: |
3255089a | 1773 | case CACHE_EXT_INDEXENTRYOFFSETTABLE: |
3b1d9e04 BP |
1774 | /* already handled in do_read_index() */ |
1775 | break; | |
cd42415f DS |
1776 | case CACHE_EXT_SPARSE_DIRECTORIES: |
1777 | /* no content, only an indicator */ | |
9fadb373 | 1778 | istate->sparse_index = INDEX_COLLAPSED; |
cd42415f | 1779 | break; |
bad68ec9 JH |
1780 | default: |
1781 | if (*ext < 'A' || 'Z' < *ext) | |
9d0a9e90 | 1782 | return error(_("index uses %.4s extension, which we do not understand"), |
bad68ec9 | 1783 | ext); |
9d0a9e90 | 1784 | fprintf_ln(stderr, _("ignoring %.4s extension"), ext); |
bad68ec9 JH |
1785 | break; |
1786 | } | |
1787 | return 0; | |
1788 | } | |
1789 | ||
4a6ed30f VD |
1790 | /* |
1791 | * Parses the contents of the cache entry contained within the 'ondisk' buffer | |
1792 | * into a new incore 'cache_entry'. | |
1793 | * | |
1794 | * Note that 'char *ondisk' may not be aligned to a 4-byte address interval in | |
1795 | * index v4, so we cannot cast it to 'struct ondisk_cache_entry *' and access | |
1796 | * its members. Instead, we use the byte offsets of members within the struct to | |
1797 | * identify where 'get_be16()', 'get_be32()', and 'oidread()' (which can all | |
1798 | * read from an unaligned memory buffer) should read from the 'ondisk' buffer | |
1799 | * into the corresponding incore 'cache_entry' members. | |
1800 | */ | |
77ff1127 BP |
1801 | static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool, |
1802 | unsigned int version, | |
4a6ed30f | 1803 | const char *ondisk, |
6c9cd161 | 1804 | unsigned long *ent_size, |
252d079c | 1805 | const struct cache_entry *previous_ce) |
7a51ed66 | 1806 | { |
debed2a6 | 1807 | struct cache_entry *ce; |
7fec10b7 | 1808 | size_t len; |
06aaaa0b | 1809 | const char *name; |
575fa8a3 | 1810 | const unsigned hashsz = the_hash_algo->rawsz; |
4a6ed30f | 1811 | const char *flagsp = ondisk + offsetof(struct ondisk_cache_entry, data) + hashsz; |
debed2a6 | 1812 | unsigned int flags; |
f5c4a9af | 1813 | size_t copy_len = 0; |
252d079c NTND |
1814 | /* |
1815 | * Adjacent cache entries tend to share the leading paths, so it makes | |
1816 | * sense to only store the differences in later entries. In the v4 | |
1817 | * on-disk format of the index, each on-disk cache entry stores the | |
1818 | * number of bytes to be stripped from the end of the previous name, | |
1819 | * and the bytes to append to the result, to come up with its name. | |
1820 | */ | |
77ff1127 | 1821 | int expand_name_field = version == 4; |
7fec10b7 | 1822 | |
7a51ed66 | 1823 | /* On-disk flags are just 16 bits */ |
575fa8a3 | 1824 | flags = get_be16(flagsp); |
debed2a6 | 1825 | len = flags & CE_NAMEMASK; |
7fec10b7 | 1826 | |
debed2a6 | 1827 | if (flags & CE_EXTENDED) { |
06aaaa0b | 1828 | int extended_flags; |
4a6ed30f | 1829 | extended_flags = get_be16(flagsp + sizeof(uint16_t)) << 16; |
06aaaa0b NTND |
1830 | /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */ |
1831 | if (extended_flags & ~CE_EXTENDED_FLAGS) | |
9d0a9e90 | 1832 | die(_("unknown index entry format 0x%08x"), extended_flags); |
debed2a6 | 1833 | flags |= extended_flags; |
4a6ed30f | 1834 | name = (const char *)(flagsp + 2 * sizeof(uint16_t)); |
06aaaa0b NTND |
1835 | } |
1836 | else | |
4a6ed30f | 1837 | name = (const char *)(flagsp + sizeof(uint16_t)); |
06aaaa0b | 1838 | |
252d079c NTND |
1839 | if (expand_name_field) { |
1840 | const unsigned char *cp = (const unsigned char *)name; | |
1841 | size_t strip_len, previous_len; | |
6c9cd161 | 1842 | |
15beaaa3 | 1843 | /* If we're at the beginning of a block, ignore the previous name */ |
252d079c | 1844 | strip_len = decode_varint(&cp); |
77ff1127 BP |
1845 | if (previous_ce) { |
1846 | previous_len = previous_ce->ce_namelen; | |
1847 | if (previous_len < strip_len) | |
252d079c | 1848 | die(_("malformed name field in the index, near path '%s'"), |
77ff1127 BP |
1849 | previous_ce->name); |
1850 | copy_len = previous_len - strip_len; | |
252d079c | 1851 | } |
252d079c NTND |
1852 | name = (const char *)cp; |
1853 | } | |
6c9cd161 | 1854 | |
252d079c NTND |
1855 | if (len == CE_NAMEMASK) { |
1856 | len = strlen(name); | |
1857 | if (expand_name_field) | |
1858 | len += copy_len; | |
1859 | } | |
6c9cd161 | 1860 | |
77ff1127 | 1861 | ce = mem_pool__ce_alloc(ce_mem_pool, len); |
252d079c | 1862 | |
4a6ed30f VD |
1863 | /* |
1864 | * NEEDSWORK: using 'offsetof()' is cumbersome and should be replaced | |
1865 | * with something more akin to 'load_bitmap_entries_v1()'s use of | |
1866 | * 'read_be16'/'read_be32'. For consistency with the corresponding | |
1867 | * ondisk entry write function ('copy_cache_entry_to_ondisk()'), this | |
1868 | * should be done at the same time as removing references to | |
1869 | * 'ondisk_cache_entry' there. | |
1870 | */ | |
1871 | ce->ce_stat_data.sd_ctime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime) | |
1872 | + offsetof(struct cache_time, sec)); | |
1873 | ce->ce_stat_data.sd_mtime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime) | |
1874 | + offsetof(struct cache_time, sec)); | |
1875 | ce->ce_stat_data.sd_ctime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime) | |
1876 | + offsetof(struct cache_time, nsec)); | |
1877 | ce->ce_stat_data.sd_mtime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime) | |
1878 | + offsetof(struct cache_time, nsec)); | |
1879 | ce->ce_stat_data.sd_dev = get_be32(ondisk + offsetof(struct ondisk_cache_entry, dev)); | |
1880 | ce->ce_stat_data.sd_ino = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ino)); | |
1881 | ce->ce_mode = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mode)); | |
1882 | ce->ce_stat_data.sd_uid = get_be32(ondisk + offsetof(struct ondisk_cache_entry, uid)); | |
1883 | ce->ce_stat_data.sd_gid = get_be32(ondisk + offsetof(struct ondisk_cache_entry, gid)); | |
1884 | ce->ce_stat_data.sd_size = get_be32(ondisk + offsetof(struct ondisk_cache_entry, size)); | |
252d079c NTND |
1885 | ce->ce_flags = flags & ~CE_NAMEMASK; |
1886 | ce->ce_namelen = len; | |
1887 | ce->index = 0; | |
4a6ed30f | 1888 | oidread(&ce->oid, (const unsigned char *)ondisk + offsetof(struct ondisk_cache_entry, data)); |
6c9cd161 | 1889 | |
252d079c NTND |
1890 | if (expand_name_field) { |
1891 | if (copy_len) | |
1892 | memcpy(ce->name, previous_ce->name, copy_len); | |
1893 | memcpy(ce->name + copy_len, name, len + 1 - copy_len); | |
1894 | *ent_size = (name - ((char *)ondisk)) + len + 1 - copy_len; | |
1895 | } else { | |
1896 | memcpy(ce->name, name, len + 1); | |
1897 | *ent_size = ondisk_ce_size(ce); | |
6c9cd161 | 1898 | } |
debed2a6 | 1899 | return ce; |
cf558704 LT |
1900 | } |
1901 | ||
03f15a79 | 1902 | static void check_ce_order(struct index_state *istate) |
15999d0b | 1903 | { |
03f15a79 TG |
1904 | unsigned int i; |
1905 | ||
00ec50e5 BP |
1906 | if (!verify_ce_order) |
1907 | return; | |
1908 | ||
03f15a79 TG |
1909 | for (i = 1; i < istate->cache_nr; i++) { |
1910 | struct cache_entry *ce = istate->cache[i - 1]; | |
1911 | struct cache_entry *next_ce = istate->cache[i]; | |
1912 | int name_compare = strcmp(ce->name, next_ce->name); | |
1913 | ||
1914 | if (0 < name_compare) | |
9d0a9e90 | 1915 | die(_("unordered stage entries in index")); |
03f15a79 TG |
1916 | if (!name_compare) { |
1917 | if (!ce_stage(ce)) | |
9d0a9e90 | 1918 | die(_("multiple stage entries for merged file '%s'"), |
03f15a79 TG |
1919 | ce->name); |
1920 | if (ce_stage(ce) > ce_stage(next_ce)) | |
9d0a9e90 | 1921 | die(_("unordered stage entries for '%s'"), |
03f15a79 TG |
1922 | ce->name); |
1923 | } | |
15999d0b JSP |
1924 | } |
1925 | } | |
1926 | ||
435ec090 CC |
1927 | static void tweak_untracked_cache(struct index_state *istate) |
1928 | { | |
ad0fb659 DS |
1929 | struct repository *r = the_repository; |
1930 | ||
1931 | prepare_repo_settings(r); | |
1932 | ||
f1bee828 ÆAB |
1933 | switch (r->settings.core_untracked_cache) { |
1934 | case UNTRACKED_CACHE_REMOVE: | |
435ec090 | 1935 | remove_untracked_cache(istate); |
f1bee828 ÆAB |
1936 | break; |
1937 | case UNTRACKED_CACHE_WRITE: | |
ad0fb659 | 1938 | add_untracked_cache(istate); |
f1bee828 ÆAB |
1939 | break; |
1940 | case UNTRACKED_CACHE_KEEP: | |
3050b6df ÆAB |
1941 | /* |
1942 | * Either an explicit "core.untrackedCache=keep", the | |
1943 | * default if "core.untrackedCache" isn't configured, | |
1944 | * or a fallback on an unknown "core.untrackedCache" | |
1945 | * value. | |
1946 | */ | |
f1bee828 | 1947 | break; |
f1bee828 | 1948 | } |
435ec090 CC |
1949 | } |
1950 | ||
43925312 CC |
1951 | static void tweak_split_index(struct index_state *istate) |
1952 | { | |
1953 | switch (git_config_get_split_index()) { | |
1954 | case -1: /* unset: do nothing */ | |
1955 | break; | |
1956 | case 0: /* false */ | |
1957 | remove_split_index(istate); | |
1958 | break; | |
1959 | case 1: /* true */ | |
1960 | add_split_index(istate); | |
1961 | break; | |
1962 | default: /* unknown value: do nothing */ | |
1963 | break; | |
1964 | } | |
1965 | } | |
1966 | ||
435ec090 CC |
1967 | static void post_read_index_from(struct index_state *istate) |
1968 | { | |
1969 | check_ce_order(istate); | |
1970 | tweak_untracked_cache(istate); | |
43925312 | 1971 | tweak_split_index(istate); |
883e248b | 1972 | tweak_fsmonitor(istate); |
435ec090 CC |
1973 | } |
1974 | ||
8e72d675 JM |
1975 | static size_t estimate_cache_size_from_compressed(unsigned int entries) |
1976 | { | |
1977 | return entries * (sizeof(struct cache_entry) + CACHE_ENTRY_PATH_LENGTH); | |
1978 | } | |
1979 | ||
1980 | static size_t estimate_cache_size(size_t ondisk_size, unsigned int entries) | |
1981 | { | |
1982 | long per_entry = sizeof(struct cache_entry) - sizeof(struct ondisk_cache_entry); | |
1983 | ||
1984 | /* | |
1985 | * Account for potential alignment differences. | |
1986 | */ | |
c097b95a | 1987 | per_entry += align_padding_size(per_entry, 0); |
8e72d675 JM |
1988 | return ondisk_size + entries * per_entry; |
1989 | } | |
1990 | ||
3255089a BP |
1991 | struct index_entry_offset |
1992 | { | |
1993 | /* starting byte offset into index file, count of index entries in this block */ | |
1994 | int offset, nr; | |
1995 | }; | |
1996 | ||
1997 | struct index_entry_offset_table | |
1998 | { | |
1999 | int nr; | |
2000 | struct index_entry_offset entries[FLEX_ARRAY]; | |
2001 | }; | |
2002 | ||
3255089a BP |
2003 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset); |
2004 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot); | |
3255089a | 2005 | |
3b1d9e04 BP |
2006 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size); |
2007 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset); | |
2008 | ||
abb4bb83 BP |
2009 | struct load_index_extensions |
2010 | { | |
abb4bb83 | 2011 | pthread_t pthread; |
abb4bb83 BP |
2012 | struct index_state *istate; |
2013 | const char *mmap; | |
2014 | size_t mmap_size; | |
2015 | unsigned long src_offset; | |
2016 | }; | |
2017 | ||
2018 | static void *load_index_extensions(void *_data) | |
2019 | { | |
2020 | struct load_index_extensions *p = _data; | |
2021 | unsigned long src_offset = p->src_offset; | |
2022 | ||
2023 | while (src_offset <= p->mmap_size - the_hash_algo->rawsz - 8) { | |
2024 | /* After an array of active_nr index entries, | |
2025 | * there can be arbitrary number of extended | |
2026 | * sections, each of which is prefixed with | |
2027 | * extension name (4-byte) and section length | |
2028 | * in 4-byte network byte order. | |
2029 | */ | |
2030 | uint32_t extsize = get_be32(p->mmap + src_offset + 4); | |
2031 | if (read_index_extension(p->istate, | |
2032 | p->mmap + src_offset, | |
2033 | p->mmap + src_offset + 8, | |
2034 | extsize) < 0) { | |
2035 | munmap((void *)p->mmap, p->mmap_size); | |
2036 | die(_("index file corrupt")); | |
2037 | } | |
2038 | src_offset += 8; | |
2039 | src_offset += extsize; | |
2040 | } | |
2041 | ||
2042 | return NULL; | |
2043 | } | |
2044 | ||
77ff1127 BP |
2045 | /* |
2046 | * A helper function that will load the specified range of cache entries | |
2047 | * from the memory mapped file and add them to the given index. | |
2048 | */ | |
2049 | static unsigned long load_cache_entry_block(struct index_state *istate, | |
2050 | struct mem_pool *ce_mem_pool, int offset, int nr, const char *mmap, | |
2051 | unsigned long start_offset, const struct cache_entry *previous_ce) | |
2052 | { | |
2053 | int i; | |
2054 | unsigned long src_offset = start_offset; | |
2055 | ||
2056 | for (i = offset; i < offset + nr; i++) { | |
77ff1127 BP |
2057 | struct cache_entry *ce; |
2058 | unsigned long consumed; | |
2059 | ||
4a6ed30f VD |
2060 | ce = create_from_disk(ce_mem_pool, istate->version, |
2061 | mmap + src_offset, | |
2062 | &consumed, previous_ce); | |
77ff1127 BP |
2063 | set_index_entry(istate, i, ce); |
2064 | ||
2065 | src_offset += consumed; | |
2066 | previous_ce = ce; | |
2067 | } | |
2068 | return src_offset - start_offset; | |
2069 | } | |
2070 | ||
2071 | static unsigned long load_all_cache_entries(struct index_state *istate, | |
2072 | const char *mmap, size_t mmap_size, unsigned long src_offset) | |
2073 | { | |
2074 | unsigned long consumed; | |
2075 | ||
44c7e1a7 | 2076 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2077 | if (istate->version == 4) { |
44c7e1a7 | 2078 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2079 | estimate_cache_size_from_compressed(istate->cache_nr)); |
2080 | } else { | |
44c7e1a7 | 2081 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2082 | estimate_cache_size(mmap_size, istate->cache_nr)); |
2083 | } | |
2084 | ||
2085 | consumed = load_cache_entry_block(istate, istate->ce_mem_pool, | |
2086 | 0, istate->cache_nr, mmap, src_offset, NULL); | |
2087 | return consumed; | |
2088 | } | |
2089 | ||
3255089a BP |
2090 | /* |
2091 | * Mostly randomly chosen maximum thread counts: we | |
2092 | * cap the parallelism to online_cpus() threads, and we want | |
2093 | * to have at least 10000 cache entries per thread for it to | |
2094 | * be worth starting a thread. | |
2095 | */ | |
2096 | ||
2097 | #define THREAD_COST (10000) | |
2098 | ||
77ff1127 BP |
2099 | struct load_cache_entries_thread_data |
2100 | { | |
2101 | pthread_t pthread; | |
2102 | struct index_state *istate; | |
2103 | struct mem_pool *ce_mem_pool; | |
2104 | int offset; | |
2105 | const char *mmap; | |
2106 | struct index_entry_offset_table *ieot; | |
2107 | int ieot_start; /* starting index into the ieot array */ | |
2108 | int ieot_blocks; /* count of ieot entries to process */ | |
2109 | unsigned long consumed; /* return # of bytes in index file processed */ | |
2110 | }; | |
2111 | ||
2112 | /* | |
2113 | * A thread proc to run the load_cache_entries() computation | |
2114 | * across multiple background threads. | |
2115 | */ | |
2116 | static void *load_cache_entries_thread(void *_data) | |
2117 | { | |
2118 | struct load_cache_entries_thread_data *p = _data; | |
2119 | int i; | |
2120 | ||
2121 | /* iterate across all ieot blocks assigned to this thread */ | |
2122 | for (i = p->ieot_start; i < p->ieot_start + p->ieot_blocks; i++) { | |
2123 | p->consumed += load_cache_entry_block(p->istate, p->ce_mem_pool, | |
2124 | p->offset, p->ieot->entries[i].nr, p->mmap, p->ieot->entries[i].offset, NULL); | |
2125 | p->offset += p->ieot->entries[i].nr; | |
2126 | } | |
2127 | return NULL; | |
2128 | } | |
2129 | ||
2130 | static unsigned long load_cache_entries_threaded(struct index_state *istate, const char *mmap, size_t mmap_size, | |
7bd9631b | 2131 | int nr_threads, struct index_entry_offset_table *ieot) |
77ff1127 BP |
2132 | { |
2133 | int i, offset, ieot_blocks, ieot_start, err; | |
2134 | struct load_cache_entries_thread_data *data; | |
2135 | unsigned long consumed = 0; | |
2136 | ||
2137 | /* a little sanity checking */ | |
2138 | if (istate->name_hash_initialized) | |
2139 | BUG("the name hash isn't thread safe"); | |
2140 | ||
44c7e1a7 EN |
2141 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
2142 | mem_pool_init(istate->ce_mem_pool, 0); | |
77ff1127 BP |
2143 | |
2144 | /* ensure we have no more threads than we have blocks to process */ | |
2145 | if (nr_threads > ieot->nr) | |
2146 | nr_threads = ieot->nr; | |
ca56dadb | 2147 | CALLOC_ARRAY(data, nr_threads); |
77ff1127 BP |
2148 | |
2149 | offset = ieot_start = 0; | |
2150 | ieot_blocks = DIV_ROUND_UP(ieot->nr, nr_threads); | |
2151 | for (i = 0; i < nr_threads; i++) { | |
2152 | struct load_cache_entries_thread_data *p = &data[i]; | |
2153 | int nr, j; | |
2154 | ||
2155 | if (ieot_start + ieot_blocks > ieot->nr) | |
2156 | ieot_blocks = ieot->nr - ieot_start; | |
2157 | ||
2158 | p->istate = istate; | |
2159 | p->offset = offset; | |
2160 | p->mmap = mmap; | |
2161 | p->ieot = ieot; | |
2162 | p->ieot_start = ieot_start; | |
2163 | p->ieot_blocks = ieot_blocks; | |
2164 | ||
2165 | /* create a mem_pool for each thread */ | |
2166 | nr = 0; | |
2167 | for (j = p->ieot_start; j < p->ieot_start + p->ieot_blocks; j++) | |
2168 | nr += p->ieot->entries[j].nr; | |
bcd2c5ee | 2169 | p->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2170 | if (istate->version == 4) { |
44c7e1a7 | 2171 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2172 | estimate_cache_size_from_compressed(nr)); |
2173 | } else { | |
44c7e1a7 | 2174 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2175 | estimate_cache_size(mmap_size, nr)); |
2176 | } | |
2177 | ||
2178 | err = pthread_create(&p->pthread, NULL, load_cache_entries_thread, p); | |
2179 | if (err) | |
2180 | die(_("unable to create load_cache_entries thread: %s"), strerror(err)); | |
2181 | ||
2182 | /* increment by the number of cache entries in the ieot block being processed */ | |
2183 | for (j = 0; j < ieot_blocks; j++) | |
2184 | offset += ieot->entries[ieot_start + j].nr; | |
2185 | ieot_start += ieot_blocks; | |
2186 | } | |
2187 | ||
2188 | for (i = 0; i < nr_threads; i++) { | |
2189 | struct load_cache_entries_thread_data *p = &data[i]; | |
2190 | ||
2191 | err = pthread_join(p->pthread, NULL); | |
2192 | if (err) | |
2193 | die(_("unable to join load_cache_entries thread: %s"), strerror(err)); | |
2194 | mem_pool_combine(istate->ce_mem_pool, p->ce_mem_pool); | |
2195 | consumed += p->consumed; | |
2196 | } | |
2197 | ||
2198 | free(data); | |
2199 | ||
2200 | return consumed; | |
2201 | } | |
77ff1127 | 2202 | |
491df5f6 VD |
2203 | static void set_new_index_sparsity(struct index_state *istate) |
2204 | { | |
2205 | /* | |
2206 | * If the index's repo exists, mark it sparse according to | |
2207 | * repo settings. | |
2208 | */ | |
5bdf6d4a ÆAB |
2209 | prepare_repo_settings(istate->repo); |
2210 | if (!istate->repo->settings.command_requires_full_index && | |
2211 | is_sparse_index_allowed(istate, 0)) | |
2212 | istate->sparse_index = 1; | |
491df5f6 VD |
2213 | } |
2214 | ||
8fd2cb40 | 2215 | /* remember to discard_cache() before reading a different cache! */ |
3e52f70b | 2216 | int do_read_index(struct index_state *istate, const char *path, int must_exist) |
e83c5163 | 2217 | { |
77ff1127 | 2218 | int fd; |
e83c5163 | 2219 | struct stat st; |
debed2a6 | 2220 | unsigned long src_offset; |
371ed0de BP |
2221 | const struct cache_header *hdr; |
2222 | const char *mmap; | |
7a51ed66 | 2223 | size_t mmap_size; |
abb4bb83 BP |
2224 | struct load_index_extensions p; |
2225 | size_t extension_offset = 0; | |
77ff1127 BP |
2226 | int nr_threads, cpus; |
2227 | struct index_entry_offset_table *ieot = NULL; | |
e83c5163 | 2228 | |
913e0e99 | 2229 | if (istate->initialized) |
4aab5b46 | 2230 | return istate->cache_nr; |
5d1a5c02 | 2231 | |
fba2f38a KB |
2232 | istate->timestamp.sec = 0; |
2233 | istate->timestamp.nsec = 0; | |
8fd2cb40 | 2234 | fd = open(path, O_RDONLY); |
5d1a5c02 | 2235 | if (fd < 0) { |
491df5f6 VD |
2236 | if (!must_exist && errno == ENOENT) { |
2237 | set_new_index_sparsity(istate); | |
5d1a5c02 | 2238 | return 0; |
491df5f6 | 2239 | } |
9d0a9e90 | 2240 | die_errno(_("%s: index file open failed"), path); |
5d1a5c02 | 2241 | } |
e83c5163 | 2242 | |
3511a377 | 2243 | if (fstat(fd, &st)) |
9d0a9e90 | 2244 | die_errno(_("%s: cannot stat the open index"), path); |
3511a377 | 2245 | |
7a51ed66 | 2246 | mmap_size = xsize_t(st.st_size); |
aab61359 | 2247 | if (mmap_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
9d0a9e90 | 2248 | die(_("%s: index file smaller than expected"), path); |
3511a377 | 2249 | |
02638d1e | 2250 | mmap = xmmap_gently(NULL, mmap_size, PROT_READ, MAP_PRIVATE, fd, 0); |
7a51ed66 | 2251 | if (mmap == MAP_FAILED) |
dc059294 EW |
2252 | die_errno(_("%s: unable to map index file%s"), path, |
2253 | mmap_os_err()); | |
57d84f8d | 2254 | close(fd); |
e83c5163 | 2255 | |
371ed0de | 2256 | hdr = (const struct cache_header *)mmap; |
7a51ed66 | 2257 | if (verify_hdr(hdr, mmap_size) < 0) |
e83c5163 LT |
2258 | goto unmap; |
2259 | ||
92e2cab9 | 2260 | oidread(&istate->oid, (const unsigned char *)hdr + mmap_size - the_hash_algo->rawsz); |
9d227781 | 2261 | istate->version = ntohl(hdr->hdr_version); |
4aab5b46 JH |
2262 | istate->cache_nr = ntohl(hdr->hdr_entries); |
2263 | istate->cache_alloc = alloc_nr(istate->cache_nr); | |
ca56dadb | 2264 | CALLOC_ARRAY(istate->cache, istate->cache_alloc); |
913e0e99 | 2265 | istate->initialized = 1; |
7a51ed66 | 2266 | |
abb4bb83 BP |
2267 | p.istate = istate; |
2268 | p.mmap = mmap; | |
2269 | p.mmap_size = mmap_size; | |
6c9cd161 | 2270 | |
7a51ed66 | 2271 | src_offset = sizeof(*hdr); |
4aab5b46 | 2272 | |
2a9dedef JN |
2273 | if (git_config_get_index_threads(&nr_threads)) |
2274 | nr_threads = 1; | |
7a51ed66 | 2275 | |
77ff1127 BP |
2276 | /* TODO: does creating more threads than cores help? */ |
2277 | if (!nr_threads) { | |
2278 | nr_threads = istate->cache_nr / THREAD_COST; | |
2279 | cpus = online_cpus(); | |
2280 | if (nr_threads > cpus) | |
2281 | nr_threads = cpus; | |
e83c5163 | 2282 | } |
abb4bb83 | 2283 | |
88168b9b NTND |
2284 | if (!HAVE_THREADS) |
2285 | nr_threads = 1; | |
2286 | ||
abb4bb83 BP |
2287 | if (nr_threads > 1) { |
2288 | extension_offset = read_eoie_extension(mmap, mmap_size); | |
2289 | if (extension_offset) { | |
2290 | int err; | |
2291 | ||
2292 | p.src_offset = extension_offset; | |
2293 | err = pthread_create(&p.pthread, NULL, load_index_extensions, &p); | |
2294 | if (err) | |
2295 | die(_("unable to create load_index_extensions thread: %s"), strerror(err)); | |
2296 | ||
2297 | nr_threads--; | |
2298 | } | |
2299 | } | |
abb4bb83 | 2300 | |
77ff1127 BP |
2301 | /* |
2302 | * Locate and read the index entry offset table so that we can use it | |
2303 | * to multi-thread the reading of the cache entries. | |
2304 | */ | |
2305 | if (extension_offset && nr_threads > 1) | |
2306 | ieot = read_ieot_extension(mmap, mmap_size, extension_offset); | |
2307 | ||
2308 | if (ieot) { | |
7bd9631b | 2309 | src_offset += load_cache_entries_threaded(istate, mmap, mmap_size, nr_threads, ieot); |
77ff1127 | 2310 | free(ieot); |
8e72d675 | 2311 | } else { |
77ff1127 | 2312 | src_offset += load_all_cache_entries(istate, mmap, mmap_size, src_offset); |
8e72d675 | 2313 | } |
6c9cd161 | 2314 | |
fba2f38a | 2315 | istate->timestamp.sec = st.st_mtime; |
c06ff490 | 2316 | istate->timestamp.nsec = ST_MTIME_NSEC(st); |
fba2f38a | 2317 | |
abb4bb83 | 2318 | /* if we created a thread, join it otherwise load the extensions on the primary thread */ |
abb4bb83 BP |
2319 | if (extension_offset) { |
2320 | int ret = pthread_join(p.pthread, NULL); | |
2321 | if (ret) | |
2322 | die(_("unable to join load_index_extensions thread: %s"), strerror(ret)); | |
88168b9b | 2323 | } else { |
abb4bb83 BP |
2324 | p.src_offset = src_offset; |
2325 | load_index_extensions(&p); | |
bad68ec9 | 2326 | } |
371ed0de | 2327 | munmap((void *)mmap, mmap_size); |
42fee7a3 JH |
2328 | |
2329 | /* | |
2330 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2331 | * that is associated with the given "istate". | |
2332 | */ | |
2333 | trace2_data_intmax("index", the_repository, "read/version", | |
2334 | istate->version); | |
2335 | trace2_data_intmax("index", the_repository, "read/cache_nr", | |
2336 | istate->cache_nr); | |
2337 | ||
7ca4fc88 VD |
2338 | /* |
2339 | * If the command explicitly requires a full index, force it | |
2340 | * to be full. Otherwise, correct the sparsity based on repository | |
2341 | * settings and other properties of the index (if necessary). | |
2342 | */ | |
4300f844 DS |
2343 | prepare_repo_settings(istate->repo); |
2344 | if (istate->repo->settings.command_requires_full_index) | |
2345 | ensure_full_index(istate); | |
7ca4fc88 VD |
2346 | else |
2347 | ensure_correct_sparsity(istate); | |
4300f844 | 2348 | |
4aab5b46 | 2349 | return istate->cache_nr; |
e83c5163 LT |
2350 | |
2351 | unmap: | |
371ed0de | 2352 | munmap((void *)mmap, mmap_size); |
9d0a9e90 | 2353 | die(_("index file corrupt")); |
e83c5163 LT |
2354 | } |
2355 | ||
0d59ffb4 CC |
2356 | /* |
2357 | * Signal that the shared index is used by updating its mtime. | |
2358 | * | |
2359 | * This way, shared index can be removed if they have not been used | |
2360 | * for some time. | |
2361 | */ | |
a125a223 | 2362 | static void freshen_shared_index(const char *shared_index, int warn) |
0d59ffb4 | 2363 | { |
0d59ffb4 | 2364 | if (!check_and_freshen_file(shared_index, 1) && warn) |
9d0a9e90 | 2365 | warning(_("could not freshen shared index '%s'"), shared_index); |
0d59ffb4 CC |
2366 | } |
2367 | ||
a125a223 TG |
2368 | int read_index_from(struct index_state *istate, const char *path, |
2369 | const char *gitdir) | |
5fc2fc8f NTND |
2370 | { |
2371 | struct split_index *split_index; | |
2372 | int ret; | |
2182abd9 | 2373 | char *base_oid_hex; |
a125a223 | 2374 | char *base_path; |
5fc2fc8f NTND |
2375 | |
2376 | /* istate->initialized covers both .git/index and .git/sharedindex.xxx */ | |
2377 | if (istate->initialized) | |
2378 | return istate->cache_nr; | |
2379 | ||
42fee7a3 JH |
2380 | /* |
2381 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2382 | * that is associated with the given "istate". | |
2383 | */ | |
2384 | trace2_region_enter_printf("index", "do_read_index", the_repository, | |
2385 | "%s", path); | |
c46c406a | 2386 | trace_performance_enter(); |
5fc2fc8f | 2387 | ret = do_read_index(istate, path, 0); |
c46c406a | 2388 | trace_performance_leave("read cache %s", path); |
42fee7a3 JH |
2389 | trace2_region_leave_printf("index", "do_read_index", the_repository, |
2390 | "%s", path); | |
435ec090 | 2391 | |
5fc2fc8f | 2392 | split_index = istate->split_index; |
2182abd9 | 2393 | if (!split_index || is_null_oid(&split_index->base_oid)) { |
435ec090 | 2394 | post_read_index_from(istate); |
5fc2fc8f | 2395 | return ret; |
03f15a79 | 2396 | } |
5fc2fc8f | 2397 | |
c46c406a | 2398 | trace_performance_enter(); |
5fc2fc8f | 2399 | if (split_index->base) |
2f6b1eb7 | 2400 | release_index(split_index->base); |
5fc2fc8f | 2401 | else |
2f6b1eb7 | 2402 | ALLOC_ARRAY(split_index->base, 1); |
6269f8ea | 2403 | index_state_init(split_index->base, istate->repo); |
de6ae5f9 | 2404 | |
2182abd9 | 2405 | base_oid_hex = oid_to_hex(&split_index->base_oid); |
2406 | base_path = xstrfmt("%s/sharedindex.%s", gitdir, base_oid_hex); | |
42fee7a3 JH |
2407 | trace2_region_enter_printf("index", "shared/do_read_index", |
2408 | the_repository, "%s", base_path); | |
998330ac | 2409 | ret = do_read_index(split_index->base, base_path, 0); |
42fee7a3 JH |
2410 | trace2_region_leave_printf("index", "shared/do_read_index", |
2411 | the_repository, "%s", base_path); | |
998330ac SG |
2412 | if (!ret) { |
2413 | char *path_copy = xstrdup(path); | |
652891de JS |
2414 | char *base_path2 = xstrfmt("%s/sharedindex.%s", |
2415 | dirname(path_copy), base_oid_hex); | |
998330ac SG |
2416 | free(path_copy); |
2417 | trace2_region_enter_printf("index", "shared/do_read_index", | |
2418 | the_repository, "%s", base_path2); | |
2419 | ret = do_read_index(split_index->base, base_path2, 1); | |
2420 | trace2_region_leave_printf("index", "shared/do_read_index", | |
2421 | the_repository, "%s", base_path2); | |
652891de | 2422 | free(base_path2); |
998330ac | 2423 | } |
9001dc2a | 2424 | if (!oideq(&split_index->base_oid, &split_index->base->oid)) |
9d0a9e90 | 2425 | die(_("broken index, expect %s in %s, got %s"), |
2182abd9 | 2426 | base_oid_hex, base_path, |
75691ea3 | 2427 | oid_to_hex(&split_index->base->oid)); |
de6ae5f9 | 2428 | |
a125a223 | 2429 | freshen_shared_index(base_path, 0); |
5fc2fc8f | 2430 | merge_base_index(istate); |
435ec090 | 2431 | post_read_index_from(istate); |
c46c406a | 2432 | trace_performance_leave("read cache %s", base_path); |
b42ad7d5 | 2433 | free(base_path); |
5fc2fc8f NTND |
2434 | return ret; |
2435 | } | |
2436 | ||
fa7b3c2f JH |
2437 | int is_index_unborn(struct index_state *istate) |
2438 | { | |
debed2a6 | 2439 | return (!istate->cache_nr && !istate->timestamp.sec); |
fa7b3c2f JH |
2440 | } |
2441 | ||
6269f8ea | 2442 | void index_state_init(struct index_state *istate, struct repository *r) |
2f6b1eb7 | 2443 | { |
6269f8ea | 2444 | struct index_state blank = INDEX_STATE_INIT(r); |
2f6b1eb7 ÆAB |
2445 | memcpy(istate, &blank, sizeof(*istate)); |
2446 | } | |
2447 | ||
2448 | void release_index(struct index_state *istate) | |
6d297f81 | 2449 | { |
8e72d675 JM |
2450 | /* |
2451 | * Cache entries in istate->cache[] should have been allocated | |
2452 | * from the memory pool associated with this index, or from an | |
2453 | * associated split_index. There is no need to free individual | |
8616a2d0 JM |
2454 | * cache entries. validate_cache_entries can detect when this |
2455 | * assertion does not hold. | |
8e72d675 | 2456 | */ |
8616a2d0 | 2457 | validate_cache_entries(istate); |
debed2a6 | 2458 | |
cfc5789a | 2459 | resolve_undo_clear_index(istate); |
2092678c | 2460 | free_name_hash(istate); |
4aab5b46 | 2461 | cache_tree_free(&(istate->cache_tree)); |
2f6b1eb7 ÆAB |
2462 | free(istate->fsmonitor_last_update); |
2463 | free(istate->cache); | |
5fc2fc8f | 2464 | discard_split_index(istate); |
f9e6c649 | 2465 | free_untracked_cache(istate->untracked); |
8e72d675 | 2466 | |
b5fcb1c0 ÆAB |
2467 | if (istate->sparse_checkout_patterns) { |
2468 | clear_pattern_list(istate->sparse_checkout_patterns); | |
2469 | FREE_AND_NULL(istate->sparse_checkout_patterns); | |
2470 | } | |
2471 | ||
8e72d675 | 2472 | if (istate->ce_mem_pool) { |
8616a2d0 | 2473 | mem_pool_discard(istate->ce_mem_pool, should_validate_cache_entries()); |
44c7e1a7 | 2474 | FREE_AND_NULL(istate->ce_mem_pool); |
8e72d675 | 2475 | } |
6d297f81 JS |
2476 | } |
2477 | ||
2f6b1eb7 ÆAB |
2478 | void discard_index(struct index_state *istate) |
2479 | { | |
2480 | release_index(istate); | |
6269f8ea | 2481 | index_state_init(istate, istate->repo); |
2f6b1eb7 ÆAB |
2482 | } |
2483 | ||
8616a2d0 JM |
2484 | /* |
2485 | * Validate the cache entries of this index. | |
2486 | * All cache entries associated with this index | |
2487 | * should have been allocated by the memory pool | |
2488 | * associated with this index, or by a referenced | |
2489 | * split index. | |
2490 | */ | |
2491 | void validate_cache_entries(const struct index_state *istate) | |
2492 | { | |
2493 | int i; | |
2494 | ||
2495 | if (!should_validate_cache_entries() ||!istate || !istate->initialized) | |
2496 | return; | |
2497 | ||
2498 | for (i = 0; i < istate->cache_nr; i++) { | |
2499 | if (!istate) { | |
391408e5 | 2500 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2501 | } else if (!istate->ce_mem_pool || |
2502 | !mem_pool_contains(istate->ce_mem_pool, istate->cache[i])) { | |
2503 | if (!istate->split_index || | |
2504 | !istate->split_index->base || | |
2505 | !istate->split_index->base->ce_mem_pool || | |
2506 | !mem_pool_contains(istate->split_index->base->ce_mem_pool, istate->cache[i])) { | |
391408e5 | 2507 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2508 | } |
2509 | } | |
2510 | } | |
2511 | ||
2512 | if (istate->split_index) | |
2513 | validate_cache_entries(istate->split_index->base); | |
2514 | } | |
2515 | ||
d1f128b0 | 2516 | int unmerged_index(const struct index_state *istate) |
94a5728c DB |
2517 | { |
2518 | int i; | |
2519 | for (i = 0; i < istate->cache_nr; i++) { | |
2520 | if (ce_stage(istate->cache[i])) | |
2521 | return 1; | |
2522 | } | |
2523 | return 0; | |
2524 | } | |
2525 | ||
150fe065 NTND |
2526 | int repo_index_has_changes(struct repository *repo, |
2527 | struct tree *tree, | |
2528 | struct strbuf *sb) | |
cffbfad5 | 2529 | { |
150fe065 | 2530 | struct index_state *istate = repo->index; |
e1f8694f | 2531 | struct object_id cmp; |
cffbfad5 EN |
2532 | int i; |
2533 | ||
e1f8694f EN |
2534 | if (tree) |
2535 | cmp = tree->object.oid; | |
4a93b899 | 2536 | if (tree || !repo_get_oid_tree(repo, "HEAD", &cmp)) { |
cffbfad5 EN |
2537 | struct diff_options opt; |
2538 | ||
150fe065 | 2539 | repo_diff_setup(repo, &opt); |
cffbfad5 EN |
2540 | opt.flags.exit_with_status = 1; |
2541 | if (!sb) | |
2542 | opt.flags.quick = 1; | |
8d833e93 | 2543 | diff_setup_done(&opt); |
e1f8694f | 2544 | do_diff_cache(&cmp, &opt); |
cffbfad5 EN |
2545 | diffcore_std(&opt); |
2546 | for (i = 0; sb && i < diff_queued_diff.nr; i++) { | |
2547 | if (i) | |
2548 | strbuf_addch(sb, ' '); | |
2549 | strbuf_addstr(sb, diff_queued_diff.queue[i]->two->path); | |
2550 | } | |
2551 | diff_flush(&opt); | |
2552 | return opt.flags.has_changes != 0; | |
2553 | } else { | |
0c18c059 DS |
2554 | /* TODO: audit for interaction with sparse-index. */ |
2555 | ensure_full_index(istate); | |
1b9fbefb | 2556 | for (i = 0; sb && i < istate->cache_nr; i++) { |
cffbfad5 EN |
2557 | if (i) |
2558 | strbuf_addch(sb, ' '); | |
1b9fbefb | 2559 | strbuf_addstr(sb, istate->cache[i]->name); |
cffbfad5 | 2560 | } |
1b9fbefb | 2561 | return !!istate->cache_nr; |
cffbfad5 EN |
2562 | } |
2563 | } | |
2564 | ||
410334ed DS |
2565 | static int write_index_ext_header(struct hashfile *f, |
2566 | git_hash_ctx *eoie_f, | |
2567 | unsigned int ext, | |
2568 | unsigned int sz) | |
bad68ec9 | 2569 | { |
410334ed DS |
2570 | hashwrite_be32(f, ext); |
2571 | hashwrite_be32(f, sz); | |
2572 | ||
2573 | if (eoie_f) { | |
2574 | ext = htonl(ext); | |
2575 | sz = htonl(sz); | |
2576 | the_hash_algo->update_fn(eoie_f, &ext, sizeof(ext)); | |
2577 | the_hash_algo->update_fn(eoie_f, &sz, sizeof(sz)); | |
3b1d9e04 | 2578 | } |
410334ed | 2579 | return 0; |
bad68ec9 JH |
2580 | } |
2581 | ||
58bf2a4c NTND |
2582 | static void ce_smudge_racily_clean_entry(struct index_state *istate, |
2583 | struct cache_entry *ce) | |
407c8eb0 JH |
2584 | { |
2585 | /* | |
2586 | * The only thing we care about in this function is to smudge the | |
2587 | * falsely clean entry due to touch-update-touch race, so we leave | |
2588 | * everything else as they are. We are called for entries whose | |
c21d39d7 | 2589 | * ce_stat_data.sd_mtime match the index file mtime. |
c70115b4 JH |
2590 | * |
2591 | * Note that this actually does not do much for gitlinks, for | |
2592 | * which ce_match_stat_basic() always goes to the actual | |
2593 | * contents. The caller checks with is_racy_timestamp() which | |
2594 | * always says "no" for gitlinks, so we are not called for them ;-) | |
407c8eb0 JH |
2595 | */ |
2596 | struct stat st; | |
2597 | ||
2598 | if (lstat(ce->name, &st) < 0) | |
2599 | return; | |
2600 | if (ce_match_stat_basic(ce, &st)) | |
2601 | return; | |
58bf2a4c | 2602 | if (ce_modified_check_fs(istate, ce, &st)) { |
4b3511b0 JH |
2603 | /* This is "racily clean"; smudge it. Note that this |
2604 | * is a tricky code. At first glance, it may appear | |
2605 | * that it can break with this sequence: | |
2606 | * | |
2607 | * $ echo xyzzy >frotz | |
2608 | * $ git-update-index --add frotz | |
2609 | * $ : >frotz | |
2610 | * $ sleep 3 | |
2611 | * $ echo filfre >nitfol | |
2612 | * $ git-update-index --add nitfol | |
2613 | * | |
b7e58b17 | 2614 | * but it does not. When the second update-index runs, |
4b3511b0 JH |
2615 | * it notices that the entry "frotz" has the same timestamp |
2616 | * as index, and if we were to smudge it by resetting its | |
2617 | * size to zero here, then the object name recorded | |
2618 | * in index is the 6-byte file but the cached stat information | |
2619 | * becomes zero --- which would then match what we would | |
a6080a0a | 2620 | * obtain from the filesystem next time we stat("frotz"). |
4b3511b0 JH |
2621 | * |
2622 | * However, the second update-index, before calling | |
2623 | * this function, notices that the cached size is 6 | |
2624 | * bytes and what is on the filesystem is an empty | |
2625 | * file, and never calls us, so the cached size information | |
2626 | * for "frotz" stays 6 which does not match the filesystem. | |
2627 | */ | |
c21d39d7 | 2628 | ce->ce_stat_data.sd_size = 0; |
407c8eb0 JH |
2629 | } |
2630 | } | |
2631 | ||
f136f7bf | 2632 | /* Copy miscellaneous fields but not the name */ |
ce012deb | 2633 | static void copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk, |
f136f7bf | 2634 | struct cache_entry *ce) |
7a51ed66 | 2635 | { |
b60e188c | 2636 | short flags; |
575fa8a3 | 2637 | const unsigned hashsz = the_hash_algo->rawsz; |
2638 | uint16_t *flagsp = (uint16_t *)(ondisk->data + hashsz); | |
b60e188c | 2639 | |
c21d39d7 MH |
2640 | ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec); |
2641 | ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec); | |
2642 | ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec); | |
2643 | ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec); | |
2644 | ondisk->dev = htonl(ce->ce_stat_data.sd_dev); | |
2645 | ondisk->ino = htonl(ce->ce_stat_data.sd_ino); | |
7a51ed66 | 2646 | ondisk->mode = htonl(ce->ce_mode); |
c21d39d7 MH |
2647 | ondisk->uid = htonl(ce->ce_stat_data.sd_uid); |
2648 | ondisk->gid = htonl(ce->ce_stat_data.sd_gid); | |
2649 | ondisk->size = htonl(ce->ce_stat_data.sd_size); | |
575fa8a3 | 2650 | hashcpy(ondisk->data, ce->oid.hash); |
b60e188c | 2651 | |
ce51bf09 | 2652 | flags = ce->ce_flags & ~CE_NAMEMASK; |
b60e188c | 2653 | flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce)); |
575fa8a3 | 2654 | flagsp[0] = htons(flags); |
06aaaa0b | 2655 | if (ce->ce_flags & CE_EXTENDED) { |
575fa8a3 | 2656 | flagsp[1] = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16); |
f136f7bf JH |
2657 | } |
2658 | } | |
2659 | ||
410334ed | 2660 | static int ce_write_entry(struct hashfile *f, struct cache_entry *ce, |
ce012deb | 2661 | struct strbuf *previous_name, struct ondisk_cache_entry *ondisk) |
f136f7bf | 2662 | { |
9d227781 | 2663 | int size; |
00a4b035 RJ |
2664 | unsigned int saved_namelen; |
2665 | int stripped_name = 0; | |
ce012deb | 2666 | static unsigned char padding[8] = { 0x00 }; |
f136f7bf | 2667 | |
b3c96fb1 NTND |
2668 | if (ce->ce_flags & CE_STRIP_NAME) { |
2669 | saved_namelen = ce_namelen(ce); | |
2670 | ce->ce_namelen = 0; | |
00a4b035 | 2671 | stripped_name = 1; |
b3c96fb1 NTND |
2672 | } |
2673 | ||
575fa8a3 | 2674 | size = offsetof(struct ondisk_cache_entry,data) + ondisk_data_size(ce->ce_flags, 0); |
ce012deb | 2675 | |
9d227781 | 2676 | if (!previous_name) { |
ce012deb KW |
2677 | int len = ce_namelen(ce); |
2678 | copy_cache_entry_to_ondisk(ondisk, ce); | |
410334ed DS |
2679 | hashwrite(f, ondisk, size); |
2680 | hashwrite(f, ce->name, len); | |
2681 | hashwrite(f, padding, align_padding_size(size, len)); | |
9d227781 JH |
2682 | } else { |
2683 | int common, to_remove, prefix_size; | |
2684 | unsigned char to_remove_vi[16]; | |
2685 | for (common = 0; | |
2686 | (ce->name[common] && | |
2687 | common < previous_name->len && | |
2688 | ce->name[common] == previous_name->buf[common]); | |
2689 | common++) | |
2690 | ; /* still matching */ | |
2691 | to_remove = previous_name->len - common; | |
2692 | prefix_size = encode_varint(to_remove, to_remove_vi); | |
2693 | ||
ce012deb | 2694 | copy_cache_entry_to_ondisk(ondisk, ce); |
410334ed DS |
2695 | hashwrite(f, ondisk, size); |
2696 | hashwrite(f, to_remove_vi, prefix_size); | |
2697 | hashwrite(f, ce->name + common, ce_namelen(ce) - common); | |
2698 | hashwrite(f, padding, 1); | |
9d227781 JH |
2699 | |
2700 | strbuf_splice(previous_name, common, to_remove, | |
2701 | ce->name + common, ce_namelen(ce) - common); | |
06aaaa0b | 2702 | } |
00a4b035 | 2703 | if (stripped_name) { |
b3c96fb1 NTND |
2704 | ce->ce_namelen = saved_namelen; |
2705 | ce->ce_flags &= ~CE_STRIP_NAME; | |
2706 | } | |
7a51ed66 | 2707 | |
410334ed | 2708 | return 0; |
7a51ed66 LT |
2709 | } |
2710 | ||
426ddeea YM |
2711 | /* |
2712 | * This function verifies if index_state has the correct sha1 of the | |
2713 | * index file. Don't die if we have any other failure, just return 0. | |
2714 | */ | |
2715 | static int verify_index_from(const struct index_state *istate, const char *path) | |
2716 | { | |
2717 | int fd; | |
2718 | ssize_t n; | |
2719 | struct stat st; | |
aab61359 | 2720 | unsigned char hash[GIT_MAX_RAWSZ]; |
426ddeea YM |
2721 | |
2722 | if (!istate->initialized) | |
2723 | return 0; | |
2724 | ||
2725 | fd = open(path, O_RDONLY); | |
2726 | if (fd < 0) | |
2727 | return 0; | |
2728 | ||
2729 | if (fstat(fd, &st)) | |
2730 | goto out; | |
2731 | ||
aab61359 | 2732 | if (st.st_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
426ddeea YM |
2733 | goto out; |
2734 | ||
aab61359 | 2735 | n = pread_in_full(fd, hash, the_hash_algo->rawsz, st.st_size - the_hash_algo->rawsz); |
2736 | if (n != the_hash_algo->rawsz) | |
426ddeea YM |
2737 | goto out; |
2738 | ||
67947c34 | 2739 | if (!hasheq(istate->oid.hash, hash)) |
426ddeea YM |
2740 | goto out; |
2741 | ||
2742 | close(fd); | |
2743 | return 1; | |
2744 | ||
2745 | out: | |
2746 | close(fd); | |
2747 | return 0; | |
2748 | } | |
2749 | ||
1b0d968b | 2750 | static int repo_verify_index(struct repository *repo) |
426ddeea | 2751 | { |
1b0d968b | 2752 | return verify_index_from(repo->index, repo->index_file); |
426ddeea YM |
2753 | } |
2754 | ||
2ede073f | 2755 | int has_racy_timestamp(struct index_state *istate) |
483fbe2b JH |
2756 | { |
2757 | int entries = istate->cache_nr; | |
2758 | int i; | |
2759 | ||
2760 | for (i = 0; i < entries; i++) { | |
2761 | struct cache_entry *ce = istate->cache[i]; | |
2762 | if (is_racy_timestamp(istate, ce)) | |
2763 | return 1; | |
2764 | } | |
2765 | return 0; | |
2766 | } | |
2767 | ||
1b0d968b NTND |
2768 | void repo_update_index_if_able(struct repository *repo, |
2769 | struct lock_file *lockfile) | |
ccdc4ec3 | 2770 | { |
1b0d968b NTND |
2771 | if ((repo->index->cache_changed || |
2772 | has_racy_timestamp(repo->index)) && | |
2773 | repo_verify_index(repo)) | |
2774 | write_locked_index(repo->index, lockfile, COMMIT_LOCK); | |
b74c90fb MÅ |
2775 | else |
2776 | rollback_lock_file(lockfile); | |
ccdc4ec3 JH |
2777 | } |
2778 | ||
d8465500 JN |
2779 | static int record_eoie(void) |
2780 | { | |
2781 | int val; | |
2782 | ||
2783 | if (!git_config_get_bool("index.recordendofindexentries", &val)) | |
2784 | return val; | |
2a9dedef JN |
2785 | |
2786 | /* | |
2787 | * As a convenience, the end of index entries extension | |
2788 | * used for threading is written by default if the user | |
2789 | * explicitly requested threaded index reads. | |
2790 | */ | |
2791 | return !git_config_get_index_threads(&val) && val != 1; | |
d8465500 JN |
2792 | } |
2793 | ||
42916054 JN |
2794 | static int record_ieot(void) |
2795 | { | |
2796 | int val; | |
2797 | ||
2798 | if (!git_config_get_bool("index.recordoffsettable", &val)) | |
2799 | return val; | |
2a9dedef JN |
2800 | |
2801 | /* | |
2802 | * As a convenience, the offset table used for threading is | |
2803 | * written by default if the user explicitly requested | |
2804 | * threaded index reads. | |
2805 | */ | |
2806 | return !git_config_get_index_threads(&val) && val != 1; | |
42916054 JN |
2807 | } |
2808 | ||
3b7a4475 JS |
2809 | enum write_extensions { |
2810 | WRITE_NO_EXTENSION = 0, | |
2811 | WRITE_SPLIT_INDEX_EXTENSION = 1<<0, | |
2812 | WRITE_CACHE_TREE_EXTENSION = 1<<1, | |
2813 | WRITE_RESOLVE_UNDO_EXTENSION = 1<<2, | |
2814 | WRITE_UNTRACKED_CACHE_EXTENSION = 1<<3, | |
2815 | WRITE_FSMONITOR_EXTENSION = 1<<4, | |
2816 | }; | |
2817 | #define WRITE_ALL_EXTENSIONS ((enum write_extensions)-1) | |
2818 | ||
812d6b00 MÅ |
2819 | /* |
2820 | * On success, `tempfile` is closed. If it is the temporary file | |
2821 | * of a `struct lock_file`, we will therefore effectively perform | |
2822 | * a 'close_lock_file_gently()`. Since that is an implementation | |
2823 | * detail of lockfiles, callers of `do_write_index()` should not | |
2824 | * rely on it. | |
2825 | */ | |
9f41c7a6 | 2826 | static int do_write_index(struct index_state *istate, struct tempfile *tempfile, |
3b7a4475 | 2827 | enum write_extensions write_extensions, unsigned flags) |
197ee8c9 | 2828 | { |
ca54d9ba | 2829 | uint64_t start = getnanotime(); |
410334ed DS |
2830 | struct hashfile *f; |
2831 | git_hash_ctx *eoie_c = NULL; | |
197ee8c9 | 2832 | struct cache_header hdr; |
b50386c7 | 2833 | int i, err = 0, removed, extended, hdr_version; |
4aab5b46 JH |
2834 | struct cache_entry **cache = istate->cache; |
2835 | int entries = istate->cache_nr; | |
e1afca4f | 2836 | struct stat st; |
575fa8a3 | 2837 | struct ondisk_cache_entry ondisk; |
9d227781 | 2838 | struct strbuf previous_name_buf = STRBUF_INIT, *previous_name; |
4bddd983 | 2839 | int drop_cache_tree = istate->drop_cache_tree; |
3b1d9e04 | 2840 | off_t offset; |
ba95e96d | 2841 | int csum_fsync_flag; |
77ff1127 | 2842 | int ieot_entries = 1; |
3255089a BP |
2843 | struct index_entry_offset_table *ieot = NULL; |
2844 | int nr, nr_threads; | |
6269f8ea | 2845 | struct repository *r = istate->repo; |
025a0709 | 2846 | |
410334ed DS |
2847 | f = hashfd(tempfile->fd, tempfile->filename.buf); |
2848 | ||
17194b19 DS |
2849 | prepare_repo_settings(r); |
2850 | f->skip_hash = r->settings.index_skip_hash; | |
ee1f0c24 | 2851 | |
06aaaa0b | 2852 | for (i = removed = extended = 0; i < entries; i++) { |
7a51ed66 | 2853 | if (cache[i]->ce_flags & CE_REMOVE) |
025a0709 | 2854 | removed++; |
197ee8c9 | 2855 | |
06aaaa0b NTND |
2856 | /* reduce extended entries if possible */ |
2857 | cache[i]->ce_flags &= ~CE_EXTENDED; | |
2858 | if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) { | |
2859 | extended++; | |
2860 | cache[i]->ce_flags |= CE_EXTENDED; | |
2861 | } | |
2862 | } | |
2863 | ||
e8ffd034 | 2864 | if (!istate->version) |
4a93b899 | 2865 | istate->version = get_index_format_default(r); |
9d227781 JH |
2866 | |
2867 | /* demote version 3 to version 2 when the latter suffices */ | |
2868 | if (istate->version == 3 || istate->version == 2) | |
2869 | istate->version = extended ? 3 : 2; | |
2870 | ||
2871 | hdr_version = istate->version; | |
2872 | ||
ccc4feb5 | 2873 | hdr.hdr_signature = htonl(CACHE_SIGNATURE); |
9d227781 | 2874 | hdr.hdr_version = htonl(hdr_version); |
025a0709 | 2875 | hdr.hdr_entries = htonl(entries - removed); |
197ee8c9 | 2876 | |
410334ed | 2877 | hashwrite(f, &hdr, sizeof(hdr)); |
197ee8c9 | 2878 | |
2a9dedef | 2879 | if (!HAVE_THREADS || git_config_get_index_threads(&nr_threads)) |
88168b9b | 2880 | nr_threads = 1; |
62e5ee81 | 2881 | |
42916054 | 2882 | if (nr_threads != 1 && record_ieot()) { |
3255089a BP |
2883 | int ieot_blocks, cpus; |
2884 | ||
2885 | /* | |
2886 | * ensure default number of ieot blocks maps evenly to the | |
2887 | * default number of threads that will process them leaving | |
2888 | * room for the thread to load the index extensions. | |
2889 | */ | |
2890 | if (!nr_threads) { | |
2891 | ieot_blocks = istate->cache_nr / THREAD_COST; | |
2892 | cpus = online_cpus(); | |
2893 | if (ieot_blocks > cpus - 1) | |
2894 | ieot_blocks = cpus - 1; | |
2895 | } else { | |
2896 | ieot_blocks = nr_threads; | |
77ff1127 BP |
2897 | if (ieot_blocks > istate->cache_nr) |
2898 | ieot_blocks = istate->cache_nr; | |
3255089a BP |
2899 | } |
2900 | ||
2901 | /* | |
2902 | * no reason to write out the IEOT extension if we don't | |
2903 | * have enough blocks to utilize multi-threading | |
2904 | */ | |
2905 | if (ieot_blocks > 1) { | |
2906 | ieot = xcalloc(1, sizeof(struct index_entry_offset_table) | |
2907 | + (ieot_blocks * sizeof(struct index_entry_offset))); | |
77ff1127 | 2908 | ieot_entries = DIV_ROUND_UP(entries, ieot_blocks); |
3255089a BP |
2909 | } |
2910 | } | |
3255089a | 2911 | |
410334ed DS |
2912 | offset = hashfile_total(f); |
2913 | ||
3255089a | 2914 | nr = 0; |
9d227781 | 2915 | previous_name = (hdr_version == 4) ? &previous_name_buf : NULL; |
ce012deb | 2916 | |
197ee8c9 LT |
2917 | for (i = 0; i < entries; i++) { |
2918 | struct cache_entry *ce = cache[i]; | |
7a51ed66 | 2919 | if (ce->ce_flags & CE_REMOVE) |
aa16021e | 2920 | continue; |
e06c43c7 | 2921 | if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce)) |
58bf2a4c | 2922 | ce_smudge_racily_clean_entry(istate, ce); |
99d1a986 | 2923 | if (is_null_oid(&ce->oid)) { |
83bd7437 JK |
2924 | static const char msg[] = "cache entry has null sha1: %s"; |
2925 | static int allow = -1; | |
2926 | ||
2927 | if (allow < 0) | |
2928 | allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0); | |
2929 | if (allow) | |
2930 | warning(msg, ce->name); | |
2931 | else | |
b50386c7 | 2932 | err = error(msg, ce->name); |
a96d3cc3 JK |
2933 | |
2934 | drop_cache_tree = 1; | |
83bd7437 | 2935 | } |
77ff1127 | 2936 | if (ieot && i && (i % ieot_entries == 0)) { |
3255089a BP |
2937 | ieot->entries[ieot->nr].nr = nr; |
2938 | ieot->entries[ieot->nr].offset = offset; | |
2939 | ieot->nr++; | |
2940 | /* | |
2941 | * If we have a V4 index, set the first byte to an invalid | |
2942 | * character to ensure there is nothing common with the previous | |
2943 | * entry | |
2944 | */ | |
2945 | if (previous_name) | |
2946 | previous_name->buf[0] = 0; | |
2947 | nr = 0; | |
410334ed DS |
2948 | |
2949 | offset = hashfile_total(f); | |
3255089a | 2950 | } |
410334ed | 2951 | if (ce_write_entry(f, ce, previous_name, (struct ondisk_cache_entry *)&ondisk) < 0) |
b50386c7 KW |
2952 | err = -1; |
2953 | ||
2954 | if (err) | |
2955 | break; | |
3255089a BP |
2956 | nr++; |
2957 | } | |
2958 | if (ieot && nr) { | |
2959 | ieot->entries[ieot->nr].nr = nr; | |
2960 | ieot->entries[ieot->nr].offset = offset; | |
2961 | ieot->nr++; | |
197ee8c9 | 2962 | } |
9d227781 | 2963 | strbuf_release(&previous_name_buf); |
1af1c2b6 | 2964 | |
3255089a BP |
2965 | if (err) { |
2966 | free(ieot); | |
b50386c7 | 2967 | return err; |
3255089a | 2968 | } |
b50386c7 | 2969 | |
410334ed DS |
2970 | offset = hashfile_total(f); |
2971 | ||
2972 | /* | |
2973 | * The extension headers must be hashed on their own for the | |
2974 | * EOIE extension. Create a hashfile here to compute that hash. | |
2975 | */ | |
2976 | if (offset && record_eoie()) { | |
2977 | CALLOC_ARRAY(eoie_c, 1); | |
2978 | the_hash_algo->init_fn(eoie_c); | |
3255089a | 2979 | } |
3b1d9e04 | 2980 | |
3255089a BP |
2981 | /* |
2982 | * Lets write out CACHE_EXT_INDEXENTRYOFFSETTABLE first so that we | |
2983 | * can minimize the number of extensions we have to scan through to | |
2984 | * find it during load. Write it out regardless of the | |
2985 | * strip_extensions parameter as we need it when loading the shared | |
2986 | * index. | |
2987 | */ | |
3255089a BP |
2988 | if (ieot) { |
2989 | struct strbuf sb = STRBUF_INIT; | |
2990 | ||
2991 | write_ieot_extension(&sb, ieot); | |
410334ed DS |
2992 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_INDEXENTRYOFFSETTABLE, sb.len) < 0; |
2993 | hashwrite(f, sb.buf, sb.len); | |
3255089a BP |
2994 | strbuf_release(&sb); |
2995 | free(ieot); | |
2996 | if (err) | |
2997 | return -1; | |
2998 | } | |
3255089a | 2999 | |
3b7a4475 JS |
3000 | if (write_extensions & WRITE_SPLIT_INDEX_EXTENSION && |
3001 | istate->split_index) { | |
5fc2fc8f NTND |
3002 | struct strbuf sb = STRBUF_INIT; |
3003 | ||
451b66c5 JS |
3004 | if (istate->sparse_index) |
3005 | die(_("cannot write split index for a sparse index")); | |
3006 | ||
5fc2fc8f | 3007 | err = write_link_extension(&sb, istate) < 0 || |
410334ed DS |
3008 | write_index_ext_header(f, eoie_c, CACHE_EXT_LINK, |
3009 | sb.len) < 0; | |
3010 | hashwrite(f, sb.buf, sb.len); | |
5fc2fc8f NTND |
3011 | strbuf_release(&sb); |
3012 | if (err) | |
3013 | return -1; | |
3014 | } | |
3b7a4475 JS |
3015 | if (write_extensions & WRITE_CACHE_TREE_EXTENSION && |
3016 | !drop_cache_tree && istate->cache_tree) { | |
f285a2d7 | 3017 | struct strbuf sb = STRBUF_INIT; |
1dffb8fa | 3018 | |
1dffb8fa | 3019 | cache_tree_write(&sb, istate->cache_tree); |
410334ed DS |
3020 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_TREE, sb.len) < 0; |
3021 | hashwrite(f, sb.buf, sb.len); | |
1dffb8fa PH |
3022 | strbuf_release(&sb); |
3023 | if (err) | |
bad68ec9 | 3024 | return -1; |
bad68ec9 | 3025 | } |
3b7a4475 JS |
3026 | if (write_extensions & WRITE_RESOLVE_UNDO_EXTENSION && |
3027 | istate->resolve_undo) { | |
cfc5789a JH |
3028 | struct strbuf sb = STRBUF_INIT; |
3029 | ||
3030 | resolve_undo_write(&sb, istate->resolve_undo); | |
410334ed DS |
3031 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_RESOLVE_UNDO, |
3032 | sb.len) < 0; | |
3033 | hashwrite(f, sb.buf, sb.len); | |
cfc5789a JH |
3034 | strbuf_release(&sb); |
3035 | if (err) | |
3036 | return -1; | |
3037 | } | |
3b7a4475 JS |
3038 | if (write_extensions & WRITE_UNTRACKED_CACHE_EXTENSION && |
3039 | istate->untracked) { | |
83c094ad NTND |
3040 | struct strbuf sb = STRBUF_INIT; |
3041 | ||
3042 | write_untracked_extension(&sb, istate->untracked); | |
410334ed DS |
3043 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_UNTRACKED, |
3044 | sb.len) < 0; | |
3045 | hashwrite(f, sb.buf, sb.len); | |
83c094ad NTND |
3046 | strbuf_release(&sb); |
3047 | if (err) | |
3048 | return -1; | |
3049 | } | |
3b7a4475 JS |
3050 | if (write_extensions & WRITE_FSMONITOR_EXTENSION && |
3051 | istate->fsmonitor_last_update) { | |
883e248b BP |
3052 | struct strbuf sb = STRBUF_INIT; |
3053 | ||
3054 | write_fsmonitor_extension(&sb, istate); | |
410334ed DS |
3055 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_FSMONITOR, sb.len) < 0; |
3056 | hashwrite(f, sb.buf, sb.len); | |
3b1d9e04 BP |
3057 | strbuf_release(&sb); |
3058 | if (err) | |
3059 | return -1; | |
3060 | } | |
cd42415f | 3061 | if (istate->sparse_index) { |
410334ed | 3062 | if (write_index_ext_header(f, eoie_c, CACHE_EXT_SPARSE_DIRECTORIES, 0) < 0) |
cd42415f DS |
3063 | return -1; |
3064 | } | |
3b1d9e04 BP |
3065 | |
3066 | /* | |
3067 | * CACHE_EXT_ENDOFINDEXENTRIES must be written as the last entry before the SHA1 | |
3068 | * so that it can be found and processed before all the index entries are | |
3069 | * read. Write it out regardless of the strip_extensions parameter as we need it | |
3070 | * when loading the shared index. | |
3071 | */ | |
410334ed | 3072 | if (eoie_c) { |
3b1d9e04 BP |
3073 | struct strbuf sb = STRBUF_INIT; |
3074 | ||
410334ed DS |
3075 | write_eoie_extension(&sb, eoie_c, offset); |
3076 | err = write_index_ext_header(f, NULL, CACHE_EXT_ENDOFINDEXENTRIES, sb.len) < 0; | |
3077 | hashwrite(f, sb.buf, sb.len); | |
883e248b BP |
3078 | strbuf_release(&sb); |
3079 | if (err) | |
3080 | return -1; | |
3081 | } | |
e1afca4f | 3082 | |
ba95e96d NS |
3083 | csum_fsync_flag = 0; |
3084 | if (!alternate_index_output && (flags & COMMIT_LOCK)) | |
3085 | csum_fsync_flag = CSUM_FSYNC; | |
3086 | ||
3087 | finalize_hashfile(f, istate->oid.hash, FSYNC_COMPONENT_INDEX, | |
3088 | CSUM_HASH_IN_STREAM | csum_fsync_flag); | |
3089 | ||
49bd0fc2 | 3090 | if (close_tempfile_gently(tempfile)) { |
6a8c89d0 | 3091 | error(_("could not close '%s'"), get_tempfile_path(tempfile)); |
49bd0fc2 JK |
3092 | return -1; |
3093 | } | |
6a8c89d0 | 3094 | if (stat(get_tempfile_path(tempfile), &st)) |
e1afca4f | 3095 | return -1; |
5bcf109c KB |
3096 | istate->timestamp.sec = (unsigned int)st.st_mtime; |
3097 | istate->timestamp.nsec = ST_MTIME_NSEC(st); | |
ca54d9ba | 3098 | trace_performance_since(start, "write index, changed mask = %x", istate->cache_changed); |
42fee7a3 JH |
3099 | |
3100 | /* | |
3101 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3102 | * that is associated with the given "istate". | |
3103 | */ | |
3104 | trace2_data_intmax("index", the_repository, "write/version", | |
3105 | istate->version); | |
3106 | trace2_data_intmax("index", the_repository, "write/cache_nr", | |
3107 | istate->cache_nr); | |
3108 | ||
e1afca4f | 3109 | return 0; |
197ee8c9 | 3110 | } |
e46bbcf6 | 3111 | |
626f35c8 NTND |
3112 | void set_alternate_index_output(const char *name) |
3113 | { | |
3114 | alternate_index_output = name; | |
3115 | } | |
3116 | ||
3117 | static int commit_locked_index(struct lock_file *lk) | |
3118 | { | |
751baced MH |
3119 | if (alternate_index_output) |
3120 | return commit_lock_file_to(lk, alternate_index_output); | |
3121 | else | |
626f35c8 | 3122 | return commit_lock_file(lk); |
626f35c8 NTND |
3123 | } |
3124 | ||
3b7a4475 JS |
3125 | static int do_write_locked_index(struct index_state *istate, |
3126 | struct lock_file *lock, | |
3127 | unsigned flags, | |
3128 | enum write_extensions write_extensions) | |
03b86647 | 3129 | { |
42fee7a3 | 3130 | int ret; |
9fadb373 | 3131 | int was_full = istate->sparse_index == INDEX_EXPANDED; |
6e773527 | 3132 | |
ce7a9f01 | 3133 | ret = convert_to_sparse(istate, 0); |
6e773527 DS |
3134 | |
3135 | if (ret) { | |
3136 | warning(_("failed to convert to a sparse-index")); | |
3137 | return ret; | |
3138 | } | |
42fee7a3 JH |
3139 | |
3140 | /* | |
3141 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3142 | * that is associated with the given "istate". | |
3143 | */ | |
3144 | trace2_region_enter_printf("index", "do_write_index", the_repository, | |
6a8c89d0 | 3145 | "%s", get_lock_file_path(lock)); |
3b7a4475 | 3146 | ret = do_write_index(istate, lock->tempfile, write_extensions, flags); |
42fee7a3 | 3147 | trace2_region_leave_printf("index", "do_write_index", the_repository, |
6a8c89d0 | 3148 | "%s", get_lock_file_path(lock)); |
42fee7a3 | 3149 | |
6e773527 DS |
3150 | if (was_full) |
3151 | ensure_full_index(istate); | |
3152 | ||
03b86647 NTND |
3153 | if (ret) |
3154 | return ret; | |
03b86647 | 3155 | if (flags & COMMIT_LOCK) |
1956ecd0 BP |
3156 | ret = commit_locked_index(lock); |
3157 | else | |
3158 | ret = close_lock_file_gently(lock); | |
3159 | ||
dbb1c613 | 3160 | run_hooks_l("post-index-change", |
1956ecd0 BP |
3161 | istate->updated_workdir ? "1" : "0", |
3162 | istate->updated_skipworktree ? "1" : "0", NULL); | |
3163 | istate->updated_workdir = 0; | |
3164 | istate->updated_skipworktree = 0; | |
3165 | ||
3166 | return ret; | |
03b86647 NTND |
3167 | } |
3168 | ||
5fc2fc8f NTND |
3169 | static int write_split_index(struct index_state *istate, |
3170 | struct lock_file *lock, | |
3171 | unsigned flags) | |
3172 | { | |
3173 | int ret; | |
3174 | prepare_to_write_split_index(istate); | |
3b7a4475 | 3175 | ret = do_write_locked_index(istate, lock, flags, WRITE_ALL_EXTENSIONS); |
5fc2fc8f NTND |
3176 | finish_writing_split_index(istate); |
3177 | return ret; | |
3178 | } | |
3179 | ||
b9683722 CC |
3180 | static const char *shared_index_expire = "2.weeks.ago"; |
3181 | ||
3182 | static unsigned long get_shared_index_expire_date(void) | |
3183 | { | |
3184 | static unsigned long shared_index_expire_date; | |
3185 | static int shared_index_expire_date_prepared; | |
3186 | ||
3187 | if (!shared_index_expire_date_prepared) { | |
3188 | git_config_get_expiry("splitindex.sharedindexexpire", | |
3189 | &shared_index_expire); | |
3190 | shared_index_expire_date = approxidate(shared_index_expire); | |
3191 | shared_index_expire_date_prepared = 1; | |
3192 | } | |
3193 | ||
3194 | return shared_index_expire_date; | |
3195 | } | |
3196 | ||
3197 | static int should_delete_shared_index(const char *shared_index_path) | |
3198 | { | |
3199 | struct stat st; | |
3200 | unsigned long expiration; | |
3201 | ||
3202 | /* Check timestamp */ | |
3203 | expiration = get_shared_index_expire_date(); | |
3204 | if (!expiration) | |
3205 | return 0; | |
3206 | if (stat(shared_index_path, &st)) | |
78bde923 | 3207 | return error_errno(_("could not stat '%s'"), shared_index_path); |
b9683722 CC |
3208 | if (st.st_mtime > expiration) |
3209 | return 0; | |
3210 | ||
3211 | return 1; | |
3212 | } | |
3213 | ||
3214 | static int clean_shared_index_files(const char *current_hex) | |
3215 | { | |
3216 | struct dirent *de; | |
3217 | DIR *dir = opendir(get_git_dir()); | |
3218 | ||
3219 | if (!dir) | |
3220 | return error_errno(_("unable to open git dir: %s"), get_git_dir()); | |
3221 | ||
3222 | while ((de = readdir(dir)) != NULL) { | |
3223 | const char *sha1_hex; | |
3224 | const char *shared_index_path; | |
3225 | if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex)) | |
3226 | continue; | |
3227 | if (!strcmp(sha1_hex, current_hex)) | |
3228 | continue; | |
3229 | shared_index_path = git_path("%s", de->d_name); | |
3230 | if (should_delete_shared_index(shared_index_path) > 0 && | |
3231 | unlink(shared_index_path)) | |
3232 | warning_errno(_("unable to unlink: %s"), shared_index_path); | |
3233 | } | |
3234 | closedir(dir); | |
3235 | ||
3236 | return 0; | |
3237 | } | |
3238 | ||
a0a96756 | 3239 | static int write_shared_index(struct index_state *istate, |
ba95e96d | 3240 | struct tempfile **temp, unsigned flags) |
c18b80a0 NTND |
3241 | { |
3242 | struct split_index *si = istate->split_index; | |
6e773527 | 3243 | int ret, was_full = !istate->sparse_index; |
c18b80a0 | 3244 | |
c18b80a0 | 3245 | move_cache_to_base_index(istate); |
ce7a9f01 | 3246 | convert_to_sparse(istate, 0); |
42fee7a3 JH |
3247 | |
3248 | trace2_region_enter_printf("index", "shared/do_write_index", | |
6a8c89d0 | 3249 | the_repository, "%s", get_tempfile_path(*temp)); |
3b7a4475 | 3250 | ret = do_write_index(si->base, *temp, WRITE_NO_EXTENSION, flags); |
c173542c | 3251 | trace2_region_leave_printf("index", "shared/do_write_index", |
6a8c89d0 | 3252 | the_repository, "%s", get_tempfile_path(*temp)); |
42fee7a3 | 3253 | |
6e773527 DS |
3254 | if (was_full) |
3255 | ensure_full_index(istate); | |
3256 | ||
59f9d2dd | 3257 | if (ret) |
c18b80a0 | 3258 | return ret; |
7db2d08c | 3259 | ret = adjust_shared_perm(get_tempfile_path(*temp)); |
df801f3f | 3260 | if (ret) { |
9d0a9e90 | 3261 | error(_("cannot fix permission bits on '%s'"), get_tempfile_path(*temp)); |
df801f3f CC |
3262 | return ret; |
3263 | } | |
7db2d08c | 3264 | ret = rename_tempfile(temp, |
75691ea3 | 3265 | git_path("sharedindex.%s", oid_to_hex(&si->base->oid))); |
b9683722 | 3266 | if (!ret) { |
75691ea3 | 3267 | oidcpy(&si->base_oid, &si->base->oid); |
3268 | clean_shared_index_files(oid_to_hex(&si->base->oid)); | |
b9683722 CC |
3269 | } |
3270 | ||
c18b80a0 NTND |
3271 | return ret; |
3272 | } | |
3273 | ||
e6a1dd77 CC |
3274 | static const int default_max_percent_split_change = 20; |
3275 | ||
3276 | static int too_many_not_shared_entries(struct index_state *istate) | |
3277 | { | |
3278 | int i, not_shared = 0; | |
3279 | int max_split = git_config_get_max_percent_split_change(); | |
3280 | ||
3281 | switch (max_split) { | |
3282 | case -1: | |
3283 | /* not or badly configured: use the default value */ | |
3284 | max_split = default_max_percent_split_change; | |
3285 | break; | |
3286 | case 0: | |
3287 | return 1; /* 0% means always write a new shared index */ | |
3288 | case 100: | |
3289 | return 0; /* 100% means never write a new shared index */ | |
3290 | default: | |
3291 | break; /* just use the configured value */ | |
3292 | } | |
3293 | ||
3294 | /* Count not shared entries */ | |
3295 | for (i = 0; i < istate->cache_nr; i++) { | |
3296 | struct cache_entry *ce = istate->cache[i]; | |
3297 | if (!ce->index) | |
3298 | not_shared++; | |
3299 | } | |
3300 | ||
3301 | return (int64_t)istate->cache_nr * max_split < (int64_t)not_shared * 100; | |
3302 | } | |
3303 | ||
03b86647 NTND |
3304 | int write_locked_index(struct index_state *istate, struct lock_file *lock, |
3305 | unsigned flags) | |
3306 | { | |
e8ffd034 | 3307 | int new_shared_index, ret, test_split_index_env; |
5fc2fc8f NTND |
3308 | struct split_index *si = istate->split_index; |
3309 | ||
4592e608 | 3310 | if (git_env_bool("GIT_TEST_CHECK_CACHE_TREE", 0)) |
c207e9e1 | 3311 | cache_tree_verify(the_repository, istate); |
4592e608 | 3312 | |
61000814 MÅ |
3313 | if ((flags & SKIP_IF_UNCHANGED) && !istate->cache_changed) { |
3314 | if (flags & COMMIT_LOCK) | |
3315 | rollback_lock_file(lock); | |
3316 | return 0; | |
3317 | } | |
3318 | ||
3bd28eb2 AV |
3319 | if (istate->fsmonitor_last_update) |
3320 | fill_fsmonitor_bitmap(istate); | |
3321 | ||
e8ffd034 SG |
3322 | test_split_index_env = git_env_bool("GIT_TEST_SPLIT_INDEX", 0); |
3323 | ||
3324 | if ((!si && !test_split_index_env) || | |
3325 | alternate_index_output || | |
5165dd59 | 3326 | (istate->cache_changed & ~EXTMASK)) { |
3b7a4475 JS |
3327 | ret = do_write_locked_index(istate, lock, flags, |
3328 | ~WRITE_SPLIT_INDEX_EXTENSION); | |
df60cf57 | 3329 | goto out; |
5fc2fc8f NTND |
3330 | } |
3331 | ||
e8ffd034 SG |
3332 | if (test_split_index_env) { |
3333 | if (!si) { | |
3334 | si = init_split_index(istate); | |
d6e3c181 | 3335 | istate->cache_changed |= SPLIT_INDEX_ORDERED; |
e8ffd034 SG |
3336 | } else { |
3337 | int v = si->base_oid.hash[0]; | |
3338 | if ((v & 15) < 6) | |
3339 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
3340 | } | |
d6e3c181 | 3341 | } |
e6a1dd77 CC |
3342 | if (too_many_not_shared_entries(istate)) |
3343 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
0d59ffb4 CC |
3344 | |
3345 | new_shared_index = istate->cache_changed & SPLIT_INDEX_ORDERED; | |
3346 | ||
3347 | if (new_shared_index) { | |
59f9d2dd NTND |
3348 | struct tempfile *temp; |
3349 | int saved_errno; | |
3350 | ||
c9d6c788 ÆAB |
3351 | /* Same initial permissions as the main .git/index file */ |
3352 | temp = mks_tempfile_sm(git_path("sharedindex_XXXXXX"), 0, 0666); | |
59f9d2dd | 3353 | if (!temp) { |
3b7a4475 JS |
3354 | ret = do_write_locked_index(istate, lock, flags, |
3355 | ~WRITE_SPLIT_INDEX_EXTENSION); | |
ef5b3a6c NTND |
3356 | goto out; |
3357 | } | |
ba95e96d | 3358 | ret = write_shared_index(istate, &temp, flags); |
59f9d2dd NTND |
3359 | |
3360 | saved_errno = errno; | |
3361 | if (is_tempfile_active(temp)) | |
3362 | delete_tempfile(&temp); | |
3363 | errno = saved_errno; | |
3364 | ||
c18b80a0 | 3365 | if (ret) |
df60cf57 | 3366 | goto out; |
c18b80a0 NTND |
3367 | } |
3368 | ||
0d59ffb4 CC |
3369 | ret = write_split_index(istate, lock, flags); |
3370 | ||
3371 | /* Freshen the shared index only if the split-index was written */ | |
6e37c8ed | 3372 | if (!ret && !new_shared_index && !is_null_oid(&si->base_oid)) { |
a125a223 | 3373 | const char *shared_index = git_path("sharedindex.%s", |
2182abd9 | 3374 | oid_to_hex(&si->base_oid)); |
a125a223 TG |
3375 | freshen_shared_index(shared_index, 1); |
3376 | } | |
0d59ffb4 | 3377 | |
df60cf57 MÅ |
3378 | out: |
3379 | if (flags & COMMIT_LOCK) | |
3380 | rollback_lock_file(lock); | |
0d59ffb4 | 3381 | return ret; |
03b86647 NTND |
3382 | } |
3383 | ||
e46bbcf6 MV |
3384 | /* |
3385 | * Read the index file that is potentially unmerged into given | |
ad376204 EN |
3386 | * index_state, dropping any unmerged entries to stage #0 (potentially |
3387 | * resulting in a path appearing as both a file and a directory in the | |
3388 | * index; the caller is responsible to clear out the extra entries | |
3389 | * before writing the index to a tree). Returns true if the index is | |
3390 | * unmerged. Callers who want to refuse to work from an unmerged | |
3391 | * state can call this and check its return value, instead of calling | |
3392 | * read_cache(). | |
e46bbcf6 | 3393 | */ |
e1ff0a32 | 3394 | int repo_read_index_unmerged(struct repository *repo) |
e46bbcf6 | 3395 | { |
e1ff0a32 | 3396 | struct index_state *istate; |
e46bbcf6 | 3397 | int i; |
d1a43f2a | 3398 | int unmerged = 0; |
e46bbcf6 | 3399 | |
e1ff0a32 NTND |
3400 | repo_read_index(repo); |
3401 | istate = repo->index; | |
e46bbcf6 MV |
3402 | for (i = 0; i < istate->cache_nr; i++) { |
3403 | struct cache_entry *ce = istate->cache[i]; | |
d1a43f2a | 3404 | struct cache_entry *new_ce; |
a849735b | 3405 | int len; |
d1a43f2a JH |
3406 | |
3407 | if (!ce_stage(ce)) | |
e46bbcf6 | 3408 | continue; |
d1a43f2a | 3409 | unmerged = 1; |
68c4f6a5 | 3410 | len = ce_namelen(ce); |
a849735b | 3411 | new_ce = make_empty_cache_entry(istate, len); |
d1a43f2a | 3412 | memcpy(new_ce->name, ce->name, len); |
b60e188c TG |
3413 | new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED; |
3414 | new_ce->ce_namelen = len; | |
d1a43f2a | 3415 | new_ce->ce_mode = ce->ce_mode; |
ad376204 | 3416 | if (add_index_entry(istate, new_ce, ADD_CACHE_SKIP_DFCHECK)) |
9d0a9e90 | 3417 | return error(_("%s: cannot drop to stage #0"), |
5699d17e | 3418 | new_ce->name); |
e46bbcf6 | 3419 | } |
d1a43f2a | 3420 | return unmerged; |
e46bbcf6 | 3421 | } |
041aee31 | 3422 | |
98fa4738 JK |
3423 | /* |
3424 | * Returns 1 if the path is an "other" path with respect to | |
3425 | * the index; that is, the path is not mentioned in the index at all, | |
3426 | * either as a file, a directory with some files in the index, | |
3427 | * or as an unmerged entry. | |
3428 | * | |
3429 | * We helpfully remove a trailing "/" from directories so that | |
3430 | * the output of read_directory can be used as-is. | |
3431 | */ | |
847a9e5d DS |
3432 | int index_name_is_other(struct index_state *istate, const char *name, |
3433 | int namelen) | |
98fa4738 JK |
3434 | { |
3435 | int pos; | |
3436 | if (namelen && name[namelen - 1] == '/') | |
3437 | namelen--; | |
3438 | pos = index_name_pos(istate, name, namelen); | |
3439 | if (0 <= pos) | |
3440 | return 0; /* exact match */ | |
3441 | pos = -pos - 1; | |
3442 | if (pos < istate->cache_nr) { | |
3443 | struct cache_entry *ce = istate->cache[pos]; | |
3444 | if (ce_namelen(ce) == namelen && | |
3445 | !memcmp(ce->name, name, namelen)) | |
3446 | return 0; /* Yup, this one exists unmerged */ | |
3447 | } | |
3448 | return 1; | |
3449 | } | |
29fb37b2 | 3450 | |
847a9e5d | 3451 | void *read_blob_data_from_index(struct index_state *istate, |
87542508 | 3452 | const char *path, unsigned long *size) |
29fb37b2 LF |
3453 | { |
3454 | int pos, len; | |
3455 | unsigned long sz; | |
3456 | enum object_type type; | |
3457 | void *data; | |
3458 | ||
3459 | len = strlen(path); | |
3460 | pos = index_name_pos(istate, path, len); | |
3461 | if (pos < 0) { | |
3462 | /* | |
3463 | * We might be in the middle of a merge, in which | |
3464 | * case we would read stage #2 (ours). | |
3465 | */ | |
3466 | int i; | |
3467 | for (i = -pos - 1; | |
3468 | (pos < 0 && i < istate->cache_nr && | |
3469 | !strcmp(istate->cache[i]->name, path)); | |
3470 | i++) | |
3471 | if (ce_stage(istate->cache[i]) == 2) | |
3472 | pos = i; | |
3473 | } | |
3474 | if (pos < 0) | |
3475 | return NULL; | |
bc726bd0 ÆAB |
3476 | data = repo_read_object_file(the_repository, &istate->cache[pos]->oid, |
3477 | &type, &sz); | |
29fb37b2 LF |
3478 | if (!data || type != OBJ_BLOB) { |
3479 | free(data); | |
3480 | return NULL; | |
3481 | } | |
ff366825 LF |
3482 | if (size) |
3483 | *size = sz; | |
29fb37b2 LF |
3484 | return data; |
3485 | } | |
38612532 | 3486 | |
edf3b905 DT |
3487 | void move_index_extensions(struct index_state *dst, struct index_state *src) |
3488 | { | |
3489 | dst->untracked = src->untracked; | |
3490 | src->untracked = NULL; | |
836ef2b6 NTND |
3491 | dst->cache_tree = src->cache_tree; |
3492 | src->cache_tree = NULL; | |
edf3b905 | 3493 | } |
a849735b | 3494 | |
8e72d675 JM |
3495 | struct cache_entry *dup_cache_entry(const struct cache_entry *ce, |
3496 | struct index_state *istate) | |
3497 | { | |
3498 | unsigned int size = ce_size(ce); | |
3499 | int mem_pool_allocated; | |
3500 | struct cache_entry *new_entry = make_empty_cache_entry(istate, ce_namelen(ce)); | |
3501 | mem_pool_allocated = new_entry->mem_pool_allocated; | |
3502 | ||
3503 | memcpy(new_entry, ce, size); | |
3504 | new_entry->mem_pool_allocated = mem_pool_allocated; | |
3505 | return new_entry; | |
3506 | } | |
3507 | ||
a849735b JM |
3508 | void discard_cache_entry(struct cache_entry *ce) |
3509 | { | |
8616a2d0 JM |
3510 | if (ce && should_validate_cache_entries()) |
3511 | memset(ce, 0xCD, cache_entry_size(ce->ce_namelen)); | |
3512 | ||
8e72d675 JM |
3513 | if (ce && ce->mem_pool_allocated) |
3514 | return; | |
3515 | ||
a849735b JM |
3516 | free(ce); |
3517 | } | |
8616a2d0 JM |
3518 | |
3519 | int should_validate_cache_entries(void) | |
3520 | { | |
3521 | static int validate_index_cache_entries = -1; | |
3522 | ||
3523 | if (validate_index_cache_entries < 0) { | |
3524 | if (getenv("GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES")) | |
3525 | validate_index_cache_entries = 1; | |
3526 | else | |
3527 | validate_index_cache_entries = 0; | |
3528 | } | |
3529 | ||
3530 | return validate_index_cache_entries; | |
3531 | } | |
3b1d9e04 BP |
3532 | |
3533 | #define EOIE_SIZE (4 + GIT_SHA1_RAWSZ) /* <4-byte offset> + <20-byte hash> */ | |
3534 | #define EOIE_SIZE_WITH_HEADER (4 + 4 + EOIE_SIZE) /* <4-byte signature> + <4-byte length> + EOIE_SIZE */ | |
3535 | ||
3536 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size) | |
3537 | { | |
3538 | /* | |
3539 | * The end of index entries (EOIE) extension is guaranteed to be last | |
3540 | * so that it can be found by scanning backwards from the EOF. | |
3541 | * | |
3542 | * "EOIE" | |
3543 | * <4-byte length> | |
3544 | * <4-byte offset> | |
3545 | * <20-byte hash> | |
3546 | */ | |
3547 | const char *index, *eoie; | |
3548 | uint32_t extsize; | |
3549 | size_t offset, src_offset; | |
3550 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3551 | git_hash_ctx c; | |
3552 | ||
3553 | /* ensure we have an index big enough to contain an EOIE extension */ | |
3554 | if (mmap_size < sizeof(struct cache_header) + EOIE_SIZE_WITH_HEADER + the_hash_algo->rawsz) | |
3555 | return 0; | |
3556 | ||
3557 | /* validate the extension signature */ | |
3558 | index = eoie = mmap + mmap_size - EOIE_SIZE_WITH_HEADER - the_hash_algo->rawsz; | |
3559 | if (CACHE_EXT(index) != CACHE_EXT_ENDOFINDEXENTRIES) | |
3560 | return 0; | |
3561 | index += sizeof(uint32_t); | |
3562 | ||
3563 | /* validate the extension size */ | |
3564 | extsize = get_be32(index); | |
3565 | if (extsize != EOIE_SIZE) | |
3566 | return 0; | |
3567 | index += sizeof(uint32_t); | |
3568 | ||
3569 | /* | |
3570 | * Validate the offset we're going to look for the first extension | |
3571 | * signature is after the index header and before the eoie extension. | |
3572 | */ | |
3573 | offset = get_be32(index); | |
3574 | if (mmap + offset < mmap + sizeof(struct cache_header)) | |
3575 | return 0; | |
3576 | if (mmap + offset >= eoie) | |
3577 | return 0; | |
3578 | index += sizeof(uint32_t); | |
3579 | ||
3580 | /* | |
3581 | * The hash is computed over extension types and their sizes (but not | |
3582 | * their contents). E.g. if we have "TREE" extension that is N-bytes | |
3583 | * long, "REUC" extension that is M-bytes long, followed by "EOIE", | |
3584 | * then the hash would be: | |
3585 | * | |
3586 | * SHA-1("TREE" + <binary representation of N> + | |
3587 | * "REUC" + <binary representation of M>) | |
3588 | */ | |
3589 | src_offset = offset; | |
3590 | the_hash_algo->init_fn(&c); | |
3591 | while (src_offset < mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) { | |
3592 | /* After an array of active_nr index entries, | |
3593 | * there can be arbitrary number of extended | |
3594 | * sections, each of which is prefixed with | |
3595 | * extension name (4-byte) and section length | |
3596 | * in 4-byte network byte order. | |
3597 | */ | |
3598 | uint32_t extsize; | |
3599 | memcpy(&extsize, mmap + src_offset + 4, 4); | |
3600 | extsize = ntohl(extsize); | |
3601 | ||
3602 | /* verify the extension size isn't so large it will wrap around */ | |
3603 | if (src_offset + 8 + extsize < src_offset) | |
3604 | return 0; | |
3605 | ||
3606 | the_hash_algo->update_fn(&c, mmap + src_offset, 8); | |
3607 | ||
3608 | src_offset += 8; | |
3609 | src_offset += extsize; | |
3610 | } | |
3611 | the_hash_algo->final_fn(hash, &c); | |
3612 | if (!hasheq(hash, (const unsigned char *)index)) | |
3613 | return 0; | |
3614 | ||
3615 | /* Validate that the extension offsets returned us back to the eoie extension. */ | |
3616 | if (src_offset != mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) | |
3617 | return 0; | |
3618 | ||
3619 | return offset; | |
3620 | } | |
3621 | ||
3622 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset) | |
3623 | { | |
3624 | uint32_t buffer; | |
3625 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3626 | ||
3627 | /* offset */ | |
3628 | put_be32(&buffer, offset); | |
3629 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3630 | ||
3631 | /* hash */ | |
3632 | the_hash_algo->final_fn(hash, eoie_context); | |
3633 | strbuf_add(sb, hash, the_hash_algo->rawsz); | |
3634 | } | |
3255089a | 3635 | |
3255089a BP |
3636 | #define IEOT_VERSION (1) |
3637 | ||
3638 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset) | |
3639 | { | |
ec36c42a NTND |
3640 | const char *index = NULL; |
3641 | uint32_t extsize, ext_version; | |
3642 | struct index_entry_offset_table *ieot; | |
3643 | int i, nr; | |
3644 | ||
3645 | /* find the IEOT extension */ | |
3646 | if (!offset) | |
3647 | return NULL; | |
3648 | while (offset <= mmap_size - the_hash_algo->rawsz - 8) { | |
3649 | extsize = get_be32(mmap + offset + 4); | |
3650 | if (CACHE_EXT((mmap + offset)) == CACHE_EXT_INDEXENTRYOFFSETTABLE) { | |
3651 | index = mmap + offset + 4 + 4; | |
3652 | break; | |
3653 | } | |
3654 | offset += 8; | |
3655 | offset += extsize; | |
3656 | } | |
3657 | if (!index) | |
3658 | return NULL; | |
3659 | ||
3660 | /* validate the version is IEOT_VERSION */ | |
3661 | ext_version = get_be32(index); | |
3662 | if (ext_version != IEOT_VERSION) { | |
3663 | error("invalid IEOT version %d", ext_version); | |
3664 | return NULL; | |
3665 | } | |
3666 | index += sizeof(uint32_t); | |
3667 | ||
3668 | /* extension size - version bytes / bytes per entry */ | |
3669 | nr = (extsize - sizeof(uint32_t)) / (sizeof(uint32_t) + sizeof(uint32_t)); | |
3670 | if (!nr) { | |
3671 | error("invalid number of IEOT entries %d", nr); | |
3672 | return NULL; | |
3673 | } | |
3674 | ieot = xmalloc(sizeof(struct index_entry_offset_table) | |
3675 | + (nr * sizeof(struct index_entry_offset))); | |
3676 | ieot->nr = nr; | |
3677 | for (i = 0; i < nr; i++) { | |
3678 | ieot->entries[i].offset = get_be32(index); | |
3679 | index += sizeof(uint32_t); | |
3680 | ieot->entries[i].nr = get_be32(index); | |
3681 | index += sizeof(uint32_t); | |
3682 | } | |
3683 | ||
3684 | return ieot; | |
3255089a BP |
3685 | } |
3686 | ||
3687 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot) | |
3688 | { | |
ec36c42a NTND |
3689 | uint32_t buffer; |
3690 | int i; | |
3255089a | 3691 | |
ec36c42a NTND |
3692 | /* version */ |
3693 | put_be32(&buffer, IEOT_VERSION); | |
3694 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3695 | |
ec36c42a NTND |
3696 | /* ieot */ |
3697 | for (i = 0; i < ieot->nr; i++) { | |
3255089a | 3698 | |
ec36c42a NTND |
3699 | /* offset */ |
3700 | put_be32(&buffer, ieot->entries[i].offset); | |
3701 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3702 | |
ec36c42a NTND |
3703 | /* count */ |
3704 | put_be32(&buffer, ieot->entries[i].nr); | |
3705 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3706 | } | |
3255089a | 3707 | } |
b2896d27 JT |
3708 | |
3709 | void prefetch_cache_entries(const struct index_state *istate, | |
3710 | must_prefetch_predicate must_prefetch) | |
3711 | { | |
3712 | int i; | |
3713 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
3714 | ||
3715 | for (i = 0; i < istate->cache_nr; i++) { | |
3716 | struct cache_entry *ce = istate->cache[i]; | |
3717 | ||
3718 | if (S_ISGITLINK(ce->ce_mode) || !must_prefetch(ce)) | |
3719 | continue; | |
3720 | if (!oid_object_info_extended(the_repository, &ce->oid, | |
3721 | NULL, | |
3722 | OBJECT_INFO_FOR_PREFETCH)) | |
3723 | continue; | |
3724 | oid_array_append(&to_fetch, &ce->oid); | |
3725 | } | |
3726 | promisor_remote_get_direct(the_repository, | |
3727 | to_fetch.oid, to_fetch.nr); | |
3728 | oid_array_clear(&to_fetch); | |
3729 | } | |
1a40e7be EN |
3730 | |
3731 | static int read_one_entry_opt(struct index_state *istate, | |
3732 | const struct object_id *oid, | |
3733 | struct strbuf *base, | |
3734 | const char *pathname, | |
3735 | unsigned mode, int opt) | |
3736 | { | |
3737 | int len; | |
3738 | struct cache_entry *ce; | |
3739 | ||
3740 | if (S_ISDIR(mode)) | |
3741 | return READ_TREE_RECURSIVE; | |
3742 | ||
3743 | len = strlen(pathname); | |
3744 | ce = make_empty_cache_entry(istate, base->len + len); | |
3745 | ||
3746 | ce->ce_mode = create_ce_mode(mode); | |
3747 | ce->ce_flags = create_ce_flags(1); | |
3748 | ce->ce_namelen = base->len + len; | |
3749 | memcpy(ce->name, base->buf, base->len); | |
3750 | memcpy(ce->name + base->len, pathname, len+1); | |
3751 | oidcpy(&ce->oid, oid); | |
3752 | return add_index_entry(istate, ce, opt); | |
3753 | } | |
3754 | ||
3755 | static int read_one_entry(const struct object_id *oid, struct strbuf *base, | |
3756 | const char *pathname, unsigned mode, | |
3757 | void *context) | |
3758 | { | |
3759 | struct index_state *istate = context; | |
3760 | return read_one_entry_opt(istate, oid, base, pathname, | |
3761 | mode, | |
3762 | ADD_CACHE_OK_TO_ADD|ADD_CACHE_SKIP_DFCHECK); | |
3763 | } | |
3764 | ||
3765 | /* | |
3766 | * This is used when the caller knows there is no existing entries at | |
3767 | * the stage that will conflict with the entry being added. | |
3768 | */ | |
3769 | static int read_one_entry_quick(const struct object_id *oid, struct strbuf *base, | |
3770 | const char *pathname, unsigned mode, | |
3771 | void *context) | |
3772 | { | |
3773 | struct index_state *istate = context; | |
3774 | return read_one_entry_opt(istate, oid, base, pathname, | |
3775 | mode, ADD_CACHE_JUST_APPEND); | |
3776 | } | |
3777 | ||
3778 | /* | |
3779 | * Read the tree specified with --with-tree option | |
3780 | * (typically, HEAD) into stage #1 and then | |
3781 | * squash them down to stage #0. This is used for | |
3782 | * --error-unmatch to list and check the path patterns | |
3783 | * that were given from the command line. We are not | |
3784 | * going to write this index out. | |
3785 | */ | |
3786 | void overlay_tree_on_index(struct index_state *istate, | |
3787 | const char *tree_name, const char *prefix) | |
3788 | { | |
3789 | struct tree *tree; | |
3790 | struct object_id oid; | |
3791 | struct pathspec pathspec; | |
3792 | struct cache_entry *last_stage0 = NULL; | |
3793 | int i; | |
3794 | read_tree_fn_t fn = NULL; | |
3795 | int err; | |
3796 | ||
3797 | if (repo_get_oid(the_repository, tree_name, &oid)) | |
3798 | die("tree-ish %s not found.", tree_name); | |
3799 | tree = parse_tree_indirect(&oid); | |
3800 | if (!tree) | |
3801 | die("bad tree-ish %s", tree_name); | |
3802 | ||
3803 | /* Hoist the unmerged entries up to stage #3 to make room */ | |
3804 | /* TODO: audit for interaction with sparse-index. */ | |
3805 | ensure_full_index(istate); | |
3806 | for (i = 0; i < istate->cache_nr; i++) { | |
3807 | struct cache_entry *ce = istate->cache[i]; | |
3808 | if (!ce_stage(ce)) | |
3809 | continue; | |
3810 | ce->ce_flags |= CE_STAGEMASK; | |
3811 | } | |
3812 | ||
3813 | if (prefix) { | |
3814 | static const char *(matchbuf[1]); | |
3815 | matchbuf[0] = NULL; | |
3816 | parse_pathspec(&pathspec, PATHSPEC_ALL_MAGIC, | |
3817 | PATHSPEC_PREFER_CWD, prefix, matchbuf); | |
3818 | } else | |
3819 | memset(&pathspec, 0, sizeof(pathspec)); | |
3820 | ||
3821 | /* | |
3822 | * See if we have cache entry at the stage. If so, | |
3823 | * do it the original slow way, otherwise, append and then | |
3824 | * sort at the end. | |
3825 | */ | |
3826 | for (i = 0; !fn && i < istate->cache_nr; i++) { | |
3827 | const struct cache_entry *ce = istate->cache[i]; | |
3828 | if (ce_stage(ce) == 1) | |
3829 | fn = read_one_entry; | |
3830 | } | |
3831 | ||
3832 | if (!fn) | |
3833 | fn = read_one_entry_quick; | |
3834 | err = read_tree(the_repository, tree, &pathspec, fn, istate); | |
3835 | clear_pathspec(&pathspec); | |
3836 | if (err) | |
3837 | die("unable to read tree entries %s", tree_name); | |
3838 | ||
3839 | /* | |
3840 | * Sort the cache entry -- we need to nuke the cache tree, though. | |
3841 | */ | |
3842 | if (fn == read_one_entry_quick) { | |
3843 | cache_tree_free(&istate->cache_tree); | |
3844 | QSORT(istate->cache, istate->cache_nr, cmp_cache_name_compare); | |
3845 | } | |
3846 | ||
3847 | for (i = 0; i < istate->cache_nr; i++) { | |
3848 | struct cache_entry *ce = istate->cache[i]; | |
3849 | switch (ce_stage(ce)) { | |
3850 | case 0: | |
3851 | last_stage0 = ce; | |
3852 | /* fallthru */ | |
3853 | default: | |
3854 | continue; | |
3855 | case 1: | |
3856 | /* | |
3857 | * If there is stage #0 entry for this, we do not | |
3858 | * need to show it. We use CE_UPDATE bit to mark | |
3859 | * such an entry. | |
3860 | */ | |
3861 | if (last_stage0 && | |
3862 | !strcmp(last_stage0->name, ce->name)) | |
3863 | ce->ce_flags |= CE_UPDATE; | |
3864 | } | |
3865 | } | |
3866 | } | |
6cee5ebc EN |
3867 | |
3868 | struct update_callback_data { | |
3869 | struct index_state *index; | |
3870 | int include_sparse; | |
3871 | int flags; | |
3872 | int add_errors; | |
3873 | }; | |
3874 | ||
3875 | static int fix_unmerged_status(struct diff_filepair *p, | |
3876 | struct update_callback_data *data) | |
3877 | { | |
3878 | if (p->status != DIFF_STATUS_UNMERGED) | |
3879 | return p->status; | |
3880 | if (!(data->flags & ADD_CACHE_IGNORE_REMOVAL) && !p->two->mode) | |
3881 | /* | |
3882 | * This is not an explicit add request, and the | |
3883 | * path is missing from the working tree (deleted) | |
3884 | */ | |
3885 | return DIFF_STATUS_DELETED; | |
3886 | else | |
3887 | /* | |
3888 | * Either an explicit add request, or path exists | |
3889 | * in the working tree. An attempt to explicitly | |
3890 | * add a path that does not exist in the working tree | |
3891 | * will be caught as an error by the caller immediately. | |
3892 | */ | |
3893 | return DIFF_STATUS_MODIFIED; | |
3894 | } | |
3895 | ||
3896 | static void update_callback(struct diff_queue_struct *q, | |
3897 | struct diff_options *opt UNUSED, void *cbdata) | |
3898 | { | |
3899 | int i; | |
3900 | struct update_callback_data *data = cbdata; | |
3901 | ||
3902 | for (i = 0; i < q->nr; i++) { | |
3903 | struct diff_filepair *p = q->queue[i]; | |
3904 | const char *path = p->one->path; | |
3905 | ||
3906 | if (!data->include_sparse && | |
3907 | !path_in_sparse_checkout(path, data->index)) | |
3908 | continue; | |
3909 | ||
3910 | switch (fix_unmerged_status(p, data)) { | |
3911 | default: | |
3912 | die(_("unexpected diff status %c"), p->status); | |
3913 | case DIFF_STATUS_MODIFIED: | |
3914 | case DIFF_STATUS_TYPE_CHANGED: | |
3915 | if (add_file_to_index(data->index, path, data->flags)) { | |
3916 | if (!(data->flags & ADD_CACHE_IGNORE_ERRORS)) | |
3917 | die(_("updating files failed")); | |
3918 | data->add_errors++; | |
3919 | } | |
3920 | break; | |
3921 | case DIFF_STATUS_DELETED: | |
3922 | if (data->flags & ADD_CACHE_IGNORE_REMOVAL) | |
3923 | break; | |
3924 | if (!(data->flags & ADD_CACHE_PRETEND)) | |
3925 | remove_file_from_index(data->index, path); | |
3926 | if (data->flags & (ADD_CACHE_PRETEND|ADD_CACHE_VERBOSE)) | |
3927 | printf(_("remove '%s'\n"), path); | |
3928 | break; | |
3929 | } | |
3930 | } | |
3931 | } | |
3932 | ||
3933 | int add_files_to_cache(struct repository *repo, const char *prefix, | |
3934 | const struct pathspec *pathspec, int include_sparse, | |
3935 | int flags) | |
3936 | { | |
3937 | struct update_callback_data data; | |
3938 | struct rev_info rev; | |
3939 | ||
3940 | memset(&data, 0, sizeof(data)); | |
3941 | data.index = repo->index; | |
3942 | data.include_sparse = include_sparse; | |
3943 | data.flags = flags; | |
3944 | ||
3945 | repo_init_revisions(repo, &rev, prefix); | |
3946 | setup_revisions(0, NULL, &rev, NULL); | |
3947 | if (pathspec) | |
3948 | copy_pathspec(&rev.prune_data, pathspec); | |
3949 | rev.diffopt.output_format = DIFF_FORMAT_CALLBACK; | |
3950 | rev.diffopt.format_callback = update_callback; | |
3951 | rev.diffopt.format_callback_data = &data; | |
3952 | rev.diffopt.flags.override_submodule_config = 1; | |
3953 | rev.max_count = 0; /* do not compare unmerged paths with stage #2 */ | |
3954 | ||
3955 | /* | |
3956 | * Use an ODB transaction to optimize adding multiple objects. | |
3957 | * This function is invoked from commands other than 'add', which | |
3958 | * may not have their own transaction active. | |
3959 | */ | |
3960 | begin_odb_transaction(); | |
3961 | run_diff_files(&rev, DIFF_RACY_IS_MODIFIED); | |
3962 | end_odb_transaction(); | |
3963 | ||
3964 | release_revisions(&rev); | |
3965 | return !!data.add_errors; | |
3966 | } |