]>
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" |
6cee5ebc | 7 | #include "bulk-checkin.h" |
b2141fc1 | 8 | #include "config.h" |
d4a4f929 | 9 | #include "date.h" |
cffbfad5 EN |
10 | #include "diff.h" |
11 | #include "diffcore.h" | |
41771fa4 | 12 | #include "hex.h" |
f6ecc62d | 13 | #include "tempfile.h" |
697cc8ef | 14 | #include "lockfile.h" |
bad68ec9 | 15 | #include "cache-tree.h" |
f35a6d3b | 16 | #include "refs.h" |
d616813d | 17 | #include "dir.h" |
87bed179 | 18 | #include "object-file.h" |
a034e910 | 19 | #include "object-store-ll.h" |
6f2d7430 | 20 | #include "oid-array.h" |
041aee31 JH |
21 | #include "tree.h" |
22 | #include "commit.h" | |
32a8f510 | 23 | #include "environment.h" |
f394e093 | 24 | #include "gettext.h" |
5bc07225 | 25 | #include "mem-pool.h" |
f5653856 | 26 | #include "name-hash.h" |
dabab1d6 | 27 | #include "object-name.h" |
c339932b | 28 | #include "path.h" |
fbffdfb1 | 29 | #include "preload-index.h" |
08c46a49 | 30 | #include "read-cache.h" |
cfc5789a | 31 | #include "resolve-undo.h" |
6cee5ebc | 32 | #include "revision.h" |
6c9cd161 | 33 | #include "strbuf.h" |
74ea5c95 | 34 | #include "trace2.h" |
6c9cd161 | 35 | #include "varint.h" |
5fc2fc8f | 36 | #include "split-index.h" |
cb2a5135 | 37 | #include "symlinks.h" |
a42643aa | 38 | #include "utf8.h" |
883e248b | 39 | #include "fsmonitor.h" |
abb4bb83 | 40 | #include "thread-utils.h" |
ae9af122 | 41 | #include "progress.h" |
6e773527 | 42 | #include "sparse-index.h" |
410334ed | 43 | #include "csum-file.h" |
b2896d27 | 44 | #include "promisor-remote.h" |
dbb1c613 | 45 | #include "hook.h" |
bad68ec9 | 46 | |
b60e188c TG |
47 | /* Mask for the name length in ce_flags in the on-disk index */ |
48 | ||
49 | #define CE_NAMEMASK (0x0fff) | |
50 | ||
bad68ec9 JH |
51 | /* Index extensions. |
52 | * | |
53 | * The first letter should be 'A'..'Z' for extensions that are not | |
54 | * necessary for a correct operation (i.e. optimization data). | |
55 | * When new extensions are added that _needs_ to be understood in | |
56 | * order to correctly interpret the index file, pick character that | |
57 | * is outside the range, to cause the reader to abort. | |
58 | */ | |
59 | ||
60 | #define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) ) | |
61 | #define CACHE_EXT_TREE 0x54524545 /* "TREE" */ | |
b659b49b | 62 | #define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */ |
5fc2fc8f | 63 | #define CACHE_EXT_LINK 0x6c696e6b /* "link" */ |
83c094ad | 64 | #define CACHE_EXT_UNTRACKED 0x554E5452 /* "UNTR" */ |
883e248b | 65 | #define CACHE_EXT_FSMONITOR 0x46534D4E /* "FSMN" */ |
3b1d9e04 | 66 | #define CACHE_EXT_ENDOFINDEXENTRIES 0x454F4945 /* "EOIE" */ |
3255089a | 67 | #define CACHE_EXT_INDEXENTRYOFFSETTABLE 0x49454F54 /* "IEOT" */ |
cd42415f | 68 | #define CACHE_EXT_SPARSE_DIRECTORIES 0x73646972 /* "sdir" */ |
5fc2fc8f NTND |
69 | |
70 | /* changes that can be kept in $GIT_DIR/index (basically all extensions) */ | |
e0cf0d7d | 71 | #define EXTMASK (RESOLVE_UNDO_CHANGED | CACHE_TREE_CHANGED | \ |
c18b80a0 | 72 | CE_ENTRY_ADDED | CE_ENTRY_REMOVED | CE_ENTRY_CHANGED | \ |
883e248b | 73 | SPLIT_INDEX_ORDERED | UNTRACKED_CHANGED | FSMONITOR_CHANGED) |
e83c5163 | 74 | |
8e72d675 JM |
75 | |
76 | /* | |
77 | * This is an estimate of the pathname length in the index. We use | |
78 | * this for V4 index files to guess the un-deltafied size of the index | |
79 | * in memory because of pathname deltafication. This is not required | |
80 | * for V2/V3 index formats because their pathnames are not compressed. | |
81 | * If the initial amount of memory set aside is not sufficient, the | |
82 | * mem pool will allocate extra memory. | |
83 | */ | |
84 | #define CACHE_ENTRY_PATH_LENGTH 80 | |
85 | ||
20ec2d03 VD |
86 | enum index_search_mode { |
87 | NO_EXPAND_SPARSE = 0, | |
88 | EXPAND_SPARSE = 1 | |
89 | }; | |
90 | ||
8e72d675 JM |
91 | static inline struct cache_entry *mem_pool__ce_alloc(struct mem_pool *mem_pool, size_t len) |
92 | { | |
93 | struct cache_entry *ce; | |
94 | ce = mem_pool_alloc(mem_pool, cache_entry_size(len)); | |
95 | ce->mem_pool_allocated = 1; | |
96 | return ce; | |
97 | } | |
98 | ||
99 | static inline struct cache_entry *mem_pool__ce_calloc(struct mem_pool *mem_pool, size_t len) | |
100 | { | |
101 | struct cache_entry * ce; | |
102 | ce = mem_pool_calloc(mem_pool, 1, cache_entry_size(len)); | |
103 | ce->mem_pool_allocated = 1; | |
104 | return ce; | |
105 | } | |
106 | ||
107 | static struct mem_pool *find_mem_pool(struct index_state *istate) | |
108 | { | |
109 | struct mem_pool **pool_ptr; | |
110 | ||
111 | if (istate->split_index && istate->split_index->base) | |
112 | pool_ptr = &istate->split_index->base->ce_mem_pool; | |
113 | else | |
114 | pool_ptr = &istate->ce_mem_pool; | |
115 | ||
44c7e1a7 EN |
116 | if (!*pool_ptr) { |
117 | *pool_ptr = xmalloc(sizeof(**pool_ptr)); | |
118 | mem_pool_init(*pool_ptr, 0); | |
119 | } | |
8e72d675 JM |
120 | |
121 | return *pool_ptr; | |
122 | } | |
123 | ||
626f35c8 | 124 | static const char *alternate_index_output; |
8fd2cb40 | 125 | |
9cb76b8c JH |
126 | static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
127 | { | |
4300f844 | 128 | if (S_ISSPARSEDIR(ce->ce_mode)) |
9fadb373 | 129 | istate->sparse_index = INDEX_COLLAPSED; |
4300f844 | 130 | |
9cb76b8c | 131 | istate->cache[nr] = ce; |
96872bc2 | 132 | add_name_hash(istate, ce); |
9cb76b8c JH |
133 | } |
134 | ||
cf558704 LT |
135 | static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
136 | { | |
137 | struct cache_entry *old = istate->cache[nr]; | |
138 | ||
078a58e8 | 139 | replace_index_entry_in_base(istate, old, ce); |
2092678c | 140 | remove_name_hash(istate, old); |
a849735b | 141 | discard_cache_entry(old); |
0e267b7a | 142 | ce->ce_flags &= ~CE_HASHED; |
a22c6371 | 143 | set_index_entry(istate, nr, ce); |
078a58e8 | 144 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 145 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 146 | istate->cache_changed |= CE_ENTRY_CHANGED; |
cf558704 LT |
147 | } |
148 | ||
81dc2307 PB |
149 | void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name) |
150 | { | |
b7f9130a | 151 | struct cache_entry *old_entry = istate->cache[nr], *new_entry, *refreshed; |
81dc2307 PB |
152 | int namelen = strlen(new_name); |
153 | ||
a849735b | 154 | new_entry = make_empty_cache_entry(istate, namelen); |
285c2e25 BW |
155 | copy_cache_entry(new_entry, old_entry); |
156 | new_entry->ce_flags &= ~CE_HASHED; | |
157 | new_entry->ce_namelen = namelen; | |
158 | new_entry->index = 0; | |
159 | memcpy(new_entry->name, new_name, namelen + 1); | |
81dc2307 | 160 | |
285c2e25 BW |
161 | cache_tree_invalidate_path(istate, old_entry->name); |
162 | untracked_cache_remove_from_index(istate, old_entry->name); | |
81dc2307 | 163 | remove_index_entry_at(istate, nr); |
b7f9130a VD |
164 | |
165 | /* | |
166 | * Refresh the new index entry. Using 'refresh_cache_entry' ensures | |
167 | * we only update stat info if the entry is otherwise up-to-date (i.e., | |
168 | * the contents/mode haven't changed). This ensures that we reflect the | |
169 | * 'ctime' of the rename in the index without (incorrectly) updating | |
170 | * the cached stat info to reflect unstaged changes on disk. | |
171 | */ | |
172 | refreshed = refresh_cache_entry(istate, new_entry, CE_MATCH_REFRESH); | |
173 | if (refreshed && refreshed != new_entry) { | |
174 | add_index_entry(istate, refreshed, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); | |
175 | discard_cache_entry(new_entry); | |
176 | } else | |
177 | add_index_entry(istate, new_entry, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); | |
81dc2307 PB |
178 | } |
179 | ||
415e96c8 JH |
180 | /* |
181 | * This only updates the "non-critical" parts of the directory | |
182 | * cache, ie the parts that aren't tracked by GIT, and only used | |
183 | * to validate the cache. | |
184 | */ | |
d4c0a3ac | 185 | void fill_stat_cache_info(struct index_state *istate, struct cache_entry *ce, struct stat *st) |
415e96c8 | 186 | { |
c21d39d7 | 187 | fill_stat_data(&ce->ce_stat_data, st); |
5f73076c JH |
188 | |
189 | if (assume_unchanged) | |
7a51ed66 | 190 | ce->ce_flags |= CE_VALID; |
eadb5831 | 191 | |
883e248b | 192 | if (S_ISREG(st->st_mode)) { |
eadb5831 | 193 | ce_mark_uptodate(ce); |
b5a81697 | 194 | mark_fsmonitor_valid(istate, ce); |
883e248b | 195 | } |
415e96c8 JH |
196 | } |
197 | ||
c33fa871 JH |
198 | static unsigned int st_mode_from_ce(const struct cache_entry *ce) |
199 | { | |
200 | extern int trust_executable_bit, has_symlinks; | |
201 | ||
202 | switch (ce->ce_mode & S_IFMT) { | |
203 | case S_IFLNK: | |
204 | return has_symlinks ? S_IFLNK : (S_IFREG | 0644); | |
205 | case S_IFREG: | |
206 | return (ce->ce_mode & (trust_executable_bit ? 0755 : 0644)) | S_IFREG; | |
207 | case S_IFGITLINK: | |
208 | return S_IFDIR | 0755; | |
209 | case S_IFDIR: | |
210 | return ce->ce_mode; | |
211 | default: | |
212 | BUG("unsupported ce_mode: %o", ce->ce_mode); | |
213 | } | |
214 | } | |
215 | ||
216 | int fake_lstat(const struct cache_entry *ce, struct stat *st) | |
217 | { | |
218 | fake_lstat_data(&ce->ce_stat_data, st); | |
219 | st->st_mode = st_mode_from_ce(ce); | |
220 | ||
221 | /* always succeed as lstat() replacement */ | |
222 | return 0; | |
223 | } | |
224 | ||
58bf2a4c NTND |
225 | static int ce_compare_data(struct index_state *istate, |
226 | const struct cache_entry *ce, | |
227 | struct stat *st) | |
29e4d363 JH |
228 | { |
229 | int match = -1; | |
1b8ac5ea | 230 | int fd = git_open_cloexec(ce->name, O_RDONLY); |
29e4d363 JH |
231 | |
232 | if (fd >= 0) { | |
bebfecb9 | 233 | struct object_id oid; |
58bf2a4c | 234 | if (!index_fd(istate, &oid, fd, st, OBJ_BLOB, ce->name, 0)) |
6a29d7b7 | 235 | match = !oideq(&oid, &ce->oid); |
7f8508e8 | 236 | /* index_fd() closed the file descriptor already */ |
29e4d363 JH |
237 | } |
238 | return match; | |
239 | } | |
240 | ||
21a6b9fa | 241 | static int ce_compare_link(const struct cache_entry *ce, size_t expected_size) |
29e4d363 JH |
242 | { |
243 | int match = -1; | |
29e4d363 JH |
244 | void *buffer; |
245 | unsigned long size; | |
21666f1a | 246 | enum object_type type; |
a60272b3 | 247 | struct strbuf sb = STRBUF_INIT; |
29e4d363 | 248 | |
a60272b3 | 249 | if (strbuf_readlink(&sb, ce->name, expected_size)) |
29e4d363 | 250 | return -1; |
a60272b3 | 251 | |
bc726bd0 | 252 | buffer = repo_read_object_file(the_repository, &ce->oid, &type, &size); |
a60272b3 LT |
253 | if (buffer) { |
254 | if (size == sb.len) | |
255 | match = memcmp(buffer, sb.buf, size); | |
256 | free(buffer); | |
29e4d363 | 257 | } |
a60272b3 | 258 | strbuf_release(&sb); |
29e4d363 JH |
259 | return match; |
260 | } | |
261 | ||
21a6b9fa | 262 | static int ce_compare_gitlink(const struct cache_entry *ce) |
f35a6d3b | 263 | { |
1053fe82 | 264 | struct object_id oid; |
f35a6d3b LT |
265 | |
266 | /* | |
267 | * We don't actually require that the .git directory | |
302b9282 | 268 | * under GITLINK directory be a valid git directory. It |
f35a6d3b LT |
269 | * might even be missing (in case nobody populated that |
270 | * sub-project). | |
271 | * | |
272 | * If so, we consider it always to match. | |
273 | */ | |
a98e6101 | 274 | if (resolve_gitlink_ref(ce->name, "HEAD", &oid) < 0) |
f35a6d3b | 275 | return 0; |
6a29d7b7 | 276 | return !oideq(&oid, &ce->oid); |
f35a6d3b LT |
277 | } |
278 | ||
58bf2a4c NTND |
279 | static int ce_modified_check_fs(struct index_state *istate, |
280 | const struct cache_entry *ce, | |
281 | struct stat *st) | |
29e4d363 JH |
282 | { |
283 | switch (st->st_mode & S_IFMT) { | |
284 | case S_IFREG: | |
58bf2a4c | 285 | if (ce_compare_data(istate, ce, st)) |
29e4d363 JH |
286 | return DATA_CHANGED; |
287 | break; | |
288 | case S_IFLNK: | |
dc49cd76 | 289 | if (ce_compare_link(ce, xsize_t(st->st_size))) |
29e4d363 JH |
290 | return DATA_CHANGED; |
291 | break; | |
a8ee75bc | 292 | case S_IFDIR: |
7a51ed66 | 293 | if (S_ISGITLINK(ce->ce_mode)) |
c70115b4 | 294 | return ce_compare_gitlink(ce) ? DATA_CHANGED : 0; |
1cf01a34 | 295 | /* else fallthrough */ |
29e4d363 JH |
296 | default: |
297 | return TYPE_CHANGED; | |
298 | } | |
299 | return 0; | |
300 | } | |
301 | ||
21a6b9fa | 302 | static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st) |
734aab75 LT |
303 | { |
304 | unsigned int changed = 0; | |
305 | ||
7a51ed66 LT |
306 | if (ce->ce_flags & CE_REMOVE) |
307 | return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED; | |
308 | ||
309 | switch (ce->ce_mode & S_IFMT) { | |
8ae0a8c5 KS |
310 | case S_IFREG: |
311 | changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0; | |
3e09cdfd JH |
312 | /* We consider only the owner x bit to be relevant for |
313 | * "mode changes" | |
314 | */ | |
315 | if (trust_executable_bit && | |
7a51ed66 | 316 | (0100 & (ce->ce_mode ^ st->st_mode))) |
ffbe1add | 317 | changed |= MODE_CHANGED; |
8ae0a8c5 KS |
318 | break; |
319 | case S_IFLNK: | |
78a8d641 JS |
320 | if (!S_ISLNK(st->st_mode) && |
321 | (has_symlinks || !S_ISREG(st->st_mode))) | |
322 | changed |= TYPE_CHANGED; | |
8ae0a8c5 | 323 | break; |
302b9282 | 324 | case S_IFGITLINK: |
c70115b4 | 325 | /* We ignore most of the st_xxx fields for gitlinks */ |
f35a6d3b LT |
326 | if (!S_ISDIR(st->st_mode)) |
327 | changed |= TYPE_CHANGED; | |
328 | else if (ce_compare_gitlink(ce)) | |
329 | changed |= DATA_CHANGED; | |
a8ee75bc | 330 | return changed; |
8ae0a8c5 | 331 | default: |
391408e5 | 332 | BUG("unsupported ce_mode: %o", ce->ce_mode); |
8ae0a8c5 | 333 | } |
2cb45e95 | 334 | |
c21d39d7 | 335 | changed |= match_stat_data(&ce->ce_stat_data, st); |
b0391890 | 336 | |
f49c2c22 | 337 | /* Racily smudged entry? */ |
c21d39d7 | 338 | if (!ce->ce_stat_data.sd_size) { |
99d1a986 | 339 | if (!is_empty_blob_sha1(ce->oid.hash)) |
f49c2c22 LT |
340 | changed |= DATA_CHANGED; |
341 | } | |
342 | ||
407c8eb0 JH |
343 | return changed; |
344 | } | |
345 | ||
2bb4cda1 NTND |
346 | static int is_racy_stat(const struct index_state *istate, |
347 | const struct stat_data *sd) | |
6d91da6d | 348 | { |
2bb4cda1 | 349 | return (istate->timestamp.sec && |
fba2f38a KB |
350 | #ifdef USE_NSEC |
351 | /* nanosecond timestamped files can also be racy! */ | |
2bb4cda1 NTND |
352 | (istate->timestamp.sec < sd->sd_mtime.sec || |
353 | (istate->timestamp.sec == sd->sd_mtime.sec && | |
354 | istate->timestamp.nsec <= sd->sd_mtime.nsec)) | |
fba2f38a | 355 | #else |
2bb4cda1 | 356 | istate->timestamp.sec <= sd->sd_mtime.sec |
fba2f38a | 357 | #endif |
2bb4cda1 NTND |
358 | ); |
359 | } | |
360 | ||
5581a019 | 361 | int is_racy_timestamp(const struct index_state *istate, |
2bb4cda1 NTND |
362 | const struct cache_entry *ce) |
363 | { | |
364 | return (!S_ISGITLINK(ce->ce_mode) && | |
365 | is_racy_stat(istate, &ce->ce_stat_data)); | |
6d91da6d JH |
366 | } |
367 | ||
ed4efab1 NTND |
368 | int match_stat_data_racy(const struct index_state *istate, |
369 | const struct stat_data *sd, struct stat *st) | |
370 | { | |
371 | if (is_racy_stat(istate, sd)) | |
372 | return MTIME_CHANGED; | |
373 | return match_stat_data(sd, st); | |
6d91da6d JH |
374 | } |
375 | ||
883e248b | 376 | int ie_match_stat(struct index_state *istate, |
21a6b9fa | 377 | const struct cache_entry *ce, struct stat *st, |
4bd5b7da | 378 | unsigned int options) |
407c8eb0 | 379 | { |
5f73076c | 380 | unsigned int changed; |
4bd5b7da | 381 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 382 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
4bd5b7da | 383 | int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY; |
883e248b | 384 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
5f73076c | 385 | |
883e248b BP |
386 | if (!ignore_fsmonitor) |
387 | refresh_fsmonitor(istate); | |
5f73076c JH |
388 | /* |
389 | * If it's marked as always valid in the index, it's | |
390 | * valid whatever the checked-out copy says. | |
56cac48c NTND |
391 | * |
392 | * skip-worktree has the same effect with higher precedence | |
5f73076c | 393 | */ |
56cac48c NTND |
394 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) |
395 | return 0; | |
7a51ed66 | 396 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) |
5f73076c | 397 | return 0; |
883e248b BP |
398 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) |
399 | return 0; | |
5f73076c | 400 | |
331fcb59 JH |
401 | /* |
402 | * Intent-to-add entries have not been added, so the index entry | |
403 | * by definition never matches what is in the work tree until it | |
404 | * actually gets added. | |
405 | */ | |
895ff3b2 | 406 | if (ce_intent_to_add(ce)) |
331fcb59 JH |
407 | return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED; |
408 | ||
5f73076c | 409 | changed = ce_match_stat_basic(ce, st); |
407c8eb0 | 410 | |
29e4d363 JH |
411 | /* |
412 | * Within 1 second of this sequence: | |
413 | * echo xyzzy >file && git-update-index --add file | |
414 | * running this command: | |
415 | * echo frotz >file | |
416 | * would give a falsely clean cache entry. The mtime and | |
417 | * length match the cache, and other stat fields do not change. | |
418 | * | |
419 | * We could detect this at update-index time (the cache entry | |
420 | * being registered/updated records the same time as "now") | |
421 | * and delay the return from git-update-index, but that would | |
422 | * effectively mean we can make at most one commit per second, | |
423 | * which is not acceptable. Instead, we check cache entries | |
424 | * whose mtime are the same as the index file timestamp more | |
5f73076c | 425 | * carefully than others. |
29e4d363 | 426 | */ |
6d91da6d | 427 | if (!changed && is_racy_timestamp(istate, ce)) { |
42f77406 JH |
428 | if (assume_racy_is_modified) |
429 | changed |= DATA_CHANGED; | |
430 | else | |
58bf2a4c | 431 | changed |= ce_modified_check_fs(istate, ce, st); |
42f77406 | 432 | } |
b0391890 | 433 | |
29e4d363 | 434 | return changed; |
b0391890 JH |
435 | } |
436 | ||
883e248b | 437 | int ie_modified(struct index_state *istate, |
21a6b9fa RS |
438 | const struct cache_entry *ce, |
439 | struct stat *st, unsigned int options) | |
b0391890 | 440 | { |
29e4d363 | 441 | int changed, changed_fs; |
4bd5b7da JH |
442 | |
443 | changed = ie_match_stat(istate, ce, st, options); | |
b0391890 JH |
444 | if (!changed) |
445 | return 0; | |
b0391890 JH |
446 | /* |
447 | * If the mode or type has changed, there's no point in trying | |
448 | * to refresh the entry - it's not going to match | |
449 | */ | |
450 | if (changed & (MODE_CHANGED | TYPE_CHANGED)) | |
451 | return changed; | |
452 | ||
c70115b4 JH |
453 | /* |
454 | * Immediately after read-tree or update-index --cacheinfo, | |
455 | * the length field is zero, as we have never even read the | |
456 | * lstat(2) information once, and we cannot trust DATA_CHANGED | |
457 | * returned by ie_match_stat() which in turn was returned by | |
458 | * ce_match_stat_basic() to signal that the filesize of the | |
459 | * blob changed. We have to actually go to the filesystem to | |
460 | * see if the contents match, and if so, should answer "unchanged". | |
461 | * | |
462 | * The logic does not apply to gitlinks, as ce_match_stat_basic() | |
463 | * already has checked the actual HEAD from the filesystem in the | |
464 | * subproject. If ie_match_stat() already said it is different, | |
465 | * then we know it is. | |
b0391890 | 466 | */ |
c70115b4 | 467 | if ((changed & DATA_CHANGED) && |
c21d39d7 | 468 | (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0)) |
b0391890 JH |
469 | return changed; |
470 | ||
58bf2a4c | 471 | changed_fs = ce_modified_check_fs(istate, ce, st); |
29e4d363 JH |
472 | if (changed_fs) |
473 | return changed | changed_fs; | |
b0391890 JH |
474 | return 0; |
475 | } | |
476 | ||
aabc5617 EN |
477 | static int cache_name_stage_compare(const char *name1, int len1, int stage1, |
478 | const char *name2, int len2, int stage2) | |
958ba6c9 | 479 | { |
958ba6c9 LT |
480 | int cmp; |
481 | ||
ccdd4a0f | 482 | cmp = name_compare(name1, len1, name2, len2); |
0ab9e1e8 LT |
483 | if (cmp) |
484 | return cmp; | |
0ab9e1e8 | 485 | |
b60e188c | 486 | if (stage1 < stage2) |
eb38c22f | 487 | return -1; |
b60e188c | 488 | if (stage1 > stage2) |
eb38c22f | 489 | return 1; |
ccdd4a0f JM |
490 | return 0; |
491 | } | |
492 | ||
aabc5617 | 493 | int cmp_cache_name_compare(const void *a_, const void *b_) |
ccdd4a0f | 494 | { |
aabc5617 | 495 | const struct cache_entry *ce1, *ce2; |
ccdd4a0f | 496 | |
aabc5617 EN |
497 | ce1 = *((const struct cache_entry **)a_); |
498 | ce2 = *((const struct cache_entry **)b_); | |
499 | return cache_name_stage_compare(ce1->name, ce1->ce_namelen, ce_stage(ce1), | |
500 | ce2->name, ce2->ce_namelen, ce_stage(ce2)); | |
eb38c22f LT |
501 | } |
502 | ||
20ec2d03 VD |
503 | static int index_name_stage_pos(struct index_state *istate, |
504 | const char *name, int namelen, | |
505 | int stage, | |
506 | enum index_search_mode search_mode) | |
eb38c22f LT |
507 | { |
508 | int first, last; | |
509 | ||
510 | first = 0; | |
4aab5b46 | 511 | last = istate->cache_nr; |
eb38c22f | 512 | while (last > first) { |
568a05c5 | 513 | int next = first + ((last - first) >> 1); |
4aab5b46 | 514 | struct cache_entry *ce = istate->cache[next]; |
b60e188c | 515 | int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce)); |
eb38c22f | 516 | if (!cmp) |
76e7f4ec | 517 | return next; |
eb38c22f LT |
518 | if (cmp < 0) { |
519 | last = next; | |
520 | continue; | |
521 | } | |
522 | first = next+1; | |
523 | } | |
95e0321c | 524 | |
20ec2d03 | 525 | if (search_mode == EXPAND_SPARSE && istate->sparse_index && |
95e0321c DS |
526 | first > 0) { |
527 | /* Note: first <= istate->cache_nr */ | |
528 | struct cache_entry *ce = istate->cache[first - 1]; | |
529 | ||
530 | /* | |
531 | * If we are in a sparse-index _and_ the entry before the | |
532 | * insertion position is a sparse-directory entry that is | |
533 | * an ancestor of 'name', then we need to expand the index | |
534 | * and search again. This will only trigger once, because | |
535 | * thereafter the index is fully expanded. | |
536 | */ | |
537 | if (S_ISSPARSEDIR(ce->ce_mode) && | |
538 | ce_namelen(ce) < namelen && | |
539 | !strncmp(name, ce->name, ce_namelen(ce))) { | |
540 | ensure_full_index(istate); | |
20ec2d03 | 541 | return index_name_stage_pos(istate, name, namelen, stage, search_mode); |
95e0321c DS |
542 | } |
543 | } | |
544 | ||
76e7f4ec | 545 | return -first-1; |
eb38c22f LT |
546 | } |
547 | ||
847a9e5d | 548 | int index_name_pos(struct index_state *istate, const char *name, int namelen) |
b60e188c | 549 | { |
20ec2d03 VD |
550 | return index_name_stage_pos(istate, name, namelen, 0, EXPAND_SPARSE); |
551 | } | |
552 | ||
9553aa0f VD |
553 | int index_name_pos_sparse(struct index_state *istate, const char *name, int namelen) |
554 | { | |
555 | return index_name_stage_pos(istate, name, namelen, 0, NO_EXPAND_SPARSE); | |
556 | } | |
557 | ||
20ec2d03 VD |
558 | int index_entry_exists(struct index_state *istate, const char *name, int namelen) |
559 | { | |
560 | return index_name_stage_pos(istate, name, namelen, 0, NO_EXPAND_SPARSE) >= 0; | |
b60e188c TG |
561 | } |
562 | ||
4aab5b46 | 563 | int remove_index_entry_at(struct index_state *istate, int pos) |
7b937ca3 | 564 | { |
cf558704 LT |
565 | struct cache_entry *ce = istate->cache[pos]; |
566 | ||
cfc5789a | 567 | record_resolve_undo(istate, ce); |
2092678c | 568 | remove_name_hash(istate, ce); |
045113a5 | 569 | save_or_free_index_entry(istate, ce); |
e636a7b4 | 570 | istate->cache_changed |= CE_ENTRY_REMOVED; |
4aab5b46 JH |
571 | istate->cache_nr--; |
572 | if (pos >= istate->cache_nr) | |
7b937ca3 | 573 | return 0; |
f331ab9d RS |
574 | MOVE_ARRAY(istate->cache + pos, istate->cache + pos + 1, |
575 | istate->cache_nr - pos); | |
7b937ca3 LT |
576 | return 1; |
577 | } | |
578 | ||
36419c8e | 579 | /* |
98e023de | 580 | * Remove all cache entries marked for removal, that is where |
36419c8e KB |
581 | * CE_REMOVE is set in ce_flags. This is much more effective than |
582 | * calling remove_index_entry_at() for each entry to be removed. | |
583 | */ | |
6fdc2057 | 584 | void remove_marked_cache_entries(struct index_state *istate, int invalidate) |
36419c8e KB |
585 | { |
586 | struct cache_entry **ce_array = istate->cache; | |
587 | unsigned int i, j; | |
588 | ||
589 | for (i = j = 0; i < istate->cache_nr; i++) { | |
5699d17e | 590 | if (ce_array[i]->ce_flags & CE_REMOVE) { |
6fdc2057 TG |
591 | if (invalidate) { |
592 | cache_tree_invalidate_path(istate, | |
593 | ce_array[i]->name); | |
594 | untracked_cache_remove_from_index(istate, | |
595 | ce_array[i]->name); | |
596 | } | |
2092678c | 597 | remove_name_hash(istate, ce_array[i]); |
045113a5 | 598 | save_or_free_index_entry(istate, ce_array[i]); |
5699d17e | 599 | } |
36419c8e KB |
600 | else |
601 | ce_array[j++] = ce_array[i]; | |
602 | } | |
ad837d9e NTND |
603 | if (j == istate->cache_nr) |
604 | return; | |
e636a7b4 | 605 | istate->cache_changed |= CE_ENTRY_REMOVED; |
36419c8e KB |
606 | istate->cache_nr = j; |
607 | } | |
608 | ||
4aab5b46 | 609 | int remove_file_from_index(struct index_state *istate, const char *path) |
197ee8c9 | 610 | { |
4aab5b46 | 611 | int pos = index_name_pos(istate, path, strlen(path)); |
c4e3cca1 JH |
612 | if (pos < 0) |
613 | pos = -pos-1; | |
a5400efe | 614 | cache_tree_invalidate_path(istate, path); |
e931371a | 615 | untracked_cache_remove_from_index(istate, path); |
4aab5b46 JH |
616 | while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path)) |
617 | remove_index_entry_at(istate, pos); | |
197ee8c9 LT |
618 | return 0; |
619 | } | |
620 | ||
20314271 JS |
621 | static int compare_name(struct cache_entry *ce, const char *path, int namelen) |
622 | { | |
623 | return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen); | |
624 | } | |
625 | ||
626 | static int index_name_pos_also_unmerged(struct index_state *istate, | |
627 | const char *path, int namelen) | |
628 | { | |
629 | int pos = index_name_pos(istate, path, namelen); | |
630 | struct cache_entry *ce; | |
631 | ||
632 | if (pos >= 0) | |
633 | return pos; | |
634 | ||
635 | /* maybe unmerged? */ | |
636 | pos = -1 - pos; | |
637 | if (pos >= istate->cache_nr || | |
638 | compare_name((ce = istate->cache[pos]), path, namelen)) | |
639 | return -1; | |
640 | ||
641 | /* order of preference: stage 2, 1, 3 */ | |
642 | if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr && | |
643 | ce_stage((ce = istate->cache[pos + 1])) == 2 && | |
644 | !compare_name(ce, path, namelen)) | |
645 | pos++; | |
646 | return pos; | |
647 | } | |
648 | ||
1102952b LT |
649 | static int different_name(struct cache_entry *ce, struct cache_entry *alias) |
650 | { | |
651 | int len = ce_namelen(ce); | |
652 | return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len); | |
653 | } | |
654 | ||
655 | /* | |
656 | * If we add a filename that aliases in the cache, we will use the | |
657 | * name that we already have - but we don't want to update the same | |
658 | * alias twice, because that implies that there were actually two | |
659 | * different files with aliasing names! | |
660 | * | |
661 | * So we use the CE_ADDED flag to verify that the alias was an old | |
662 | * one before we accept it as | |
663 | */ | |
045113a5 NTND |
664 | static struct cache_entry *create_alias_ce(struct index_state *istate, |
665 | struct cache_entry *ce, | |
666 | struct cache_entry *alias) | |
1102952b LT |
667 | { |
668 | int len; | |
285c2e25 | 669 | struct cache_entry *new_entry; |
1102952b LT |
670 | |
671 | if (alias->ce_flags & CE_ADDED) | |
9d0a9e90 NTND |
672 | die(_("will not add file alias '%s' ('%s' already exists in index)"), |
673 | ce->name, alias->name); | |
1102952b LT |
674 | |
675 | /* Ok, create the new entry using the name of the existing alias */ | |
676 | len = ce_namelen(alias); | |
a849735b | 677 | new_entry = make_empty_cache_entry(istate, len); |
285c2e25 BW |
678 | memcpy(new_entry->name, alias->name, len); |
679 | copy_cache_entry(new_entry, ce); | |
045113a5 | 680 | save_or_free_index_entry(istate, ce); |
285c2e25 | 681 | return new_entry; |
1102952b LT |
682 | } |
683 | ||
b4b313f9 | 684 | void set_object_name_for_intent_to_add_entry(struct cache_entry *ce) |
39425819 | 685 | { |
a09c985e | 686 | struct object_id oid; |
c80d226a | 687 | if (write_object_file("", 0, OBJ_BLOB, &oid)) |
9d0a9e90 | 688 | die(_("cannot create an empty blob in the object database")); |
a09c985e | 689 | oidcpy(&ce->oid, &oid); |
39425819 JH |
690 | } |
691 | ||
610d55af | 692 | int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags) |
11be42a4 | 693 | { |
a849735b | 694 | int namelen, was_same; |
d177cab0 | 695 | mode_t st_mode = st->st_mode; |
9472935d | 696 | struct cache_entry *ce, *alias = NULL; |
56cac48c | 697 | unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY; |
38ed1d89 JH |
698 | int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND); |
699 | int pretend = flags & ADD_CACHE_PRETEND; | |
39425819 JH |
700 | int intent_only = flags & ADD_CACHE_INTENT; |
701 | int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE| | |
702 | (intent_only ? ADD_CACHE_NEW_ONLY : 0)); | |
e578d031 | 703 | unsigned hash_flags = pretend ? 0 : HASH_WRITE_OBJECT; |
f937bc2f | 704 | struct object_id oid; |
9472935d | 705 | |
9e5da3d0 JK |
706 | if (flags & ADD_CACHE_RENORMALIZE) |
707 | hash_flags |= HASH_RENORMALIZE; | |
11be42a4 | 708 | |
d177cab0 | 709 | if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode)) |
9d0a9e90 | 710 | return error(_("%s: can only add regular files, symbolic links or git-directories"), path); |
11be42a4 JS |
711 | |
712 | namelen = strlen(path); | |
d177cab0 | 713 | if (S_ISDIR(st_mode)) { |
f937bc2f KM |
714 | if (resolve_gitlink_ref(path, "HEAD", &oid) < 0) |
715 | return error(_("'%s' does not have a commit checked out"), path); | |
09595258 LT |
716 | while (namelen && path[namelen-1] == '/') |
717 | namelen--; | |
718 | } | |
a849735b | 719 | ce = make_empty_cache_entry(istate, namelen); |
11be42a4 | 720 | memcpy(ce->name, path, namelen); |
b60e188c | 721 | ce->ce_namelen = namelen; |
39425819 | 722 | if (!intent_only) |
d4c0a3ac | 723 | fill_stat_cache_info(istate, ce, st); |
388b2acd JH |
724 | else |
725 | ce->ce_flags |= CE_INTENT_TO_ADD; | |
11be42a4 | 726 | |
610d55af TG |
727 | |
728 | if (trust_executable_bit && has_symlinks) { | |
d177cab0 | 729 | ce->ce_mode = create_ce_mode(st_mode); |
610d55af | 730 | } else { |
78a8d641 JS |
731 | /* If there is an existing entry, pick the mode bits and type |
732 | * from it, otherwise assume unexecutable regular file. | |
11be42a4 | 733 | */ |
185c975f | 734 | struct cache_entry *ent; |
20314271 | 735 | int pos = index_name_pos_also_unmerged(istate, path, namelen); |
185c975f | 736 | |
4aab5b46 | 737 | ent = (0 <= pos) ? istate->cache[pos] : NULL; |
d177cab0 | 738 | ce->ce_mode = ce_mode_from_stat(ent, st_mode); |
11be42a4 JS |
739 | } |
740 | ||
dc1ae704 JJ |
741 | /* When core.ignorecase=true, determine if a directory of the same name but differing |
742 | * case already exists within the Git repository. If it does, ensure the directory | |
743 | * case of the file being added to the repository matches (is folded into) the existing | |
744 | * entry's directory case. | |
745 | */ | |
746 | if (ignore_case) { | |
41284eb0 | 747 | adjust_dirname_case(istate, ce->name); |
dc1ae704 | 748 | } |
e2c2a375 | 749 | if (!(flags & ADD_CACHE_RENORMALIZE)) { |
9472935d TB |
750 | alias = index_file_exists(istate, ce->name, |
751 | ce_namelen(ce), ignore_case); | |
752 | if (alias && | |
753 | !ce_stage(alias) && | |
754 | !ie_match_stat(istate, alias, st, ce_option)) { | |
755 | /* Nothing changed, really */ | |
756 | if (!S_ISGITLINK(alias->ce_mode)) | |
757 | ce_mark_uptodate(alias); | |
758 | alias->ce_flags |= CE_ADDED; | |
dc1ae704 | 759 | |
a849735b | 760 | discard_cache_entry(ce); |
9472935d TB |
761 | return 0; |
762 | } | |
0781b8a9 | 763 | } |
39425819 | 764 | if (!intent_only) { |
1c418243 | 765 | if (index_path(istate, &ce->oid, path, st, hash_flags)) { |
a849735b | 766 | discard_cache_entry(ce); |
9d0a9e90 | 767 | return error(_("unable to index file '%s'"), path); |
2d9426b0 | 768 | } |
39425819 | 769 | } else |
b4b313f9 | 770 | set_object_name_for_intent_to_add_entry(ce); |
39425819 | 771 | |
1102952b | 772 | if (ignore_case && alias && different_name(ce, alias)) |
045113a5 | 773 | ce = create_alias_ce(istate, ce, alias); |
1102952b | 774 | ce->ce_flags |= CE_ADDED; |
38ed1d89 | 775 | |
3bf0dd1f | 776 | /* It was suspected to be racily clean, but it turns out to be Ok */ |
38ed1d89 JH |
777 | was_same = (alias && |
778 | !ce_stage(alias) && | |
4a7e27e9 | 779 | oideq(&alias->oid, &ce->oid) && |
38ed1d89 JH |
780 | ce->ce_mode == alias->ce_mode); |
781 | ||
782 | if (pretend) | |
a849735b | 783 | discard_cache_entry(ce); |
067178ed | 784 | else if (add_index_entry(istate, ce, add_option)) { |
a849735b | 785 | discard_cache_entry(ce); |
9d0a9e90 | 786 | return error(_("unable to add '%s' to index"), path); |
067178ed | 787 | } |
38ed1d89 | 788 | if (verbose && !was_same) |
11be42a4 | 789 | printf("add '%s'\n", path); |
11be42a4 JS |
790 | return 0; |
791 | } | |
792 | ||
610d55af | 793 | int add_file_to_index(struct index_state *istate, const char *path, int flags) |
d177cab0 LT |
794 | { |
795 | struct stat st; | |
796 | if (lstat(path, &st)) | |
9d0a9e90 | 797 | die_errno(_("unable to stat '%s'"), path); |
610d55af | 798 | return add_to_index(istate, path, &st, flags); |
d177cab0 LT |
799 | } |
800 | ||
a849735b JM |
801 | struct cache_entry *make_empty_cache_entry(struct index_state *istate, size_t len) |
802 | { | |
8e72d675 | 803 | return mem_pool__ce_calloc(find_mem_pool(istate), len); |
a849735b JM |
804 | } |
805 | ||
96168827 MT |
806 | struct cache_entry *make_empty_transient_cache_entry(size_t len, |
807 | struct mem_pool *ce_mem_pool) | |
a849735b | 808 | { |
96168827 MT |
809 | if (ce_mem_pool) |
810 | return mem_pool__ce_calloc(ce_mem_pool, len); | |
a849735b JM |
811 | return xcalloc(1, cache_entry_size(len)); |
812 | } | |
813 | ||
2a1ae649 RS |
814 | enum verify_path_result { |
815 | PATH_OK, | |
816 | PATH_INVALID, | |
817 | PATH_DIR_WITH_SEP, | |
818 | }; | |
819 | ||
820 | static enum verify_path_result verify_path_internal(const char *, unsigned); | |
821 | ||
822 | int verify_path(const char *path, unsigned mode) | |
823 | { | |
c8ad9d04 | 824 | return verify_path_internal(path, mode) == PATH_OK; |
2a1ae649 RS |
825 | } |
826 | ||
a849735b JM |
827 | struct cache_entry *make_cache_entry(struct index_state *istate, |
828 | unsigned int mode, | |
825ed4d9 JM |
829 | const struct object_id *oid, |
830 | const char *path, | |
831 | int stage, | |
832 | unsigned int refresh_options) | |
6640f881 | 833 | { |
bc1c2caa | 834 | struct cache_entry *ce, *ret; |
a849735b | 835 | int len; |
6640f881 | 836 | |
c8ad9d04 | 837 | if (verify_path_internal(path, mode) == PATH_INVALID) { |
9d0a9e90 | 838 | error(_("invalid path '%s'"), path); |
6640f881 | 839 | return NULL; |
7e7abea9 | 840 | } |
6640f881 CR |
841 | |
842 | len = strlen(path); | |
a849735b | 843 | ce = make_empty_cache_entry(istate, len); |
6640f881 | 844 | |
825ed4d9 | 845 | oidcpy(&ce->oid, oid); |
6640f881 | 846 | memcpy(ce->name, path, len); |
b60e188c TG |
847 | ce->ce_flags = create_ce_flags(stage); |
848 | ce->ce_namelen = len; | |
6640f881 CR |
849 | ce->ce_mode = create_ce_mode(mode); |
850 | ||
d7b665c3 | 851 | ret = refresh_cache_entry(istate, ce, refresh_options); |
915e44c6 | 852 | if (ret != ce) |
a849735b | 853 | discard_cache_entry(ce); |
915e44c6 | 854 | return ret; |
6640f881 CR |
855 | } |
856 | ||
96168827 MT |
857 | struct cache_entry *make_transient_cache_entry(unsigned int mode, |
858 | const struct object_id *oid, | |
859 | const char *path, | |
860 | int stage, | |
861 | struct mem_pool *ce_mem_pool) | |
a849735b JM |
862 | { |
863 | struct cache_entry *ce; | |
864 | int len; | |
865 | ||
866 | if (!verify_path(path, mode)) { | |
9d0a9e90 | 867 | error(_("invalid path '%s'"), path); |
a849735b JM |
868 | return NULL; |
869 | } | |
870 | ||
871 | len = strlen(path); | |
96168827 | 872 | ce = make_empty_transient_cache_entry(len, ce_mem_pool); |
a849735b JM |
873 | |
874 | oidcpy(&ce->oid, oid); | |
875 | memcpy(ce->name, path, len); | |
876 | ce->ce_flags = create_ce_flags(stage); | |
877 | ce->ce_namelen = len; | |
878 | ce->ce_mode = create_ce_mode(mode); | |
879 | ||
880 | return ce; | |
881 | } | |
882 | ||
d9d70966 TG |
883 | /* |
884 | * Chmod an index entry with either +x or -x. | |
885 | * | |
886 | * Returns -1 if the chmod for the particular cache entry failed (if it's | |
887 | * not a regular file), -2 if an invalid flip argument is passed in, 0 | |
888 | * otherwise. | |
889 | */ | |
890 | int chmod_index_entry(struct index_state *istate, struct cache_entry *ce, | |
891 | char flip) | |
892 | { | |
893 | if (!S_ISREG(ce->ce_mode)) | |
894 | return -1; | |
895 | switch (flip) { | |
896 | case '+': | |
897 | ce->ce_mode |= 0111; | |
898 | break; | |
899 | case '-': | |
900 | ce->ce_mode &= ~0111; | |
901 | break; | |
902 | default: | |
903 | return -2; | |
904 | } | |
905 | cache_tree_invalidate_path(istate, ce->name); | |
906 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
883e248b | 907 | mark_fsmonitor_invalid(istate, ce); |
d9d70966 TG |
908 | istate->cache_changed |= CE_ENTRY_CHANGED; |
909 | ||
910 | return 0; | |
911 | } | |
912 | ||
9c5e6c80 | 913 | int ce_same_name(const struct cache_entry *a, const struct cache_entry *b) |
7b937ca3 LT |
914 | { |
915 | int len = ce_namelen(a); | |
916 | return ce_namelen(b) == len && !memcmp(a->name, b->name, len); | |
917 | } | |
918 | ||
8dcf39c4 LT |
919 | /* |
920 | * We fundamentally don't like some paths: we don't want | |
921 | * dot or dot-dot anywhere, and for obvious reasons don't | |
922 | * want to recurse into ".git" either. | |
923 | * | |
924 | * Also, we don't want double slashes or slashes at the | |
925 | * end that can make pathnames ambiguous. | |
926 | */ | |
10ecfa76 | 927 | static int verify_dotfile(const char *rest, unsigned mode) |
8dcf39c4 LT |
928 | { |
929 | /* | |
930 | * The first character was '.', but that | |
931 | * has already been discarded, we now test | |
932 | * the rest. | |
933 | */ | |
e0f530ff | 934 | |
8dcf39c4 | 935 | /* "." is not allowed */ |
e0f530ff | 936 | if (*rest == '\0' || is_dir_sep(*rest)) |
8dcf39c4 LT |
937 | return 0; |
938 | ||
e0f530ff | 939 | switch (*rest) { |
8dcf39c4 | 940 | /* |
641084b6 JK |
941 | * ".git" followed by NUL or slash is bad. Note that we match |
942 | * case-insensitively here, even if ignore_case is not set. | |
943 | * This outlaws ".GIT" everywhere out of an abundance of caution, | |
944 | * since there's really no good reason to allow it. | |
10ecfa76 JK |
945 | * |
946 | * Once we've seen ".git", we can also find ".gitmodules", etc (also | |
947 | * case-insensitively). | |
8dcf39c4 LT |
948 | */ |
949 | case 'g': | |
cc2fc7c2 JK |
950 | case 'G': |
951 | if (rest[1] != 'i' && rest[1] != 'I') | |
8dcf39c4 | 952 | break; |
cc2fc7c2 | 953 | if (rest[2] != 't' && rest[2] != 'T') |
8dcf39c4 | 954 | break; |
e19e5e66 JK |
955 | if (rest[3] == '\0' || is_dir_sep(rest[3])) |
956 | return 0; | |
10ecfa76 JK |
957 | if (S_ISLNK(mode)) { |
958 | rest += 3; | |
959 | if (skip_iprefix(rest, "modules", &rest) && | |
960 | (*rest == '\0' || is_dir_sep(*rest))) | |
961 | return 0; | |
962 | } | |
e19e5e66 | 963 | break; |
8dcf39c4 | 964 | case '.': |
e0f530ff | 965 | if (rest[1] == '\0' || is_dir_sep(rest[1])) |
8dcf39c4 LT |
966 | return 0; |
967 | } | |
968 | return 1; | |
969 | } | |
970 | ||
2a1ae649 RS |
971 | static enum verify_path_result verify_path_internal(const char *path, |
972 | unsigned mode) | |
8dcf39c4 | 973 | { |
49e268e2 | 974 | char c = 0; |
8dcf39c4 | 975 | |
56948cb6 | 976 | if (has_dos_drive_prefix(path)) |
2a1ae649 | 977 | return PATH_INVALID; |
56948cb6 | 978 | |
d2c84dad | 979 | if (!is_valid_path(path)) |
2a1ae649 | 980 | return PATH_INVALID; |
d2c84dad | 981 | |
8dcf39c4 LT |
982 | goto inside; |
983 | for (;;) { | |
984 | if (!c) | |
2a1ae649 | 985 | return PATH_OK; |
56948cb6 | 986 | if (is_dir_sep(c)) { |
8dcf39c4 | 987 | inside: |
10ecfa76 | 988 | if (protect_hfs) { |
49e268e2 | 989 | |
10ecfa76 | 990 | if (is_hfs_dotgit(path)) |
2a1ae649 | 991 | return PATH_INVALID; |
10ecfa76 JK |
992 | if (S_ISLNK(mode)) { |
993 | if (is_hfs_dotgitmodules(path)) | |
2a1ae649 | 994 | return PATH_INVALID; |
10ecfa76 JK |
995 | } |
996 | } | |
997 | if (protect_ntfs) { | |
bccc37fd | 998 | #if defined GIT_WINDOWS_NATIVE || defined __CYGWIN__ |
49e268e2 | 999 | if (c == '\\') |
2a1ae649 | 1000 | return PATH_INVALID; |
49e268e2 | 1001 | #endif |
10ecfa76 | 1002 | if (is_ntfs_dotgit(path)) |
2a1ae649 | 1003 | return PATH_INVALID; |
10ecfa76 JK |
1004 | if (S_ISLNK(mode)) { |
1005 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 1006 | return PATH_INVALID; |
10ecfa76 JK |
1007 | } |
1008 | } | |
1009 | ||
8dcf39c4 | 1010 | c = *path++; |
10ecfa76 | 1011 | if ((c == '.' && !verify_dotfile(path, mode)) || |
6e773527 | 1012 | is_dir_sep(c)) |
2a1ae649 | 1013 | return PATH_INVALID; |
6e773527 DS |
1014 | /* |
1015 | * allow terminating directory separators for | |
1016 | * sparse directory entries. | |
1017 | */ | |
1018 | if (c == '\0') | |
2a1ae649 RS |
1019 | return S_ISDIR(mode) ? PATH_DIR_WITH_SEP : |
1020 | PATH_INVALID; | |
288a74bc JS |
1021 | } else if (c == '\\' && protect_ntfs) { |
1022 | if (is_ntfs_dotgit(path)) | |
2a1ae649 | 1023 | return PATH_INVALID; |
288a74bc JS |
1024 | if (S_ISLNK(mode)) { |
1025 | if (is_ntfs_dotgitmodules(path)) | |
2a1ae649 | 1026 | return PATH_INVALID; |
288a74bc | 1027 | } |
8dcf39c4 | 1028 | } |
288a74bc | 1029 | |
8dcf39c4 LT |
1030 | c = *path++; |
1031 | } | |
1032 | } | |
1033 | ||
12676608 LT |
1034 | /* |
1035 | * Do we have another file that has the beginning components being a | |
1036 | * proper superset of the name we're trying to add? | |
0f1e4f04 | 1037 | */ |
4aab5b46 JH |
1038 | static int has_file_name(struct index_state *istate, |
1039 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
0f1e4f04 | 1040 | { |
12676608 LT |
1041 | int retval = 0; |
1042 | int len = ce_namelen(ce); | |
b155725d | 1043 | int stage = ce_stage(ce); |
12676608 | 1044 | const char *name = ce->name; |
0f1e4f04 | 1045 | |
4aab5b46 JH |
1046 | while (pos < istate->cache_nr) { |
1047 | struct cache_entry *p = istate->cache[pos++]; | |
0f1e4f04 | 1048 | |
12676608 | 1049 | if (len >= ce_namelen(p)) |
0f1e4f04 | 1050 | break; |
12676608 LT |
1051 | if (memcmp(name, p->name, len)) |
1052 | break; | |
b155725d JH |
1053 | if (ce_stage(p) != stage) |
1054 | continue; | |
12676608 LT |
1055 | if (p->name[len] != '/') |
1056 | continue; | |
7a51ed66 | 1057 | if (p->ce_flags & CE_REMOVE) |
21cd8d00 | 1058 | continue; |
12676608 LT |
1059 | retval = -1; |
1060 | if (!ok_to_replace) | |
1061 | break; | |
4aab5b46 | 1062 | remove_index_entry_at(istate, --pos); |
0f1e4f04 | 1063 | } |
12676608 LT |
1064 | return retval; |
1065 | } | |
0f1e4f04 | 1066 | |
a6db3fbb JH |
1067 | |
1068 | /* | |
1069 | * Like strcmp(), but also return the offset of the first change. | |
1070 | * If strings are equal, return the length. | |
1071 | */ | |
1072 | int strcmp_offset(const char *s1, const char *s2, size_t *first_change) | |
1073 | { | |
1074 | size_t k; | |
1075 | ||
1076 | if (!first_change) | |
1077 | return strcmp(s1, s2); | |
1078 | ||
1079 | for (k = 0; s1[k] == s2[k]; k++) | |
1080 | if (s1[k] == '\0') | |
1081 | break; | |
1082 | ||
1083 | *first_change = k; | |
1084 | return (unsigned char)s1[k] - (unsigned char)s2[k]; | |
1085 | } | |
1086 | ||
12676608 LT |
1087 | /* |
1088 | * Do we have another file with a pathname that is a proper | |
1089 | * subset of the name we're trying to add? | |
06b6d81b JH |
1090 | * |
1091 | * That is, is there another file in the index with a path | |
1092 | * that matches a sub-directory in the given entry? | |
12676608 | 1093 | */ |
4aab5b46 JH |
1094 | static int has_dir_name(struct index_state *istate, |
1095 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
12676608 LT |
1096 | { |
1097 | int retval = 0; | |
b155725d | 1098 | int stage = ce_stage(ce); |
12676608 LT |
1099 | const char *name = ce->name; |
1100 | const char *slash = name + ce_namelen(ce); | |
06b6d81b JH |
1101 | size_t len_eq_last; |
1102 | int cmp_last = 0; | |
1103 | ||
1104 | /* | |
1105 | * We are frequently called during an iteration on a sorted | |
1106 | * list of pathnames and while building a new index. Therefore, | |
1107 | * there is a high probability that this entry will eventually | |
1108 | * be appended to the index, rather than inserted in the middle. | |
1109 | * If we can confirm that, we can avoid binary searches on the | |
1110 | * components of the pathname. | |
1111 | * | |
1112 | * Compare the entry's full path with the last path in the index. | |
1113 | */ | |
1114 | if (istate->cache_nr > 0) { | |
1115 | cmp_last = strcmp_offset(name, | |
1116 | istate->cache[istate->cache_nr - 1]->name, | |
1117 | &len_eq_last); | |
1118 | if (cmp_last > 0) { | |
1119 | if (len_eq_last == 0) { | |
1120 | /* | |
1121 | * The entry sorts AFTER the last one in the | |
1122 | * index and their paths have no common prefix, | |
1123 | * so there cannot be a F/D conflict. | |
1124 | */ | |
1125 | return retval; | |
1126 | } else { | |
1127 | /* | |
1128 | * The entry sorts AFTER the last one in the | |
1129 | * index, but has a common prefix. Fall through | |
1130 | * to the loop below to disect the entry's path | |
1131 | * and see where the difference is. | |
1132 | */ | |
1133 | } | |
1134 | } else if (cmp_last == 0) { | |
1135 | /* | |
1136 | * The entry exactly matches the last one in the | |
1137 | * index, but because of multiple stage and CE_REMOVE | |
1138 | * items, we fall through and let the regular search | |
1139 | * code handle it. | |
1140 | */ | |
1141 | } | |
1142 | } | |
0f1e4f04 | 1143 | |
12676608 | 1144 | for (;;) { |
b986df5c | 1145 | size_t len; |
0f1e4f04 | 1146 | |
12676608 LT |
1147 | for (;;) { |
1148 | if (*--slash == '/') | |
1149 | break; | |
1150 | if (slash <= ce->name) | |
1151 | return retval; | |
1152 | } | |
1153 | len = slash - name; | |
0f1e4f04 | 1154 | |
b986df5c JH |
1155 | if (cmp_last > 0) { |
1156 | /* | |
1157 | * (len + 1) is a directory boundary (including | |
1158 | * the trailing slash). And since the loop is | |
1159 | * decrementing "slash", the first iteration is | |
1160 | * the longest directory prefix; subsequent | |
1161 | * iterations consider parent directories. | |
1162 | */ | |
1163 | ||
1164 | if (len + 1 <= len_eq_last) { | |
1165 | /* | |
1166 | * The directory prefix (including the trailing | |
1167 | * slash) also appears as a prefix in the last | |
1168 | * entry, so the remainder cannot collide (because | |
1169 | * strcmp said the whole path was greater). | |
1170 | * | |
1171 | * EQ: last: xxx/A | |
1172 | * this: xxx/B | |
1173 | * | |
1174 | * LT: last: xxx/file_A | |
1175 | * this: xxx/file_B | |
1176 | */ | |
1177 | return retval; | |
1178 | } | |
1179 | ||
1180 | if (len > len_eq_last) { | |
1181 | /* | |
1182 | * This part of the directory prefix (excluding | |
1183 | * the trailing slash) is longer than the known | |
1184 | * equal portions, so this sub-directory cannot | |
1185 | * collide with a file. | |
1186 | * | |
1187 | * GT: last: xxxA | |
1188 | * this: xxxB/file | |
1189 | */ | |
1190 | return retval; | |
1191 | } | |
1192 | ||
b986df5c JH |
1193 | /* |
1194 | * This is a possible collision. Fall through and | |
1195 | * let the regular search code handle it. | |
1196 | * | |
1197 | * last: xxx | |
1198 | * this: xxx/file | |
1199 | */ | |
1200 | } | |
1201 | ||
20ec2d03 | 1202 | pos = index_name_stage_pos(istate, name, len, stage, EXPAND_SPARSE); |
12676608 | 1203 | if (pos >= 0) { |
21cd8d00 JH |
1204 | /* |
1205 | * Found one, but not so fast. This could | |
1206 | * be a marker that says "I was here, but | |
1207 | * I am being removed". Such an entry is | |
1208 | * not a part of the resulting tree, and | |
1209 | * it is Ok to have a directory at the same | |
1210 | * path. | |
1211 | */ | |
077c48df | 1212 | if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) { |
21cd8d00 JH |
1213 | retval = -1; |
1214 | if (!ok_to_replace) | |
1215 | break; | |
4aab5b46 | 1216 | remove_index_entry_at(istate, pos); |
21cd8d00 JH |
1217 | continue; |
1218 | } | |
12676608 | 1219 | } |
21cd8d00 JH |
1220 | else |
1221 | pos = -pos-1; | |
12676608 LT |
1222 | |
1223 | /* | |
1224 | * Trivial optimization: if we find an entry that | |
1225 | * already matches the sub-directory, then we know | |
b155725d | 1226 | * we're ok, and we can exit. |
12676608 | 1227 | */ |
4aab5b46 JH |
1228 | while (pos < istate->cache_nr) { |
1229 | struct cache_entry *p = istate->cache[pos]; | |
b155725d JH |
1230 | if ((ce_namelen(p) <= len) || |
1231 | (p->name[len] != '/') || | |
1232 | memcmp(p->name, name, len)) | |
1233 | break; /* not our subdirectory */ | |
077c48df JH |
1234 | if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE)) |
1235 | /* | |
1236 | * p is at the same stage as our entry, and | |
b155725d JH |
1237 | * is a subdirectory of what we are looking |
1238 | * at, so we cannot have conflicts at our | |
1239 | * level or anything shorter. | |
1240 | */ | |
1241 | return retval; | |
1242 | pos++; | |
192268c1 | 1243 | } |
0f1e4f04 | 1244 | } |
12676608 LT |
1245 | return retval; |
1246 | } | |
1247 | ||
1248 | /* We may be in a situation where we already have path/file and path | |
1249 | * is being added, or we already have path and path/file is being | |
1250 | * added. Either one would result in a nonsense tree that has path | |
1251 | * twice when git-write-tree tries to write it out. Prevent it. | |
a6080a0a | 1252 | * |
12676608 LT |
1253 | * If ok-to-replace is specified, we remove the conflicting entries |
1254 | * from the cache so the caller should recompute the insert position. | |
1255 | * When this happens, we return non-zero. | |
1256 | */ | |
4aab5b46 JH |
1257 | static int check_file_directory_conflict(struct index_state *istate, |
1258 | const struct cache_entry *ce, | |
1259 | int pos, int ok_to_replace) | |
12676608 | 1260 | { |
21cd8d00 JH |
1261 | int retval; |
1262 | ||
1263 | /* | |
1264 | * When ce is an "I am going away" entry, we allow it to be added | |
1265 | */ | |
7a51ed66 | 1266 | if (ce->ce_flags & CE_REMOVE) |
21cd8d00 JH |
1267 | return 0; |
1268 | ||
12676608 LT |
1269 | /* |
1270 | * We check if the path is a sub-path of a subsequent pathname | |
1271 | * first, since removing those will not change the position | |
21cd8d00 | 1272 | * in the array. |
12676608 | 1273 | */ |
4aab5b46 | 1274 | retval = has_file_name(istate, ce, pos, ok_to_replace); |
21cd8d00 | 1275 | |
12676608 LT |
1276 | /* |
1277 | * Then check if the path might have a clashing sub-directory | |
1278 | * before it. | |
1279 | */ | |
4aab5b46 | 1280 | return retval + has_dir_name(istate, ce, pos, ok_to_replace); |
0f1e4f04 JH |
1281 | } |
1282 | ||
af3785dc | 1283 | static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option) |
197ee8c9 LT |
1284 | { |
1285 | int pos; | |
192268c1 JH |
1286 | int ok_to_add = option & ADD_CACHE_OK_TO_ADD; |
1287 | int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE; | |
b155725d | 1288 | int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK; |
39425819 | 1289 | int new_only = option & ADD_CACHE_NEW_ONLY; |
5f73076c | 1290 | |
e5494631 JH |
1291 | /* |
1292 | * If this entry's path sorts after the last entry in the index, | |
1293 | * we can avoid searching for it. | |
1294 | */ | |
1295 | if (istate->cache_nr > 0 && | |
1296 | strcmp(ce->name, istate->cache[istate->cache_nr - 1]->name) > 0) | |
c097b95a | 1297 | pos = index_pos_to_insert_pos(istate->cache_nr); |
e5494631 | 1298 | else |
20ec2d03 | 1299 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce), EXPAND_SPARSE); |
197ee8c9 | 1300 | |
c35e9f5e VD |
1301 | /* |
1302 | * Cache tree path should be invalidated only after index_name_stage_pos, | |
1303 | * in case it expands a sparse index. | |
1304 | */ | |
1305 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) | |
1306 | cache_tree_invalidate_path(istate, ce->name); | |
1307 | ||
3e09cdfd | 1308 | /* existing match? Just replace it. */ |
76e7f4ec | 1309 | if (pos >= 0) { |
39425819 JH |
1310 | if (!new_only) |
1311 | replace_index_entry(istate, pos, ce); | |
197ee8c9 LT |
1312 | return 0; |
1313 | } | |
76e7f4ec | 1314 | pos = -pos-1; |
197ee8c9 | 1315 | |
ffcc9ba7 NTND |
1316 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1317 | untracked_cache_add_to_index(istate, ce->name); | |
e931371a | 1318 | |
7b937ca3 LT |
1319 | /* |
1320 | * Inserting a merged entry ("stage 0") into the index | |
1321 | * will always replace all non-merged entries.. | |
1322 | */ | |
4aab5b46 JH |
1323 | if (pos < istate->cache_nr && ce_stage(ce) == 0) { |
1324 | while (ce_same_name(istate->cache[pos], ce)) { | |
7b937ca3 | 1325 | ok_to_add = 1; |
4aab5b46 | 1326 | if (!remove_index_entry_at(istate, pos)) |
7b937ca3 LT |
1327 | break; |
1328 | } | |
1329 | } | |
1330 | ||
121481ab LT |
1331 | if (!ok_to_add) |
1332 | return -1; | |
c8ad9d04 | 1333 | if (verify_path_internal(ce->name, ce->ce_mode) == PATH_INVALID) |
9d0a9e90 | 1334 | return error(_("invalid path '%s'"), ce->name); |
121481ab | 1335 | |
3e09cdfd | 1336 | if (!skip_df_check && |
4aab5b46 | 1337 | check_file_directory_conflict(istate, ce, pos, ok_to_replace)) { |
192268c1 | 1338 | if (!ok_to_replace) |
9d0a9e90 | 1339 | return error(_("'%s' appears as both a file and as a directory"), |
4aab5b46 | 1340 | ce->name); |
20ec2d03 | 1341 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce), EXPAND_SPARSE); |
192268c1 JH |
1342 | pos = -pos-1; |
1343 | } | |
af3785dc JH |
1344 | return pos + 1; |
1345 | } | |
1346 | ||
1347 | int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option) | |
1348 | { | |
1349 | int pos; | |
1350 | ||
1351 | if (option & ADD_CACHE_JUST_APPEND) | |
1352 | pos = istate->cache_nr; | |
1353 | else { | |
1354 | int ret; | |
1355 | ret = add_index_entry_with_check(istate, ce, option); | |
1356 | if (ret <= 0) | |
1357 | return ret; | |
1358 | pos = ret - 1; | |
1359 | } | |
0f1e4f04 | 1360 | |
197ee8c9 | 1361 | /* Make sure the array is big enough .. */ |
999f5660 | 1362 | ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc); |
197ee8c9 LT |
1363 | |
1364 | /* Add it in.. */ | |
4aab5b46 | 1365 | istate->cache_nr++; |
af3785dc | 1366 | if (istate->cache_nr > pos + 1) |
f919ffeb SG |
1367 | MOVE_ARRAY(istate->cache + pos + 1, istate->cache + pos, |
1368 | istate->cache_nr - pos - 1); | |
cf558704 | 1369 | set_index_entry(istate, pos, ce); |
e636a7b4 | 1370 | istate->cache_changed |= CE_ENTRY_ADDED; |
197ee8c9 LT |
1371 | return 0; |
1372 | } | |
1373 | ||
405e5b2f LT |
1374 | /* |
1375 | * "refresh" does not calculate a new sha1 file or bring the | |
1376 | * cache up-to-date for mode/content changes. But what it | |
1377 | * _does_ do is to "re-match" the stat information of a file | |
1378 | * with the cache, so that you can refresh the cache for a | |
1379 | * file that hasn't been changed but where the stat entry is | |
1380 | * out of date. | |
1381 | * | |
1382 | * For example, you'd want to do this after doing a "git-read-tree", | |
1383 | * to link up the stat cache details with the proper files. | |
1384 | */ | |
4aab5b46 | 1385 | static struct cache_entry *refresh_cache_ent(struct index_state *istate, |
4bd5b7da | 1386 | struct cache_entry *ce, |
d05e6970 | 1387 | unsigned int options, int *err, |
a98e0f2d | 1388 | int *changed_ret, |
15268d12 JH |
1389 | int *t2_did_lstat, |
1390 | int *t2_did_scan) | |
405e5b2f LT |
1391 | { |
1392 | struct stat st; | |
1393 | struct cache_entry *updated; | |
a849735b | 1394 | int changed; |
25762726 | 1395 | int refresh = options & CE_MATCH_REFRESH; |
4bd5b7da | 1396 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 1397 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
2e2e7ec1 | 1398 | int ignore_missing = options & CE_MATCH_IGNORE_MISSING; |
883e248b | 1399 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
405e5b2f | 1400 | |
25762726 | 1401 | if (!refresh || ce_uptodate(ce)) |
eadb5831 JH |
1402 | return ce; |
1403 | ||
883e248b BP |
1404 | if (!ignore_fsmonitor) |
1405 | refresh_fsmonitor(istate); | |
aa9349d4 | 1406 | /* |
56cac48c NTND |
1407 | * CE_VALID or CE_SKIP_WORKTREE means the user promised us |
1408 | * that the change to the work tree does not matter and told | |
1409 | * us not to worry. | |
aa9349d4 | 1410 | */ |
56cac48c NTND |
1411 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) { |
1412 | ce_mark_uptodate(ce); | |
1413 | return ce; | |
1414 | } | |
aa9349d4 MSO |
1415 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) { |
1416 | ce_mark_uptodate(ce); | |
1417 | return ce; | |
1418 | } | |
883e248b BP |
1419 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) { |
1420 | ce_mark_uptodate(ce); | |
1421 | return ce; | |
1422 | } | |
aa9349d4 | 1423 | |
ccad42d4 RS |
1424 | if (has_symlink_leading_path(ce->name, ce_namelen(ce))) { |
1425 | if (ignore_missing) | |
1426 | return ce; | |
1427 | if (err) | |
1428 | *err = ENOENT; | |
1429 | return NULL; | |
1430 | } | |
1431 | ||
a98e0f2d JH |
1432 | if (t2_did_lstat) |
1433 | *t2_did_lstat = 1; | |
8fd2cb40 | 1434 | if (lstat(ce->name, &st) < 0) { |
2e2e7ec1 BK |
1435 | if (ignore_missing && errno == ENOENT) |
1436 | return ce; | |
ec0cc704 JH |
1437 | if (err) |
1438 | *err = errno; | |
8fd2cb40 JS |
1439 | return NULL; |
1440 | } | |
405e5b2f | 1441 | |
4bd5b7da | 1442 | changed = ie_match_stat(istate, ce, &st, options); |
d05e6970 JK |
1443 | if (changed_ret) |
1444 | *changed_ret = changed; | |
405e5b2f | 1445 | if (!changed) { |
4bd5b7da JH |
1446 | /* |
1447 | * The path is unchanged. If we were told to ignore | |
1448 | * valid bit, then we did the actual stat check and | |
1449 | * found that the entry is unmodified. If the entry | |
1450 | * is not marked VALID, this is the place to mark it | |
1451 | * valid again, under "assume unchanged" mode. | |
1452 | */ | |
1453 | if (ignore_valid && assume_unchanged && | |
7a51ed66 | 1454 | !(ce->ce_flags & CE_VALID)) |
405e5b2f | 1455 | ; /* mark this one VALID again */ |
eadb5831 JH |
1456 | else { |
1457 | /* | |
1458 | * We do not mark the index itself "modified" | |
1459 | * because CE_UPTODATE flag is in-core only; | |
1460 | * we are not going to write this change out. | |
1461 | */ | |
883e248b | 1462 | if (!S_ISGITLINK(ce->ce_mode)) { |
125fd984 | 1463 | ce_mark_uptodate(ce); |
b5a81697 | 1464 | mark_fsmonitor_valid(istate, ce); |
883e248b | 1465 | } |
8fd2cb40 | 1466 | return ce; |
eadb5831 | 1467 | } |
405e5b2f LT |
1468 | } |
1469 | ||
15268d12 JH |
1470 | if (t2_did_scan) |
1471 | *t2_did_scan = 1; | |
4bd5b7da | 1472 | if (ie_modified(istate, ce, &st, options)) { |
ec0cc704 JH |
1473 | if (err) |
1474 | *err = EINVAL; | |
8fd2cb40 JS |
1475 | return NULL; |
1476 | } | |
405e5b2f | 1477 | |
a849735b | 1478 | updated = make_empty_cache_entry(istate, ce_namelen(ce)); |
0e267b7a BP |
1479 | copy_cache_entry(updated, ce); |
1480 | memcpy(updated->name, ce->name, ce->ce_namelen + 1); | |
d4c0a3ac | 1481 | fill_stat_cache_info(istate, updated, &st); |
4bd5b7da JH |
1482 | /* |
1483 | * If ignore_valid is not set, we should leave CE_VALID bit | |
1484 | * alone. Otherwise, paths marked with --no-assume-unchanged | |
1485 | * (i.e. things to be edited) will reacquire CE_VALID bit | |
1486 | * automatically, which is not really what we want. | |
405e5b2f | 1487 | */ |
4bd5b7da | 1488 | if (!ignore_valid && assume_unchanged && |
7a51ed66 LT |
1489 | !(ce->ce_flags & CE_VALID)) |
1490 | updated->ce_flags &= ~CE_VALID; | |
405e5b2f | 1491 | |
e636a7b4 | 1492 | /* istate->cache_changed is updated in the caller */ |
405e5b2f LT |
1493 | return updated; |
1494 | } | |
1495 | ||
3deffc52 | 1496 | static void show_file(const char * fmt, const char * name, int in_porcelain, |
046613c5 | 1497 | int * first, const char *header_msg) |
3deffc52 MM |
1498 | { |
1499 | if (in_porcelain && *first && header_msg) { | |
1500 | printf("%s\n", header_msg); | |
cd2b8ae9 | 1501 | *first = 0; |
3deffc52 MM |
1502 | } |
1503 | printf(fmt, name); | |
1504 | } | |
1505 | ||
22184497 TG |
1506 | int repo_refresh_and_write_index(struct repository *repo, |
1507 | unsigned int refresh_flags, | |
1508 | unsigned int write_flags, | |
1509 | int gentle, | |
1510 | const struct pathspec *pathspec, | |
1511 | char *seen, const char *header_msg) | |
1512 | { | |
1513 | struct lock_file lock_file = LOCK_INIT; | |
1514 | int fd, ret = 0; | |
1515 | ||
1516 | fd = repo_hold_locked_index(repo, &lock_file, 0); | |
1517 | if (!gentle && fd < 0) | |
1518 | return -1; | |
1519 | if (refresh_index(repo->index, refresh_flags, pathspec, seen, header_msg)) | |
1520 | ret = 1; | |
1521 | if (0 <= fd && write_locked_index(repo->index, &lock_file, COMMIT_LOCK | write_flags)) | |
1522 | ret = -1; | |
1523 | return ret; | |
1524 | } | |
1525 | ||
1526 | ||
9b2d6149 NTND |
1527 | int refresh_index(struct index_state *istate, unsigned int flags, |
1528 | const struct pathspec *pathspec, | |
046613c5 | 1529 | char *seen, const char *header_msg) |
405e5b2f LT |
1530 | { |
1531 | int i; | |
1532 | int has_errors = 0; | |
1533 | int really = (flags & REFRESH_REALLY) != 0; | |
1534 | int allow_unmerged = (flags & REFRESH_UNMERGED) != 0; | |
1535 | int quiet = (flags & REFRESH_QUIET) != 0; | |
1536 | int not_new = (flags & REFRESH_IGNORE_MISSING) != 0; | |
5fdeacb0 | 1537 | int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0; |
b243012c | 1538 | int ignore_skip_worktree = (flags & REFRESH_IGNORE_SKIP_WORKTREE) != 0; |
3deffc52 MM |
1539 | int first = 1; |
1540 | int in_porcelain = (flags & REFRESH_IN_PORCELAIN); | |
25762726 BK |
1541 | unsigned int options = (CE_MATCH_REFRESH | |
1542 | (really ? CE_MATCH_IGNORE_VALID : 0) | | |
2e2e7ec1 | 1543 | (not_new ? CE_MATCH_IGNORE_MISSING : 0)); |
4bd4e730 | 1544 | const char *modified_fmt; |
73b7eae6 JK |
1545 | const char *deleted_fmt; |
1546 | const char *typechange_fmt; | |
1547 | const char *added_fmt; | |
4bd4e730 | 1548 | const char *unmerged_fmt; |
ae9af122 | 1549 | struct progress *progress = NULL; |
a98e0f2d | 1550 | int t2_sum_lstat = 0; |
15268d12 | 1551 | int t2_sum_scan = 0; |
ae9af122 NTND |
1552 | |
1553 | if (flags & REFRESH_PROGRESS && isatty(2)) | |
1554 | progress = start_delayed_progress(_("Refresh index"), | |
1555 | istate->cache_nr); | |
405e5b2f | 1556 | |
c46c406a | 1557 | trace_performance_enter(); |
a71806a7 NTND |
1558 | modified_fmt = in_porcelain ? "M\t%s\n" : "%s: needs update\n"; |
1559 | deleted_fmt = in_porcelain ? "D\t%s\n" : "%s: needs update\n"; | |
1560 | typechange_fmt = in_porcelain ? "T\t%s\n" : "%s: needs update\n"; | |
1561 | added_fmt = in_porcelain ? "A\t%s\n" : "%s: needs update\n"; | |
1562 | unmerged_fmt = in_porcelain ? "U\t%s\n" : "%s: needs merge\n"; | |
99ce720c BP |
1563 | /* |
1564 | * Use the multi-threaded preload_index() to refresh most of the | |
1565 | * cache entries quickly then in the single threaded loop below, | |
1566 | * we only have to do the special cases that are left. | |
1567 | */ | |
1568 | preload_index(istate, pathspec, 0); | |
a98e0f2d | 1569 | trace2_region_enter("index", "refresh", NULL); |
d76723ee | 1570 | |
4aab5b46 | 1571 | for (i = 0; i < istate->cache_nr; i++) { |
285c2e25 | 1572 | struct cache_entry *ce, *new_entry; |
ec0cc704 | 1573 | int cache_errno = 0; |
73b7eae6 | 1574 | int changed = 0; |
3d1f148c | 1575 | int filtered = 0; |
a98e0f2d | 1576 | int t2_did_lstat = 0; |
15268d12 | 1577 | int t2_did_scan = 0; |
ec0cc704 | 1578 | |
4aab5b46 | 1579 | ce = istate->cache[i]; |
5fdeacb0 JS |
1580 | if (ignore_submodules && S_ISGITLINK(ce->ce_mode)) |
1581 | continue; | |
b243012c MT |
1582 | if (ignore_skip_worktree && ce_skip_worktree(ce)) |
1583 | continue; | |
5fdeacb0 | 1584 | |
d76723ee DS |
1585 | /* |
1586 | * If this entry is a sparse directory, then there isn't | |
1587 | * any stat() information to update. Ignore the entry. | |
1588 | */ | |
1589 | if (S_ISSPARSEDIR(ce->ce_mode)) | |
1590 | continue; | |
1591 | ||
d7b665c3 | 1592 | if (pathspec && !ce_path_match(istate, ce, pathspec, seen)) |
3d1f148c JH |
1593 | filtered = 1; |
1594 | ||
405e5b2f | 1595 | if (ce_stage(ce)) { |
4aab5b46 JH |
1596 | while ((i < istate->cache_nr) && |
1597 | ! strcmp(istate->cache[i]->name, ce->name)) | |
405e5b2f LT |
1598 | i++; |
1599 | i--; | |
1600 | if (allow_unmerged) | |
1601 | continue; | |
3d1f148c JH |
1602 | if (!filtered) |
1603 | show_file(unmerged_fmt, ce->name, in_porcelain, | |
1604 | &first, header_msg); | |
405e5b2f LT |
1605 | has_errors = 1; |
1606 | continue; | |
1607 | } | |
1608 | ||
3d1f148c | 1609 | if (filtered) |
d616813d AJ |
1610 | continue; |
1611 | ||
a98e0f2d JH |
1612 | new_entry = refresh_cache_ent(istate, ce, options, |
1613 | &cache_errno, &changed, | |
15268d12 | 1614 | &t2_did_lstat, &t2_did_scan); |
a98e0f2d | 1615 | t2_sum_lstat += t2_did_lstat; |
15268d12 | 1616 | t2_sum_scan += t2_did_scan; |
285c2e25 | 1617 | if (new_entry == ce) |
405e5b2f | 1618 | continue; |
b7b793d1 | 1619 | display_progress(progress, i); |
285c2e25 | 1620 | if (!new_entry) { |
73b7eae6 JK |
1621 | const char *fmt; |
1622 | ||
8fd2cb40 | 1623 | if (really && cache_errno == EINVAL) { |
405e5b2f LT |
1624 | /* If we are doing --really-refresh that |
1625 | * means the index is not valid anymore. | |
1626 | */ | |
7a51ed66 | 1627 | ce->ce_flags &= ~CE_VALID; |
078a58e8 | 1628 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 1629 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 1630 | istate->cache_changed |= CE_ENTRY_CHANGED; |
405e5b2f LT |
1631 | } |
1632 | if (quiet) | |
1633 | continue; | |
73b7eae6 JK |
1634 | |
1635 | if (cache_errno == ENOENT) | |
1636 | fmt = deleted_fmt; | |
895ff3b2 | 1637 | else if (ce_intent_to_add(ce)) |
73b7eae6 JK |
1638 | fmt = added_fmt; /* must be before other checks */ |
1639 | else if (changed & TYPE_CHANGED) | |
1640 | fmt = typechange_fmt; | |
1641 | else | |
1642 | fmt = modified_fmt; | |
1643 | show_file(fmt, | |
1644 | ce->name, in_porcelain, &first, header_msg); | |
405e5b2f LT |
1645 | has_errors = 1; |
1646 | continue; | |
1647 | } | |
cf558704 | 1648 | |
285c2e25 | 1649 | replace_index_entry(istate, i, new_entry); |
405e5b2f | 1650 | } |
a98e0f2d | 1651 | trace2_data_intmax("index", NULL, "refresh/sum_lstat", t2_sum_lstat); |
15268d12 | 1652 | trace2_data_intmax("index", NULL, "refresh/sum_scan", t2_sum_scan); |
a98e0f2d | 1653 | trace2_region_leave("index", "refresh", NULL); |
b7b793d1 ÆAB |
1654 | display_progress(progress, istate->cache_nr); |
1655 | stop_progress(&progress); | |
c46c406a | 1656 | trace_performance_leave("refresh index"); |
405e5b2f LT |
1657 | return has_errors; |
1658 | } | |
1659 | ||
768d7965 JM |
1660 | struct cache_entry *refresh_cache_entry(struct index_state *istate, |
1661 | struct cache_entry *ce, | |
1662 | unsigned int options) | |
ec0cc704 | 1663 | { |
15268d12 | 1664 | return refresh_cache_ent(istate, ce, options, NULL, NULL, NULL, NULL); |
ec0cc704 JH |
1665 | } |
1666 | ||
db3b313c JH |
1667 | |
1668 | /***************************************************************** | |
1669 | * Index File I/O | |
1670 | *****************************************************************/ | |
1671 | ||
9d227781 JH |
1672 | #define INDEX_FORMAT_DEFAULT 3 |
1673 | ||
7211b9e7 | 1674 | static unsigned int get_index_format_default(struct repository *r) |
136347d7 TG |
1675 | { |
1676 | char *envversion = getenv("GIT_INDEX_VERSION"); | |
3c09d684 TG |
1677 | char *endp; |
1678 | unsigned int version = INDEX_FORMAT_DEFAULT; | |
1679 | ||
136347d7 | 1680 | if (!envversion) { |
7211b9e7 DS |
1681 | prepare_repo_settings(r); |
1682 | ||
1683 | if (r->settings.index_version >= 0) | |
1684 | version = r->settings.index_version; | |
3c09d684 TG |
1685 | if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { |
1686 | warning(_("index.version set, but the value is invalid.\n" | |
136347d7 | 1687 | "Using version %i"), INDEX_FORMAT_DEFAULT); |
3c09d684 | 1688 | return INDEX_FORMAT_DEFAULT; |
136347d7 TG |
1689 | } |
1690 | return version; | |
1691 | } | |
3c09d684 TG |
1692 | |
1693 | version = strtoul(envversion, &endp, 10); | |
1694 | if (*endp || | |
1695 | version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { | |
1696 | warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n" | |
1697 | "Using version %i"), INDEX_FORMAT_DEFAULT); | |
1698 | version = INDEX_FORMAT_DEFAULT; | |
1699 | } | |
1700 | return version; | |
136347d7 TG |
1701 | } |
1702 | ||
db3b313c JH |
1703 | /* |
1704 | * dev/ino/uid/gid/size are also just tracked to the low 32 bits | |
1705 | * Again - this is just a (very strong in practice) heuristic that | |
1706 | * the inode hasn't changed. | |
1707 | * | |
1708 | * We save the fields in big-endian order to allow using the | |
1709 | * index file over NFS transparently. | |
1710 | */ | |
1711 | struct ondisk_cache_entry { | |
1712 | struct cache_time ctime; | |
1713 | struct cache_time mtime; | |
7800c1eb TG |
1714 | uint32_t dev; |
1715 | uint32_t ino; | |
1716 | uint32_t mode; | |
1717 | uint32_t uid; | |
1718 | uint32_t gid; | |
1719 | uint32_t size; | |
575fa8a3 | 1720 | /* |
1721 | * unsigned char hash[hashsz]; | |
1722 | * uint16_t flags; | |
1723 | * if (flags & CE_EXTENDED) | |
1724 | * uint16_t flags2; | |
1725 | */ | |
1726 | unsigned char data[GIT_MAX_RAWSZ + 2 * sizeof(uint16_t)]; | |
1727 | char name[FLEX_ARRAY]; | |
db3b313c JH |
1728 | }; |
1729 | ||
6c9cd161 | 1730 | /* These are only used for v3 or lower */ |
ce012deb | 1731 | #define align_padding_size(size, len) ((size + (len) + 8) & ~7) - (size + len) |
575fa8a3 | 1732 | #define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,data) + (len) + 8) & ~7) |
db3b313c | 1733 | #define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len) |
575fa8a3 | 1734 | #define ondisk_data_size(flags, len) (the_hash_algo->rawsz + \ |
1735 | ((flags & CE_EXTENDED) ? 2 : 1) * sizeof(uint16_t) + len) | |
1736 | #define ondisk_data_size_max(len) (ondisk_data_size(CE_EXTENDED, len)) | |
1737 | #define ondisk_ce_size(ce) (ondisk_cache_entry_size(ondisk_data_size((ce)->ce_flags, ce_namelen(ce)))) | |
db3b313c | 1738 | |
a33fc72f JH |
1739 | /* Allow fsck to force verification of the index checksum. */ |
1740 | int verify_index_checksum; | |
1741 | ||
00ec50e5 BP |
1742 | /* Allow fsck to force verification of the cache entry order. */ |
1743 | int verify_ce_order; | |
1744 | ||
371ed0de | 1745 | static int verify_hdr(const struct cache_header *hdr, unsigned long size) |
e83c5163 | 1746 | { |
aab61359 | 1747 | git_hash_ctx c; |
1748 | unsigned char hash[GIT_MAX_RAWSZ]; | |
0136bac9 | 1749 | int hdr_version; |
ee1f0c24 DS |
1750 | unsigned char *start, *end; |
1751 | struct object_id oid; | |
e83c5163 | 1752 | |
ccc4feb5 | 1753 | if (hdr->hdr_signature != htonl(CACHE_SIGNATURE)) |
9d0a9e90 | 1754 | return error(_("bad signature 0x%08x"), hdr->hdr_signature); |
0136bac9 | 1755 | hdr_version = ntohl(hdr->hdr_version); |
b82a7b5b | 1756 | if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version) |
9d0a9e90 | 1757 | return error(_("bad index version %d"), hdr_version); |
a33fc72f JH |
1758 | |
1759 | if (!verify_index_checksum) | |
1760 | return 0; | |
1761 | ||
ee1f0c24 DS |
1762 | end = (unsigned char *)hdr + size; |
1763 | start = end - the_hash_algo->rawsz; | |
1764 | oidread(&oid, start); | |
1765 | if (oideq(&oid, null_oid())) | |
1766 | return 0; | |
1767 | ||
aab61359 | 1768 | the_hash_algo->init_fn(&c); |
1769 | the_hash_algo->update_fn(&c, hdr, size - the_hash_algo->rawsz); | |
1770 | the_hash_algo->final_fn(hash, &c); | |
ee1f0c24 | 1771 | if (!hasheq(hash, start)) |
9d0a9e90 | 1772 | return error(_("bad index file sha1 signature")); |
e83c5163 LT |
1773 | return 0; |
1774 | } | |
1775 | ||
4aab5b46 | 1776 | static int read_index_extension(struct index_state *istate, |
371ed0de | 1777 | const char *ext, const char *data, unsigned long sz) |
bad68ec9 JH |
1778 | { |
1779 | switch (CACHE_EXT(ext)) { | |
1780 | case CACHE_EXT_TREE: | |
4aab5b46 | 1781 | istate->cache_tree = cache_tree_read(data, sz); |
bad68ec9 | 1782 | break; |
cfc5789a JH |
1783 | case CACHE_EXT_RESOLVE_UNDO: |
1784 | istate->resolve_undo = resolve_undo_read(data, sz); | |
1785 | break; | |
5fc2fc8f NTND |
1786 | case CACHE_EXT_LINK: |
1787 | if (read_link_extension(istate, data, sz)) | |
1788 | return -1; | |
1789 | break; | |
f9e6c649 NTND |
1790 | case CACHE_EXT_UNTRACKED: |
1791 | istate->untracked = read_untracked_extension(data, sz); | |
1792 | break; | |
883e248b BP |
1793 | case CACHE_EXT_FSMONITOR: |
1794 | read_fsmonitor_extension(istate, data, sz); | |
1795 | break; | |
3b1d9e04 | 1796 | case CACHE_EXT_ENDOFINDEXENTRIES: |
3255089a | 1797 | case CACHE_EXT_INDEXENTRYOFFSETTABLE: |
3b1d9e04 BP |
1798 | /* already handled in do_read_index() */ |
1799 | break; | |
cd42415f DS |
1800 | case CACHE_EXT_SPARSE_DIRECTORIES: |
1801 | /* no content, only an indicator */ | |
9fadb373 | 1802 | istate->sparse_index = INDEX_COLLAPSED; |
cd42415f | 1803 | break; |
bad68ec9 JH |
1804 | default: |
1805 | if (*ext < 'A' || 'Z' < *ext) | |
9d0a9e90 | 1806 | return error(_("index uses %.4s extension, which we do not understand"), |
bad68ec9 | 1807 | ext); |
9d0a9e90 | 1808 | fprintf_ln(stderr, _("ignoring %.4s extension"), ext); |
bad68ec9 JH |
1809 | break; |
1810 | } | |
1811 | return 0; | |
1812 | } | |
1813 | ||
4a6ed30f VD |
1814 | /* |
1815 | * Parses the contents of the cache entry contained within the 'ondisk' buffer | |
1816 | * into a new incore 'cache_entry'. | |
1817 | * | |
1818 | * Note that 'char *ondisk' may not be aligned to a 4-byte address interval in | |
1819 | * index v4, so we cannot cast it to 'struct ondisk_cache_entry *' and access | |
1820 | * its members. Instead, we use the byte offsets of members within the struct to | |
1821 | * identify where 'get_be16()', 'get_be32()', and 'oidread()' (which can all | |
1822 | * read from an unaligned memory buffer) should read from the 'ondisk' buffer | |
1823 | * into the corresponding incore 'cache_entry' members. | |
1824 | */ | |
77ff1127 BP |
1825 | static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool, |
1826 | unsigned int version, | |
4a6ed30f | 1827 | const char *ondisk, |
6c9cd161 | 1828 | unsigned long *ent_size, |
252d079c | 1829 | const struct cache_entry *previous_ce) |
7a51ed66 | 1830 | { |
debed2a6 | 1831 | struct cache_entry *ce; |
7fec10b7 | 1832 | size_t len; |
06aaaa0b | 1833 | const char *name; |
575fa8a3 | 1834 | const unsigned hashsz = the_hash_algo->rawsz; |
4a6ed30f | 1835 | const char *flagsp = ondisk + offsetof(struct ondisk_cache_entry, data) + hashsz; |
debed2a6 | 1836 | unsigned int flags; |
f5c4a9af | 1837 | size_t copy_len = 0; |
252d079c NTND |
1838 | /* |
1839 | * Adjacent cache entries tend to share the leading paths, so it makes | |
1840 | * sense to only store the differences in later entries. In the v4 | |
1841 | * on-disk format of the index, each on-disk cache entry stores the | |
1842 | * number of bytes to be stripped from the end of the previous name, | |
1843 | * and the bytes to append to the result, to come up with its name. | |
1844 | */ | |
77ff1127 | 1845 | int expand_name_field = version == 4; |
7fec10b7 | 1846 | |
7a51ed66 | 1847 | /* On-disk flags are just 16 bits */ |
575fa8a3 | 1848 | flags = get_be16(flagsp); |
debed2a6 | 1849 | len = flags & CE_NAMEMASK; |
7fec10b7 | 1850 | |
debed2a6 | 1851 | if (flags & CE_EXTENDED) { |
06aaaa0b | 1852 | int extended_flags; |
4a6ed30f | 1853 | extended_flags = get_be16(flagsp + sizeof(uint16_t)) << 16; |
06aaaa0b NTND |
1854 | /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */ |
1855 | if (extended_flags & ~CE_EXTENDED_FLAGS) | |
9d0a9e90 | 1856 | die(_("unknown index entry format 0x%08x"), extended_flags); |
debed2a6 | 1857 | flags |= extended_flags; |
4a6ed30f | 1858 | name = (const char *)(flagsp + 2 * sizeof(uint16_t)); |
06aaaa0b NTND |
1859 | } |
1860 | else | |
4a6ed30f | 1861 | name = (const char *)(flagsp + sizeof(uint16_t)); |
06aaaa0b | 1862 | |
252d079c NTND |
1863 | if (expand_name_field) { |
1864 | const unsigned char *cp = (const unsigned char *)name; | |
1865 | size_t strip_len, previous_len; | |
6c9cd161 | 1866 | |
15beaaa3 | 1867 | /* If we're at the beginning of a block, ignore the previous name */ |
252d079c | 1868 | strip_len = decode_varint(&cp); |
77ff1127 BP |
1869 | if (previous_ce) { |
1870 | previous_len = previous_ce->ce_namelen; | |
1871 | if (previous_len < strip_len) | |
252d079c | 1872 | die(_("malformed name field in the index, near path '%s'"), |
77ff1127 BP |
1873 | previous_ce->name); |
1874 | copy_len = previous_len - strip_len; | |
252d079c | 1875 | } |
252d079c NTND |
1876 | name = (const char *)cp; |
1877 | } | |
6c9cd161 | 1878 | |
252d079c NTND |
1879 | if (len == CE_NAMEMASK) { |
1880 | len = strlen(name); | |
1881 | if (expand_name_field) | |
1882 | len += copy_len; | |
1883 | } | |
6c9cd161 | 1884 | |
77ff1127 | 1885 | ce = mem_pool__ce_alloc(ce_mem_pool, len); |
252d079c | 1886 | |
4a6ed30f VD |
1887 | /* |
1888 | * NEEDSWORK: using 'offsetof()' is cumbersome and should be replaced | |
1889 | * with something more akin to 'load_bitmap_entries_v1()'s use of | |
1890 | * 'read_be16'/'read_be32'. For consistency with the corresponding | |
1891 | * ondisk entry write function ('copy_cache_entry_to_ondisk()'), this | |
1892 | * should be done at the same time as removing references to | |
1893 | * 'ondisk_cache_entry' there. | |
1894 | */ | |
1895 | ce->ce_stat_data.sd_ctime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime) | |
1896 | + offsetof(struct cache_time, sec)); | |
1897 | ce->ce_stat_data.sd_mtime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime) | |
1898 | + offsetof(struct cache_time, sec)); | |
1899 | ce->ce_stat_data.sd_ctime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime) | |
1900 | + offsetof(struct cache_time, nsec)); | |
1901 | ce->ce_stat_data.sd_mtime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime) | |
1902 | + offsetof(struct cache_time, nsec)); | |
1903 | ce->ce_stat_data.sd_dev = get_be32(ondisk + offsetof(struct ondisk_cache_entry, dev)); | |
1904 | ce->ce_stat_data.sd_ino = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ino)); | |
1905 | ce->ce_mode = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mode)); | |
1906 | ce->ce_stat_data.sd_uid = get_be32(ondisk + offsetof(struct ondisk_cache_entry, uid)); | |
1907 | ce->ce_stat_data.sd_gid = get_be32(ondisk + offsetof(struct ondisk_cache_entry, gid)); | |
1908 | ce->ce_stat_data.sd_size = get_be32(ondisk + offsetof(struct ondisk_cache_entry, size)); | |
252d079c NTND |
1909 | ce->ce_flags = flags & ~CE_NAMEMASK; |
1910 | ce->ce_namelen = len; | |
1911 | ce->index = 0; | |
4a6ed30f | 1912 | oidread(&ce->oid, (const unsigned char *)ondisk + offsetof(struct ondisk_cache_entry, data)); |
6c9cd161 | 1913 | |
252d079c NTND |
1914 | if (expand_name_field) { |
1915 | if (copy_len) | |
1916 | memcpy(ce->name, previous_ce->name, copy_len); | |
1917 | memcpy(ce->name + copy_len, name, len + 1 - copy_len); | |
1918 | *ent_size = (name - ((char *)ondisk)) + len + 1 - copy_len; | |
1919 | } else { | |
1920 | memcpy(ce->name, name, len + 1); | |
1921 | *ent_size = ondisk_ce_size(ce); | |
6c9cd161 | 1922 | } |
debed2a6 | 1923 | return ce; |
cf558704 LT |
1924 | } |
1925 | ||
03f15a79 | 1926 | static void check_ce_order(struct index_state *istate) |
15999d0b | 1927 | { |
03f15a79 TG |
1928 | unsigned int i; |
1929 | ||
00ec50e5 BP |
1930 | if (!verify_ce_order) |
1931 | return; | |
1932 | ||
03f15a79 TG |
1933 | for (i = 1; i < istate->cache_nr; i++) { |
1934 | struct cache_entry *ce = istate->cache[i - 1]; | |
1935 | struct cache_entry *next_ce = istate->cache[i]; | |
1936 | int name_compare = strcmp(ce->name, next_ce->name); | |
1937 | ||
1938 | if (0 < name_compare) | |
9d0a9e90 | 1939 | die(_("unordered stage entries in index")); |
03f15a79 TG |
1940 | if (!name_compare) { |
1941 | if (!ce_stage(ce)) | |
9d0a9e90 | 1942 | die(_("multiple stage entries for merged file '%s'"), |
03f15a79 TG |
1943 | ce->name); |
1944 | if (ce_stage(ce) > ce_stage(next_ce)) | |
9d0a9e90 | 1945 | die(_("unordered stage entries for '%s'"), |
03f15a79 TG |
1946 | ce->name); |
1947 | } | |
15999d0b JSP |
1948 | } |
1949 | } | |
1950 | ||
435ec090 CC |
1951 | static void tweak_untracked_cache(struct index_state *istate) |
1952 | { | |
ad0fb659 DS |
1953 | struct repository *r = the_repository; |
1954 | ||
1955 | prepare_repo_settings(r); | |
1956 | ||
f1bee828 ÆAB |
1957 | switch (r->settings.core_untracked_cache) { |
1958 | case UNTRACKED_CACHE_REMOVE: | |
435ec090 | 1959 | remove_untracked_cache(istate); |
f1bee828 ÆAB |
1960 | break; |
1961 | case UNTRACKED_CACHE_WRITE: | |
ad0fb659 | 1962 | add_untracked_cache(istate); |
f1bee828 ÆAB |
1963 | break; |
1964 | case UNTRACKED_CACHE_KEEP: | |
3050b6df ÆAB |
1965 | /* |
1966 | * Either an explicit "core.untrackedCache=keep", the | |
1967 | * default if "core.untrackedCache" isn't configured, | |
1968 | * or a fallback on an unknown "core.untrackedCache" | |
1969 | * value. | |
1970 | */ | |
f1bee828 | 1971 | break; |
f1bee828 | 1972 | } |
435ec090 CC |
1973 | } |
1974 | ||
43925312 CC |
1975 | static void tweak_split_index(struct index_state *istate) |
1976 | { | |
1977 | switch (git_config_get_split_index()) { | |
1978 | case -1: /* unset: do nothing */ | |
1979 | break; | |
1980 | case 0: /* false */ | |
1981 | remove_split_index(istate); | |
1982 | break; | |
1983 | case 1: /* true */ | |
1984 | add_split_index(istate); | |
1985 | break; | |
1986 | default: /* unknown value: do nothing */ | |
1987 | break; | |
1988 | } | |
1989 | } | |
1990 | ||
435ec090 CC |
1991 | static void post_read_index_from(struct index_state *istate) |
1992 | { | |
1993 | check_ce_order(istate); | |
1994 | tweak_untracked_cache(istate); | |
43925312 | 1995 | tweak_split_index(istate); |
883e248b | 1996 | tweak_fsmonitor(istate); |
435ec090 CC |
1997 | } |
1998 | ||
8e72d675 JM |
1999 | static size_t estimate_cache_size_from_compressed(unsigned int entries) |
2000 | { | |
2001 | return entries * (sizeof(struct cache_entry) + CACHE_ENTRY_PATH_LENGTH); | |
2002 | } | |
2003 | ||
2004 | static size_t estimate_cache_size(size_t ondisk_size, unsigned int entries) | |
2005 | { | |
2006 | long per_entry = sizeof(struct cache_entry) - sizeof(struct ondisk_cache_entry); | |
2007 | ||
2008 | /* | |
2009 | * Account for potential alignment differences. | |
2010 | */ | |
c097b95a | 2011 | per_entry += align_padding_size(per_entry, 0); |
8e72d675 JM |
2012 | return ondisk_size + entries * per_entry; |
2013 | } | |
2014 | ||
3255089a BP |
2015 | struct index_entry_offset |
2016 | { | |
2017 | /* starting byte offset into index file, count of index entries in this block */ | |
2018 | int offset, nr; | |
2019 | }; | |
2020 | ||
2021 | struct index_entry_offset_table | |
2022 | { | |
2023 | int nr; | |
2024 | struct index_entry_offset entries[FLEX_ARRAY]; | |
2025 | }; | |
2026 | ||
3255089a BP |
2027 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset); |
2028 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot); | |
3255089a | 2029 | |
3b1d9e04 BP |
2030 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size); |
2031 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset); | |
2032 | ||
abb4bb83 BP |
2033 | struct load_index_extensions |
2034 | { | |
abb4bb83 | 2035 | pthread_t pthread; |
abb4bb83 BP |
2036 | struct index_state *istate; |
2037 | const char *mmap; | |
2038 | size_t mmap_size; | |
2039 | unsigned long src_offset; | |
2040 | }; | |
2041 | ||
2042 | static void *load_index_extensions(void *_data) | |
2043 | { | |
2044 | struct load_index_extensions *p = _data; | |
2045 | unsigned long src_offset = p->src_offset; | |
2046 | ||
2047 | while (src_offset <= p->mmap_size - the_hash_algo->rawsz - 8) { | |
2048 | /* After an array of active_nr index entries, | |
2049 | * there can be arbitrary number of extended | |
2050 | * sections, each of which is prefixed with | |
2051 | * extension name (4-byte) and section length | |
2052 | * in 4-byte network byte order. | |
2053 | */ | |
2054 | uint32_t extsize = get_be32(p->mmap + src_offset + 4); | |
2055 | if (read_index_extension(p->istate, | |
2056 | p->mmap + src_offset, | |
2057 | p->mmap + src_offset + 8, | |
2058 | extsize) < 0) { | |
2059 | munmap((void *)p->mmap, p->mmap_size); | |
2060 | die(_("index file corrupt")); | |
2061 | } | |
2062 | src_offset += 8; | |
2063 | src_offset += extsize; | |
2064 | } | |
2065 | ||
2066 | return NULL; | |
2067 | } | |
2068 | ||
77ff1127 BP |
2069 | /* |
2070 | * A helper function that will load the specified range of cache entries | |
2071 | * from the memory mapped file and add them to the given index. | |
2072 | */ | |
2073 | static unsigned long load_cache_entry_block(struct index_state *istate, | |
2074 | struct mem_pool *ce_mem_pool, int offset, int nr, const char *mmap, | |
2075 | unsigned long start_offset, const struct cache_entry *previous_ce) | |
2076 | { | |
2077 | int i; | |
2078 | unsigned long src_offset = start_offset; | |
2079 | ||
2080 | for (i = offset; i < offset + nr; i++) { | |
77ff1127 BP |
2081 | struct cache_entry *ce; |
2082 | unsigned long consumed; | |
2083 | ||
4a6ed30f VD |
2084 | ce = create_from_disk(ce_mem_pool, istate->version, |
2085 | mmap + src_offset, | |
2086 | &consumed, previous_ce); | |
77ff1127 BP |
2087 | set_index_entry(istate, i, ce); |
2088 | ||
2089 | src_offset += consumed; | |
2090 | previous_ce = ce; | |
2091 | } | |
2092 | return src_offset - start_offset; | |
2093 | } | |
2094 | ||
2095 | static unsigned long load_all_cache_entries(struct index_state *istate, | |
2096 | const char *mmap, size_t mmap_size, unsigned long src_offset) | |
2097 | { | |
2098 | unsigned long consumed; | |
2099 | ||
44c7e1a7 | 2100 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2101 | if (istate->version == 4) { |
44c7e1a7 | 2102 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2103 | estimate_cache_size_from_compressed(istate->cache_nr)); |
2104 | } else { | |
44c7e1a7 | 2105 | mem_pool_init(istate->ce_mem_pool, |
77ff1127 BP |
2106 | estimate_cache_size(mmap_size, istate->cache_nr)); |
2107 | } | |
2108 | ||
2109 | consumed = load_cache_entry_block(istate, istate->ce_mem_pool, | |
2110 | 0, istate->cache_nr, mmap, src_offset, NULL); | |
2111 | return consumed; | |
2112 | } | |
2113 | ||
3255089a BP |
2114 | /* |
2115 | * Mostly randomly chosen maximum thread counts: we | |
2116 | * cap the parallelism to online_cpus() threads, and we want | |
2117 | * to have at least 10000 cache entries per thread for it to | |
2118 | * be worth starting a thread. | |
2119 | */ | |
2120 | ||
2121 | #define THREAD_COST (10000) | |
2122 | ||
77ff1127 BP |
2123 | struct load_cache_entries_thread_data |
2124 | { | |
2125 | pthread_t pthread; | |
2126 | struct index_state *istate; | |
2127 | struct mem_pool *ce_mem_pool; | |
2128 | int offset; | |
2129 | const char *mmap; | |
2130 | struct index_entry_offset_table *ieot; | |
2131 | int ieot_start; /* starting index into the ieot array */ | |
2132 | int ieot_blocks; /* count of ieot entries to process */ | |
2133 | unsigned long consumed; /* return # of bytes in index file processed */ | |
2134 | }; | |
2135 | ||
2136 | /* | |
2137 | * A thread proc to run the load_cache_entries() computation | |
2138 | * across multiple background threads. | |
2139 | */ | |
2140 | static void *load_cache_entries_thread(void *_data) | |
2141 | { | |
2142 | struct load_cache_entries_thread_data *p = _data; | |
2143 | int i; | |
2144 | ||
2145 | /* iterate across all ieot blocks assigned to this thread */ | |
2146 | for (i = p->ieot_start; i < p->ieot_start + p->ieot_blocks; i++) { | |
2147 | p->consumed += load_cache_entry_block(p->istate, p->ce_mem_pool, | |
2148 | p->offset, p->ieot->entries[i].nr, p->mmap, p->ieot->entries[i].offset, NULL); | |
2149 | p->offset += p->ieot->entries[i].nr; | |
2150 | } | |
2151 | return NULL; | |
2152 | } | |
2153 | ||
2154 | static unsigned long load_cache_entries_threaded(struct index_state *istate, const char *mmap, size_t mmap_size, | |
7bd9631b | 2155 | int nr_threads, struct index_entry_offset_table *ieot) |
77ff1127 BP |
2156 | { |
2157 | int i, offset, ieot_blocks, ieot_start, err; | |
2158 | struct load_cache_entries_thread_data *data; | |
2159 | unsigned long consumed = 0; | |
2160 | ||
2161 | /* a little sanity checking */ | |
2162 | if (istate->name_hash_initialized) | |
2163 | BUG("the name hash isn't thread safe"); | |
2164 | ||
44c7e1a7 EN |
2165 | istate->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
2166 | mem_pool_init(istate->ce_mem_pool, 0); | |
77ff1127 BP |
2167 | |
2168 | /* ensure we have no more threads than we have blocks to process */ | |
2169 | if (nr_threads > ieot->nr) | |
2170 | nr_threads = ieot->nr; | |
ca56dadb | 2171 | CALLOC_ARRAY(data, nr_threads); |
77ff1127 BP |
2172 | |
2173 | offset = ieot_start = 0; | |
2174 | ieot_blocks = DIV_ROUND_UP(ieot->nr, nr_threads); | |
2175 | for (i = 0; i < nr_threads; i++) { | |
2176 | struct load_cache_entries_thread_data *p = &data[i]; | |
2177 | int nr, j; | |
2178 | ||
2179 | if (ieot_start + ieot_blocks > ieot->nr) | |
2180 | ieot_blocks = ieot->nr - ieot_start; | |
2181 | ||
2182 | p->istate = istate; | |
2183 | p->offset = offset; | |
2184 | p->mmap = mmap; | |
2185 | p->ieot = ieot; | |
2186 | p->ieot_start = ieot_start; | |
2187 | p->ieot_blocks = ieot_blocks; | |
2188 | ||
2189 | /* create a mem_pool for each thread */ | |
2190 | nr = 0; | |
2191 | for (j = p->ieot_start; j < p->ieot_start + p->ieot_blocks; j++) | |
2192 | nr += p->ieot->entries[j].nr; | |
bcd2c5ee | 2193 | p->ce_mem_pool = xmalloc(sizeof(*istate->ce_mem_pool)); |
77ff1127 | 2194 | if (istate->version == 4) { |
44c7e1a7 | 2195 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2196 | estimate_cache_size_from_compressed(nr)); |
2197 | } else { | |
44c7e1a7 | 2198 | mem_pool_init(p->ce_mem_pool, |
77ff1127 BP |
2199 | estimate_cache_size(mmap_size, nr)); |
2200 | } | |
2201 | ||
2202 | err = pthread_create(&p->pthread, NULL, load_cache_entries_thread, p); | |
2203 | if (err) | |
2204 | die(_("unable to create load_cache_entries thread: %s"), strerror(err)); | |
2205 | ||
2206 | /* increment by the number of cache entries in the ieot block being processed */ | |
2207 | for (j = 0; j < ieot_blocks; j++) | |
2208 | offset += ieot->entries[ieot_start + j].nr; | |
2209 | ieot_start += ieot_blocks; | |
2210 | } | |
2211 | ||
2212 | for (i = 0; i < nr_threads; i++) { | |
2213 | struct load_cache_entries_thread_data *p = &data[i]; | |
2214 | ||
2215 | err = pthread_join(p->pthread, NULL); | |
2216 | if (err) | |
2217 | die(_("unable to join load_cache_entries thread: %s"), strerror(err)); | |
2218 | mem_pool_combine(istate->ce_mem_pool, p->ce_mem_pool); | |
2219 | consumed += p->consumed; | |
2220 | } | |
2221 | ||
2222 | free(data); | |
2223 | ||
2224 | return consumed; | |
2225 | } | |
77ff1127 | 2226 | |
491df5f6 VD |
2227 | static void set_new_index_sparsity(struct index_state *istate) |
2228 | { | |
2229 | /* | |
2230 | * If the index's repo exists, mark it sparse according to | |
2231 | * repo settings. | |
2232 | */ | |
5bdf6d4a ÆAB |
2233 | prepare_repo_settings(istate->repo); |
2234 | if (!istate->repo->settings.command_requires_full_index && | |
2235 | is_sparse_index_allowed(istate, 0)) | |
2236 | istate->sparse_index = 1; | |
491df5f6 VD |
2237 | } |
2238 | ||
8fd2cb40 | 2239 | /* remember to discard_cache() before reading a different cache! */ |
3e52f70b | 2240 | int do_read_index(struct index_state *istate, const char *path, int must_exist) |
e83c5163 | 2241 | { |
77ff1127 | 2242 | int fd; |
e83c5163 | 2243 | struct stat st; |
debed2a6 | 2244 | unsigned long src_offset; |
371ed0de BP |
2245 | const struct cache_header *hdr; |
2246 | const char *mmap; | |
7a51ed66 | 2247 | size_t mmap_size; |
abb4bb83 BP |
2248 | struct load_index_extensions p; |
2249 | size_t extension_offset = 0; | |
77ff1127 BP |
2250 | int nr_threads, cpus; |
2251 | struct index_entry_offset_table *ieot = NULL; | |
e83c5163 | 2252 | |
913e0e99 | 2253 | if (istate->initialized) |
4aab5b46 | 2254 | return istate->cache_nr; |
5d1a5c02 | 2255 | |
fba2f38a KB |
2256 | istate->timestamp.sec = 0; |
2257 | istate->timestamp.nsec = 0; | |
8fd2cb40 | 2258 | fd = open(path, O_RDONLY); |
5d1a5c02 | 2259 | if (fd < 0) { |
491df5f6 VD |
2260 | if (!must_exist && errno == ENOENT) { |
2261 | set_new_index_sparsity(istate); | |
866b43e6 | 2262 | istate->initialized = 1; |
5d1a5c02 | 2263 | return 0; |
491df5f6 | 2264 | } |
9d0a9e90 | 2265 | die_errno(_("%s: index file open failed"), path); |
5d1a5c02 | 2266 | } |
e83c5163 | 2267 | |
3511a377 | 2268 | if (fstat(fd, &st)) |
9d0a9e90 | 2269 | die_errno(_("%s: cannot stat the open index"), path); |
3511a377 | 2270 | |
7a51ed66 | 2271 | mmap_size = xsize_t(st.st_size); |
aab61359 | 2272 | if (mmap_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
9d0a9e90 | 2273 | die(_("%s: index file smaller than expected"), path); |
3511a377 | 2274 | |
02638d1e | 2275 | mmap = xmmap_gently(NULL, mmap_size, PROT_READ, MAP_PRIVATE, fd, 0); |
7a51ed66 | 2276 | if (mmap == MAP_FAILED) |
dc059294 EW |
2277 | die_errno(_("%s: unable to map index file%s"), path, |
2278 | mmap_os_err()); | |
57d84f8d | 2279 | close(fd); |
e83c5163 | 2280 | |
371ed0de | 2281 | hdr = (const struct cache_header *)mmap; |
7a51ed66 | 2282 | if (verify_hdr(hdr, mmap_size) < 0) |
e83c5163 LT |
2283 | goto unmap; |
2284 | ||
92e2cab9 | 2285 | oidread(&istate->oid, (const unsigned char *)hdr + mmap_size - the_hash_algo->rawsz); |
9d227781 | 2286 | istate->version = ntohl(hdr->hdr_version); |
4aab5b46 JH |
2287 | istate->cache_nr = ntohl(hdr->hdr_entries); |
2288 | istate->cache_alloc = alloc_nr(istate->cache_nr); | |
ca56dadb | 2289 | CALLOC_ARRAY(istate->cache, istate->cache_alloc); |
913e0e99 | 2290 | istate->initialized = 1; |
7a51ed66 | 2291 | |
abb4bb83 BP |
2292 | p.istate = istate; |
2293 | p.mmap = mmap; | |
2294 | p.mmap_size = mmap_size; | |
6c9cd161 | 2295 | |
7a51ed66 | 2296 | src_offset = sizeof(*hdr); |
4aab5b46 | 2297 | |
2a9dedef JN |
2298 | if (git_config_get_index_threads(&nr_threads)) |
2299 | nr_threads = 1; | |
7a51ed66 | 2300 | |
77ff1127 BP |
2301 | /* TODO: does creating more threads than cores help? */ |
2302 | if (!nr_threads) { | |
2303 | nr_threads = istate->cache_nr / THREAD_COST; | |
2304 | cpus = online_cpus(); | |
2305 | if (nr_threads > cpus) | |
2306 | nr_threads = cpus; | |
e83c5163 | 2307 | } |
abb4bb83 | 2308 | |
88168b9b NTND |
2309 | if (!HAVE_THREADS) |
2310 | nr_threads = 1; | |
2311 | ||
abb4bb83 BP |
2312 | if (nr_threads > 1) { |
2313 | extension_offset = read_eoie_extension(mmap, mmap_size); | |
2314 | if (extension_offset) { | |
2315 | int err; | |
2316 | ||
2317 | p.src_offset = extension_offset; | |
2318 | err = pthread_create(&p.pthread, NULL, load_index_extensions, &p); | |
2319 | if (err) | |
2320 | die(_("unable to create load_index_extensions thread: %s"), strerror(err)); | |
2321 | ||
2322 | nr_threads--; | |
2323 | } | |
2324 | } | |
abb4bb83 | 2325 | |
77ff1127 BP |
2326 | /* |
2327 | * Locate and read the index entry offset table so that we can use it | |
2328 | * to multi-thread the reading of the cache entries. | |
2329 | */ | |
2330 | if (extension_offset && nr_threads > 1) | |
2331 | ieot = read_ieot_extension(mmap, mmap_size, extension_offset); | |
2332 | ||
2333 | if (ieot) { | |
7bd9631b | 2334 | src_offset += load_cache_entries_threaded(istate, mmap, mmap_size, nr_threads, ieot); |
77ff1127 | 2335 | free(ieot); |
8e72d675 | 2336 | } else { |
77ff1127 | 2337 | src_offset += load_all_cache_entries(istate, mmap, mmap_size, src_offset); |
8e72d675 | 2338 | } |
6c9cd161 | 2339 | |
fba2f38a | 2340 | istate->timestamp.sec = st.st_mtime; |
c06ff490 | 2341 | istate->timestamp.nsec = ST_MTIME_NSEC(st); |
fba2f38a | 2342 | |
abb4bb83 | 2343 | /* if we created a thread, join it otherwise load the extensions on the primary thread */ |
abb4bb83 BP |
2344 | if (extension_offset) { |
2345 | int ret = pthread_join(p.pthread, NULL); | |
2346 | if (ret) | |
2347 | die(_("unable to join load_index_extensions thread: %s"), strerror(ret)); | |
88168b9b | 2348 | } else { |
abb4bb83 BP |
2349 | p.src_offset = src_offset; |
2350 | load_index_extensions(&p); | |
bad68ec9 | 2351 | } |
371ed0de | 2352 | munmap((void *)mmap, mmap_size); |
42fee7a3 JH |
2353 | |
2354 | /* | |
2355 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2356 | * that is associated with the given "istate". | |
2357 | */ | |
2358 | trace2_data_intmax("index", the_repository, "read/version", | |
2359 | istate->version); | |
2360 | trace2_data_intmax("index", the_repository, "read/cache_nr", | |
2361 | istate->cache_nr); | |
2362 | ||
7ca4fc88 VD |
2363 | /* |
2364 | * If the command explicitly requires a full index, force it | |
2365 | * to be full. Otherwise, correct the sparsity based on repository | |
2366 | * settings and other properties of the index (if necessary). | |
2367 | */ | |
4300f844 DS |
2368 | prepare_repo_settings(istate->repo); |
2369 | if (istate->repo->settings.command_requires_full_index) | |
2370 | ensure_full_index(istate); | |
7ca4fc88 VD |
2371 | else |
2372 | ensure_correct_sparsity(istate); | |
4300f844 | 2373 | |
4aab5b46 | 2374 | return istate->cache_nr; |
e83c5163 LT |
2375 | |
2376 | unmap: | |
371ed0de | 2377 | munmap((void *)mmap, mmap_size); |
9d0a9e90 | 2378 | die(_("index file corrupt")); |
e83c5163 LT |
2379 | } |
2380 | ||
0d59ffb4 CC |
2381 | /* |
2382 | * Signal that the shared index is used by updating its mtime. | |
2383 | * | |
2384 | * This way, shared index can be removed if they have not been used | |
2385 | * for some time. | |
2386 | */ | |
a125a223 | 2387 | static void freshen_shared_index(const char *shared_index, int warn) |
0d59ffb4 | 2388 | { |
0d59ffb4 | 2389 | if (!check_and_freshen_file(shared_index, 1) && warn) |
9d0a9e90 | 2390 | warning(_("could not freshen shared index '%s'"), shared_index); |
0d59ffb4 CC |
2391 | } |
2392 | ||
a125a223 TG |
2393 | int read_index_from(struct index_state *istate, const char *path, |
2394 | const char *gitdir) | |
5fc2fc8f NTND |
2395 | { |
2396 | struct split_index *split_index; | |
2397 | int ret; | |
2182abd9 | 2398 | char *base_oid_hex; |
a125a223 | 2399 | char *base_path; |
5fc2fc8f NTND |
2400 | |
2401 | /* istate->initialized covers both .git/index and .git/sharedindex.xxx */ | |
2402 | if (istate->initialized) | |
2403 | return istate->cache_nr; | |
2404 | ||
42fee7a3 JH |
2405 | /* |
2406 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2407 | * that is associated with the given "istate". | |
2408 | */ | |
2409 | trace2_region_enter_printf("index", "do_read_index", the_repository, | |
2410 | "%s", path); | |
c46c406a | 2411 | trace_performance_enter(); |
5fc2fc8f | 2412 | ret = do_read_index(istate, path, 0); |
c46c406a | 2413 | trace_performance_leave("read cache %s", path); |
42fee7a3 JH |
2414 | trace2_region_leave_printf("index", "do_read_index", the_repository, |
2415 | "%s", path); | |
435ec090 | 2416 | |
5fc2fc8f | 2417 | split_index = istate->split_index; |
2182abd9 | 2418 | if (!split_index || is_null_oid(&split_index->base_oid)) { |
435ec090 | 2419 | post_read_index_from(istate); |
5fc2fc8f | 2420 | return ret; |
03f15a79 | 2421 | } |
5fc2fc8f | 2422 | |
c46c406a | 2423 | trace_performance_enter(); |
5fc2fc8f | 2424 | if (split_index->base) |
2f6b1eb7 | 2425 | release_index(split_index->base); |
5fc2fc8f | 2426 | else |
2f6b1eb7 | 2427 | ALLOC_ARRAY(split_index->base, 1); |
6269f8ea | 2428 | index_state_init(split_index->base, istate->repo); |
de6ae5f9 | 2429 | |
2182abd9 | 2430 | base_oid_hex = oid_to_hex(&split_index->base_oid); |
2431 | base_path = xstrfmt("%s/sharedindex.%s", gitdir, base_oid_hex); | |
7667f4f0 JS |
2432 | if (file_exists(base_path)) { |
2433 | trace2_region_enter_printf("index", "shared/do_read_index", | |
2434 | the_repository, "%s", base_path); | |
2435 | ||
2436 | ret = do_read_index(split_index->base, base_path, 0); | |
2437 | trace2_region_leave_printf("index", "shared/do_read_index", | |
2438 | the_repository, "%s", base_path); | |
2439 | } else { | |
998330ac | 2440 | char *path_copy = xstrdup(path); |
652891de JS |
2441 | char *base_path2 = xstrfmt("%s/sharedindex.%s", |
2442 | dirname(path_copy), base_oid_hex); | |
998330ac SG |
2443 | free(path_copy); |
2444 | trace2_region_enter_printf("index", "shared/do_read_index", | |
2445 | the_repository, "%s", base_path2); | |
2446 | ret = do_read_index(split_index->base, base_path2, 1); | |
2447 | trace2_region_leave_printf("index", "shared/do_read_index", | |
2448 | the_repository, "%s", base_path2); | |
652891de | 2449 | free(base_path2); |
998330ac | 2450 | } |
9001dc2a | 2451 | if (!oideq(&split_index->base_oid, &split_index->base->oid)) |
9d0a9e90 | 2452 | die(_("broken index, expect %s in %s, got %s"), |
2182abd9 | 2453 | base_oid_hex, base_path, |
75691ea3 | 2454 | oid_to_hex(&split_index->base->oid)); |
de6ae5f9 | 2455 | |
a125a223 | 2456 | freshen_shared_index(base_path, 0); |
5fc2fc8f | 2457 | merge_base_index(istate); |
435ec090 | 2458 | post_read_index_from(istate); |
c46c406a | 2459 | trace_performance_leave("read cache %s", base_path); |
b42ad7d5 | 2460 | free(base_path); |
5fc2fc8f NTND |
2461 | return ret; |
2462 | } | |
2463 | ||
fa7b3c2f JH |
2464 | int is_index_unborn(struct index_state *istate) |
2465 | { | |
debed2a6 | 2466 | return (!istate->cache_nr && !istate->timestamp.sec); |
fa7b3c2f JH |
2467 | } |
2468 | ||
6269f8ea | 2469 | void index_state_init(struct index_state *istate, struct repository *r) |
2f6b1eb7 | 2470 | { |
6269f8ea | 2471 | struct index_state blank = INDEX_STATE_INIT(r); |
2f6b1eb7 ÆAB |
2472 | memcpy(istate, &blank, sizeof(*istate)); |
2473 | } | |
2474 | ||
2475 | void release_index(struct index_state *istate) | |
6d297f81 | 2476 | { |
8e72d675 JM |
2477 | /* |
2478 | * Cache entries in istate->cache[] should have been allocated | |
2479 | * from the memory pool associated with this index, or from an | |
2480 | * associated split_index. There is no need to free individual | |
8616a2d0 JM |
2481 | * cache entries. validate_cache_entries can detect when this |
2482 | * assertion does not hold. | |
8e72d675 | 2483 | */ |
8616a2d0 | 2484 | validate_cache_entries(istate); |
debed2a6 | 2485 | |
cfc5789a | 2486 | resolve_undo_clear_index(istate); |
2092678c | 2487 | free_name_hash(istate); |
4aab5b46 | 2488 | cache_tree_free(&(istate->cache_tree)); |
2f6b1eb7 ÆAB |
2489 | free(istate->fsmonitor_last_update); |
2490 | free(istate->cache); | |
5fc2fc8f | 2491 | discard_split_index(istate); |
f9e6c649 | 2492 | free_untracked_cache(istate->untracked); |
8e72d675 | 2493 | |
b5fcb1c0 ÆAB |
2494 | if (istate->sparse_checkout_patterns) { |
2495 | clear_pattern_list(istate->sparse_checkout_patterns); | |
2496 | FREE_AND_NULL(istate->sparse_checkout_patterns); | |
2497 | } | |
2498 | ||
8e72d675 | 2499 | if (istate->ce_mem_pool) { |
8616a2d0 | 2500 | mem_pool_discard(istate->ce_mem_pool, should_validate_cache_entries()); |
44c7e1a7 | 2501 | FREE_AND_NULL(istate->ce_mem_pool); |
8e72d675 | 2502 | } |
6d297f81 JS |
2503 | } |
2504 | ||
2f6b1eb7 ÆAB |
2505 | void discard_index(struct index_state *istate) |
2506 | { | |
2507 | release_index(istate); | |
6269f8ea | 2508 | index_state_init(istate, istate->repo); |
2f6b1eb7 ÆAB |
2509 | } |
2510 | ||
8616a2d0 JM |
2511 | /* |
2512 | * Validate the cache entries of this index. | |
2513 | * All cache entries associated with this index | |
2514 | * should have been allocated by the memory pool | |
2515 | * associated with this index, or by a referenced | |
2516 | * split index. | |
2517 | */ | |
2518 | void validate_cache_entries(const struct index_state *istate) | |
2519 | { | |
2520 | int i; | |
2521 | ||
2522 | if (!should_validate_cache_entries() ||!istate || !istate->initialized) | |
2523 | return; | |
2524 | ||
2525 | for (i = 0; i < istate->cache_nr; i++) { | |
2526 | if (!istate) { | |
391408e5 | 2527 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2528 | } else if (!istate->ce_mem_pool || |
2529 | !mem_pool_contains(istate->ce_mem_pool, istate->cache[i])) { | |
2530 | if (!istate->split_index || | |
2531 | !istate->split_index->base || | |
2532 | !istate->split_index->base->ce_mem_pool || | |
2533 | !mem_pool_contains(istate->split_index->base->ce_mem_pool, istate->cache[i])) { | |
391408e5 | 2534 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2535 | } |
2536 | } | |
2537 | } | |
2538 | ||
2539 | if (istate->split_index) | |
2540 | validate_cache_entries(istate->split_index->base); | |
2541 | } | |
2542 | ||
d1f128b0 | 2543 | int unmerged_index(const struct index_state *istate) |
94a5728c DB |
2544 | { |
2545 | int i; | |
2546 | for (i = 0; i < istate->cache_nr; i++) { | |
2547 | if (ce_stage(istate->cache[i])) | |
2548 | return 1; | |
2549 | } | |
2550 | return 0; | |
2551 | } | |
2552 | ||
150fe065 NTND |
2553 | int repo_index_has_changes(struct repository *repo, |
2554 | struct tree *tree, | |
2555 | struct strbuf *sb) | |
cffbfad5 | 2556 | { |
150fe065 | 2557 | struct index_state *istate = repo->index; |
e1f8694f | 2558 | struct object_id cmp; |
cffbfad5 EN |
2559 | int i; |
2560 | ||
e1f8694f EN |
2561 | if (tree) |
2562 | cmp = tree->object.oid; | |
4a93b899 | 2563 | if (tree || !repo_get_oid_tree(repo, "HEAD", &cmp)) { |
cffbfad5 EN |
2564 | struct diff_options opt; |
2565 | ||
150fe065 | 2566 | repo_diff_setup(repo, &opt); |
cffbfad5 EN |
2567 | opt.flags.exit_with_status = 1; |
2568 | if (!sb) | |
2569 | opt.flags.quick = 1; | |
8d833e93 | 2570 | diff_setup_done(&opt); |
e1f8694f | 2571 | do_diff_cache(&cmp, &opt); |
cffbfad5 EN |
2572 | diffcore_std(&opt); |
2573 | for (i = 0; sb && i < diff_queued_diff.nr; i++) { | |
2574 | if (i) | |
2575 | strbuf_addch(sb, ' '); | |
2576 | strbuf_addstr(sb, diff_queued_diff.queue[i]->two->path); | |
2577 | } | |
2578 | diff_flush(&opt); | |
2579 | return opt.flags.has_changes != 0; | |
2580 | } else { | |
0c18c059 DS |
2581 | /* TODO: audit for interaction with sparse-index. */ |
2582 | ensure_full_index(istate); | |
1b9fbefb | 2583 | for (i = 0; sb && i < istate->cache_nr; i++) { |
cffbfad5 EN |
2584 | if (i) |
2585 | strbuf_addch(sb, ' '); | |
1b9fbefb | 2586 | strbuf_addstr(sb, istate->cache[i]->name); |
cffbfad5 | 2587 | } |
1b9fbefb | 2588 | return !!istate->cache_nr; |
cffbfad5 EN |
2589 | } |
2590 | } | |
2591 | ||
410334ed DS |
2592 | static int write_index_ext_header(struct hashfile *f, |
2593 | git_hash_ctx *eoie_f, | |
2594 | unsigned int ext, | |
2595 | unsigned int sz) | |
bad68ec9 | 2596 | { |
410334ed DS |
2597 | hashwrite_be32(f, ext); |
2598 | hashwrite_be32(f, sz); | |
2599 | ||
2600 | if (eoie_f) { | |
2601 | ext = htonl(ext); | |
2602 | sz = htonl(sz); | |
2603 | the_hash_algo->update_fn(eoie_f, &ext, sizeof(ext)); | |
2604 | the_hash_algo->update_fn(eoie_f, &sz, sizeof(sz)); | |
3b1d9e04 | 2605 | } |
410334ed | 2606 | return 0; |
bad68ec9 JH |
2607 | } |
2608 | ||
58bf2a4c NTND |
2609 | static void ce_smudge_racily_clean_entry(struct index_state *istate, |
2610 | struct cache_entry *ce) | |
407c8eb0 JH |
2611 | { |
2612 | /* | |
2613 | * The only thing we care about in this function is to smudge the | |
2614 | * falsely clean entry due to touch-update-touch race, so we leave | |
2615 | * everything else as they are. We are called for entries whose | |
c21d39d7 | 2616 | * ce_stat_data.sd_mtime match the index file mtime. |
c70115b4 JH |
2617 | * |
2618 | * Note that this actually does not do much for gitlinks, for | |
2619 | * which ce_match_stat_basic() always goes to the actual | |
2620 | * contents. The caller checks with is_racy_timestamp() which | |
2621 | * always says "no" for gitlinks, so we are not called for them ;-) | |
407c8eb0 JH |
2622 | */ |
2623 | struct stat st; | |
2624 | ||
2625 | if (lstat(ce->name, &st) < 0) | |
2626 | return; | |
2627 | if (ce_match_stat_basic(ce, &st)) | |
2628 | return; | |
58bf2a4c | 2629 | if (ce_modified_check_fs(istate, ce, &st)) { |
4b3511b0 JH |
2630 | /* This is "racily clean"; smudge it. Note that this |
2631 | * is a tricky code. At first glance, it may appear | |
2632 | * that it can break with this sequence: | |
2633 | * | |
2634 | * $ echo xyzzy >frotz | |
2635 | * $ git-update-index --add frotz | |
2636 | * $ : >frotz | |
2637 | * $ sleep 3 | |
2638 | * $ echo filfre >nitfol | |
2639 | * $ git-update-index --add nitfol | |
2640 | * | |
b7e58b17 | 2641 | * but it does not. When the second update-index runs, |
4b3511b0 JH |
2642 | * it notices that the entry "frotz" has the same timestamp |
2643 | * as index, and if we were to smudge it by resetting its | |
2644 | * size to zero here, then the object name recorded | |
2645 | * in index is the 6-byte file but the cached stat information | |
2646 | * becomes zero --- which would then match what we would | |
a6080a0a | 2647 | * obtain from the filesystem next time we stat("frotz"). |
4b3511b0 JH |
2648 | * |
2649 | * However, the second update-index, before calling | |
2650 | * this function, notices that the cached size is 6 | |
2651 | * bytes and what is on the filesystem is an empty | |
2652 | * file, and never calls us, so the cached size information | |
2653 | * for "frotz" stays 6 which does not match the filesystem. | |
2654 | */ | |
c21d39d7 | 2655 | ce->ce_stat_data.sd_size = 0; |
407c8eb0 JH |
2656 | } |
2657 | } | |
2658 | ||
f136f7bf | 2659 | /* Copy miscellaneous fields but not the name */ |
ce012deb | 2660 | static void copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk, |
f136f7bf | 2661 | struct cache_entry *ce) |
7a51ed66 | 2662 | { |
b60e188c | 2663 | short flags; |
575fa8a3 | 2664 | const unsigned hashsz = the_hash_algo->rawsz; |
2665 | uint16_t *flagsp = (uint16_t *)(ondisk->data + hashsz); | |
b60e188c | 2666 | |
c21d39d7 MH |
2667 | ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec); |
2668 | ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec); | |
2669 | ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec); | |
2670 | ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec); | |
2671 | ondisk->dev = htonl(ce->ce_stat_data.sd_dev); | |
2672 | ondisk->ino = htonl(ce->ce_stat_data.sd_ino); | |
7a51ed66 | 2673 | ondisk->mode = htonl(ce->ce_mode); |
c21d39d7 MH |
2674 | ondisk->uid = htonl(ce->ce_stat_data.sd_uid); |
2675 | ondisk->gid = htonl(ce->ce_stat_data.sd_gid); | |
2676 | ondisk->size = htonl(ce->ce_stat_data.sd_size); | |
575fa8a3 | 2677 | hashcpy(ondisk->data, ce->oid.hash); |
b60e188c | 2678 | |
ce51bf09 | 2679 | flags = ce->ce_flags & ~CE_NAMEMASK; |
b60e188c | 2680 | flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce)); |
575fa8a3 | 2681 | flagsp[0] = htons(flags); |
06aaaa0b | 2682 | if (ce->ce_flags & CE_EXTENDED) { |
575fa8a3 | 2683 | flagsp[1] = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16); |
f136f7bf JH |
2684 | } |
2685 | } | |
2686 | ||
410334ed | 2687 | static int ce_write_entry(struct hashfile *f, struct cache_entry *ce, |
ce012deb | 2688 | struct strbuf *previous_name, struct ondisk_cache_entry *ondisk) |
f136f7bf | 2689 | { |
9d227781 | 2690 | int size; |
00a4b035 RJ |
2691 | unsigned int saved_namelen; |
2692 | int stripped_name = 0; | |
ce012deb | 2693 | static unsigned char padding[8] = { 0x00 }; |
f136f7bf | 2694 | |
b3c96fb1 NTND |
2695 | if (ce->ce_flags & CE_STRIP_NAME) { |
2696 | saved_namelen = ce_namelen(ce); | |
2697 | ce->ce_namelen = 0; | |
00a4b035 | 2698 | stripped_name = 1; |
b3c96fb1 NTND |
2699 | } |
2700 | ||
575fa8a3 | 2701 | size = offsetof(struct ondisk_cache_entry,data) + ondisk_data_size(ce->ce_flags, 0); |
ce012deb | 2702 | |
9d227781 | 2703 | if (!previous_name) { |
ce012deb KW |
2704 | int len = ce_namelen(ce); |
2705 | copy_cache_entry_to_ondisk(ondisk, ce); | |
410334ed DS |
2706 | hashwrite(f, ondisk, size); |
2707 | hashwrite(f, ce->name, len); | |
2708 | hashwrite(f, padding, align_padding_size(size, len)); | |
9d227781 JH |
2709 | } else { |
2710 | int common, to_remove, prefix_size; | |
2711 | unsigned char to_remove_vi[16]; | |
2712 | for (common = 0; | |
2713 | (ce->name[common] && | |
2714 | common < previous_name->len && | |
2715 | ce->name[common] == previous_name->buf[common]); | |
2716 | common++) | |
2717 | ; /* still matching */ | |
2718 | to_remove = previous_name->len - common; | |
2719 | prefix_size = encode_varint(to_remove, to_remove_vi); | |
2720 | ||
ce012deb | 2721 | copy_cache_entry_to_ondisk(ondisk, ce); |
410334ed DS |
2722 | hashwrite(f, ondisk, size); |
2723 | hashwrite(f, to_remove_vi, prefix_size); | |
2724 | hashwrite(f, ce->name + common, ce_namelen(ce) - common); | |
2725 | hashwrite(f, padding, 1); | |
9d227781 JH |
2726 | |
2727 | strbuf_splice(previous_name, common, to_remove, | |
2728 | ce->name + common, ce_namelen(ce) - common); | |
06aaaa0b | 2729 | } |
00a4b035 | 2730 | if (stripped_name) { |
b3c96fb1 NTND |
2731 | ce->ce_namelen = saved_namelen; |
2732 | ce->ce_flags &= ~CE_STRIP_NAME; | |
2733 | } | |
7a51ed66 | 2734 | |
410334ed | 2735 | return 0; |
7a51ed66 LT |
2736 | } |
2737 | ||
426ddeea YM |
2738 | /* |
2739 | * This function verifies if index_state has the correct sha1 of the | |
2740 | * index file. Don't die if we have any other failure, just return 0. | |
2741 | */ | |
2742 | static int verify_index_from(const struct index_state *istate, const char *path) | |
2743 | { | |
2744 | int fd; | |
2745 | ssize_t n; | |
2746 | struct stat st; | |
aab61359 | 2747 | unsigned char hash[GIT_MAX_RAWSZ]; |
426ddeea YM |
2748 | |
2749 | if (!istate->initialized) | |
2750 | return 0; | |
2751 | ||
2752 | fd = open(path, O_RDONLY); | |
2753 | if (fd < 0) | |
2754 | return 0; | |
2755 | ||
2756 | if (fstat(fd, &st)) | |
2757 | goto out; | |
2758 | ||
aab61359 | 2759 | if (st.st_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
426ddeea YM |
2760 | goto out; |
2761 | ||
aab61359 | 2762 | n = pread_in_full(fd, hash, the_hash_algo->rawsz, st.st_size - the_hash_algo->rawsz); |
2763 | if (n != the_hash_algo->rawsz) | |
426ddeea YM |
2764 | goto out; |
2765 | ||
67947c34 | 2766 | if (!hasheq(istate->oid.hash, hash)) |
426ddeea YM |
2767 | goto out; |
2768 | ||
2769 | close(fd); | |
2770 | return 1; | |
2771 | ||
2772 | out: | |
2773 | close(fd); | |
2774 | return 0; | |
2775 | } | |
2776 | ||
1b0d968b | 2777 | static int repo_verify_index(struct repository *repo) |
426ddeea | 2778 | { |
1b0d968b | 2779 | return verify_index_from(repo->index, repo->index_file); |
426ddeea YM |
2780 | } |
2781 | ||
2ede073f | 2782 | int has_racy_timestamp(struct index_state *istate) |
483fbe2b JH |
2783 | { |
2784 | int entries = istate->cache_nr; | |
2785 | int i; | |
2786 | ||
2787 | for (i = 0; i < entries; i++) { | |
2788 | struct cache_entry *ce = istate->cache[i]; | |
2789 | if (is_racy_timestamp(istate, ce)) | |
2790 | return 1; | |
2791 | } | |
2792 | return 0; | |
2793 | } | |
2794 | ||
1b0d968b NTND |
2795 | void repo_update_index_if_able(struct repository *repo, |
2796 | struct lock_file *lockfile) | |
ccdc4ec3 | 2797 | { |
1b0d968b NTND |
2798 | if ((repo->index->cache_changed || |
2799 | has_racy_timestamp(repo->index)) && | |
2800 | repo_verify_index(repo)) | |
2801 | write_locked_index(repo->index, lockfile, COMMIT_LOCK); | |
b74c90fb MÅ |
2802 | else |
2803 | rollback_lock_file(lockfile); | |
ccdc4ec3 JH |
2804 | } |
2805 | ||
d8465500 JN |
2806 | static int record_eoie(void) |
2807 | { | |
2808 | int val; | |
2809 | ||
2810 | if (!git_config_get_bool("index.recordendofindexentries", &val)) | |
2811 | return val; | |
2a9dedef JN |
2812 | |
2813 | /* | |
2814 | * As a convenience, the end of index entries extension | |
2815 | * used for threading is written by default if the user | |
2816 | * explicitly requested threaded index reads. | |
2817 | */ | |
2818 | return !git_config_get_index_threads(&val) && val != 1; | |
d8465500 JN |
2819 | } |
2820 | ||
42916054 JN |
2821 | static int record_ieot(void) |
2822 | { | |
2823 | int val; | |
2824 | ||
2825 | if (!git_config_get_bool("index.recordoffsettable", &val)) | |
2826 | return val; | |
2a9dedef JN |
2827 | |
2828 | /* | |
2829 | * As a convenience, the offset table used for threading is | |
2830 | * written by default if the user explicitly requested | |
2831 | * threaded index reads. | |
2832 | */ | |
2833 | return !git_config_get_index_threads(&val) && val != 1; | |
42916054 JN |
2834 | } |
2835 | ||
3b7a4475 JS |
2836 | enum write_extensions { |
2837 | WRITE_NO_EXTENSION = 0, | |
2838 | WRITE_SPLIT_INDEX_EXTENSION = 1<<0, | |
2839 | WRITE_CACHE_TREE_EXTENSION = 1<<1, | |
2840 | WRITE_RESOLVE_UNDO_EXTENSION = 1<<2, | |
2841 | WRITE_UNTRACKED_CACHE_EXTENSION = 1<<3, | |
2842 | WRITE_FSMONITOR_EXTENSION = 1<<4, | |
2843 | }; | |
2844 | #define WRITE_ALL_EXTENSIONS ((enum write_extensions)-1) | |
2845 | ||
812d6b00 MÅ |
2846 | /* |
2847 | * On success, `tempfile` is closed. If it is the temporary file | |
2848 | * of a `struct lock_file`, we will therefore effectively perform | |
2849 | * a 'close_lock_file_gently()`. Since that is an implementation | |
2850 | * detail of lockfiles, callers of `do_write_index()` should not | |
2851 | * rely on it. | |
2852 | */ | |
9f41c7a6 | 2853 | static int do_write_index(struct index_state *istate, struct tempfile *tempfile, |
3b7a4475 | 2854 | enum write_extensions write_extensions, unsigned flags) |
197ee8c9 | 2855 | { |
ca54d9ba | 2856 | uint64_t start = getnanotime(); |
410334ed DS |
2857 | struct hashfile *f; |
2858 | git_hash_ctx *eoie_c = NULL; | |
197ee8c9 | 2859 | struct cache_header hdr; |
b50386c7 | 2860 | int i, err = 0, removed, extended, hdr_version; |
4aab5b46 JH |
2861 | struct cache_entry **cache = istate->cache; |
2862 | int entries = istate->cache_nr; | |
e1afca4f | 2863 | struct stat st; |
575fa8a3 | 2864 | struct ondisk_cache_entry ondisk; |
9d227781 | 2865 | struct strbuf previous_name_buf = STRBUF_INIT, *previous_name; |
4bddd983 | 2866 | int drop_cache_tree = istate->drop_cache_tree; |
3b1d9e04 | 2867 | off_t offset; |
ba95e96d | 2868 | int csum_fsync_flag; |
77ff1127 | 2869 | int ieot_entries = 1; |
3255089a BP |
2870 | struct index_entry_offset_table *ieot = NULL; |
2871 | int nr, nr_threads; | |
6269f8ea | 2872 | struct repository *r = istate->repo; |
025a0709 | 2873 | |
410334ed DS |
2874 | f = hashfd(tempfile->fd, tempfile->filename.buf); |
2875 | ||
17194b19 DS |
2876 | prepare_repo_settings(r); |
2877 | f->skip_hash = r->settings.index_skip_hash; | |
ee1f0c24 | 2878 | |
06aaaa0b | 2879 | for (i = removed = extended = 0; i < entries; i++) { |
7a51ed66 | 2880 | if (cache[i]->ce_flags & CE_REMOVE) |
025a0709 | 2881 | removed++; |
197ee8c9 | 2882 | |
06aaaa0b NTND |
2883 | /* reduce extended entries if possible */ |
2884 | cache[i]->ce_flags &= ~CE_EXTENDED; | |
2885 | if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) { | |
2886 | extended++; | |
2887 | cache[i]->ce_flags |= CE_EXTENDED; | |
2888 | } | |
2889 | } | |
2890 | ||
e8ffd034 | 2891 | if (!istate->version) |
4a93b899 | 2892 | istate->version = get_index_format_default(r); |
9d227781 JH |
2893 | |
2894 | /* demote version 3 to version 2 when the latter suffices */ | |
2895 | if (istate->version == 3 || istate->version == 2) | |
2896 | istate->version = extended ? 3 : 2; | |
2897 | ||
2898 | hdr_version = istate->version; | |
2899 | ||
ccc4feb5 | 2900 | hdr.hdr_signature = htonl(CACHE_SIGNATURE); |
9d227781 | 2901 | hdr.hdr_version = htonl(hdr_version); |
025a0709 | 2902 | hdr.hdr_entries = htonl(entries - removed); |
197ee8c9 | 2903 | |
410334ed | 2904 | hashwrite(f, &hdr, sizeof(hdr)); |
197ee8c9 | 2905 | |
2a9dedef | 2906 | if (!HAVE_THREADS || git_config_get_index_threads(&nr_threads)) |
88168b9b | 2907 | nr_threads = 1; |
62e5ee81 | 2908 | |
42916054 | 2909 | if (nr_threads != 1 && record_ieot()) { |
3255089a BP |
2910 | int ieot_blocks, cpus; |
2911 | ||
2912 | /* | |
2913 | * ensure default number of ieot blocks maps evenly to the | |
2914 | * default number of threads that will process them leaving | |
2915 | * room for the thread to load the index extensions. | |
2916 | */ | |
2917 | if (!nr_threads) { | |
2918 | ieot_blocks = istate->cache_nr / THREAD_COST; | |
2919 | cpus = online_cpus(); | |
2920 | if (ieot_blocks > cpus - 1) | |
2921 | ieot_blocks = cpus - 1; | |
2922 | } else { | |
2923 | ieot_blocks = nr_threads; | |
77ff1127 BP |
2924 | if (ieot_blocks > istate->cache_nr) |
2925 | ieot_blocks = istate->cache_nr; | |
3255089a BP |
2926 | } |
2927 | ||
2928 | /* | |
2929 | * no reason to write out the IEOT extension if we don't | |
2930 | * have enough blocks to utilize multi-threading | |
2931 | */ | |
2932 | if (ieot_blocks > 1) { | |
2933 | ieot = xcalloc(1, sizeof(struct index_entry_offset_table) | |
2934 | + (ieot_blocks * sizeof(struct index_entry_offset))); | |
77ff1127 | 2935 | ieot_entries = DIV_ROUND_UP(entries, ieot_blocks); |
3255089a BP |
2936 | } |
2937 | } | |
3255089a | 2938 | |
410334ed DS |
2939 | offset = hashfile_total(f); |
2940 | ||
3255089a | 2941 | nr = 0; |
9d227781 | 2942 | previous_name = (hdr_version == 4) ? &previous_name_buf : NULL; |
ce012deb | 2943 | |
197ee8c9 LT |
2944 | for (i = 0; i < entries; i++) { |
2945 | struct cache_entry *ce = cache[i]; | |
7a51ed66 | 2946 | if (ce->ce_flags & CE_REMOVE) |
aa16021e | 2947 | continue; |
e06c43c7 | 2948 | if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce)) |
58bf2a4c | 2949 | ce_smudge_racily_clean_entry(istate, ce); |
99d1a986 | 2950 | if (is_null_oid(&ce->oid)) { |
83bd7437 JK |
2951 | static const char msg[] = "cache entry has null sha1: %s"; |
2952 | static int allow = -1; | |
2953 | ||
2954 | if (allow < 0) | |
2955 | allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0); | |
2956 | if (allow) | |
2957 | warning(msg, ce->name); | |
2958 | else | |
b50386c7 | 2959 | err = error(msg, ce->name); |
a96d3cc3 JK |
2960 | |
2961 | drop_cache_tree = 1; | |
83bd7437 | 2962 | } |
77ff1127 | 2963 | if (ieot && i && (i % ieot_entries == 0)) { |
3255089a BP |
2964 | ieot->entries[ieot->nr].nr = nr; |
2965 | ieot->entries[ieot->nr].offset = offset; | |
2966 | ieot->nr++; | |
2967 | /* | |
2968 | * If we have a V4 index, set the first byte to an invalid | |
2969 | * character to ensure there is nothing common with the previous | |
2970 | * entry | |
2971 | */ | |
2972 | if (previous_name) | |
2973 | previous_name->buf[0] = 0; | |
2974 | nr = 0; | |
410334ed DS |
2975 | |
2976 | offset = hashfile_total(f); | |
3255089a | 2977 | } |
410334ed | 2978 | if (ce_write_entry(f, ce, previous_name, (struct ondisk_cache_entry *)&ondisk) < 0) |
b50386c7 KW |
2979 | err = -1; |
2980 | ||
2981 | if (err) | |
2982 | break; | |
3255089a BP |
2983 | nr++; |
2984 | } | |
2985 | if (ieot && nr) { | |
2986 | ieot->entries[ieot->nr].nr = nr; | |
2987 | ieot->entries[ieot->nr].offset = offset; | |
2988 | ieot->nr++; | |
197ee8c9 | 2989 | } |
9d227781 | 2990 | strbuf_release(&previous_name_buf); |
1af1c2b6 | 2991 | |
3255089a BP |
2992 | if (err) { |
2993 | free(ieot); | |
b50386c7 | 2994 | return err; |
3255089a | 2995 | } |
b50386c7 | 2996 | |
410334ed DS |
2997 | offset = hashfile_total(f); |
2998 | ||
2999 | /* | |
3000 | * The extension headers must be hashed on their own for the | |
3001 | * EOIE extension. Create a hashfile here to compute that hash. | |
3002 | */ | |
3003 | if (offset && record_eoie()) { | |
3004 | CALLOC_ARRAY(eoie_c, 1); | |
3005 | the_hash_algo->init_fn(eoie_c); | |
3255089a | 3006 | } |
3b1d9e04 | 3007 | |
3255089a BP |
3008 | /* |
3009 | * Lets write out CACHE_EXT_INDEXENTRYOFFSETTABLE first so that we | |
3010 | * can minimize the number of extensions we have to scan through to | |
3011 | * find it during load. Write it out regardless of the | |
3012 | * strip_extensions parameter as we need it when loading the shared | |
3013 | * index. | |
3014 | */ | |
3255089a BP |
3015 | if (ieot) { |
3016 | struct strbuf sb = STRBUF_INIT; | |
3017 | ||
3018 | write_ieot_extension(&sb, ieot); | |
410334ed DS |
3019 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_INDEXENTRYOFFSETTABLE, sb.len) < 0; |
3020 | hashwrite(f, sb.buf, sb.len); | |
3255089a BP |
3021 | strbuf_release(&sb); |
3022 | free(ieot); | |
3023 | if (err) | |
3024 | return -1; | |
3025 | } | |
3255089a | 3026 | |
3b7a4475 JS |
3027 | if (write_extensions & WRITE_SPLIT_INDEX_EXTENSION && |
3028 | istate->split_index) { | |
5fc2fc8f NTND |
3029 | struct strbuf sb = STRBUF_INIT; |
3030 | ||
451b66c5 JS |
3031 | if (istate->sparse_index) |
3032 | die(_("cannot write split index for a sparse index")); | |
3033 | ||
5fc2fc8f | 3034 | err = write_link_extension(&sb, istate) < 0 || |
410334ed DS |
3035 | write_index_ext_header(f, eoie_c, CACHE_EXT_LINK, |
3036 | sb.len) < 0; | |
3037 | hashwrite(f, sb.buf, sb.len); | |
5fc2fc8f NTND |
3038 | strbuf_release(&sb); |
3039 | if (err) | |
3040 | return -1; | |
3041 | } | |
3b7a4475 JS |
3042 | if (write_extensions & WRITE_CACHE_TREE_EXTENSION && |
3043 | !drop_cache_tree && istate->cache_tree) { | |
f285a2d7 | 3044 | struct strbuf sb = STRBUF_INIT; |
1dffb8fa | 3045 | |
1dffb8fa | 3046 | cache_tree_write(&sb, istate->cache_tree); |
410334ed DS |
3047 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_TREE, sb.len) < 0; |
3048 | hashwrite(f, sb.buf, sb.len); | |
1dffb8fa PH |
3049 | strbuf_release(&sb); |
3050 | if (err) | |
bad68ec9 | 3051 | return -1; |
bad68ec9 | 3052 | } |
3b7a4475 JS |
3053 | if (write_extensions & WRITE_RESOLVE_UNDO_EXTENSION && |
3054 | istate->resolve_undo) { | |
cfc5789a JH |
3055 | struct strbuf sb = STRBUF_INIT; |
3056 | ||
3057 | resolve_undo_write(&sb, istate->resolve_undo); | |
410334ed DS |
3058 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_RESOLVE_UNDO, |
3059 | sb.len) < 0; | |
3060 | hashwrite(f, sb.buf, sb.len); | |
cfc5789a JH |
3061 | strbuf_release(&sb); |
3062 | if (err) | |
3063 | return -1; | |
3064 | } | |
3b7a4475 JS |
3065 | if (write_extensions & WRITE_UNTRACKED_CACHE_EXTENSION && |
3066 | istate->untracked) { | |
83c094ad NTND |
3067 | struct strbuf sb = STRBUF_INIT; |
3068 | ||
3069 | write_untracked_extension(&sb, istate->untracked); | |
410334ed DS |
3070 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_UNTRACKED, |
3071 | sb.len) < 0; | |
3072 | hashwrite(f, sb.buf, sb.len); | |
83c094ad NTND |
3073 | strbuf_release(&sb); |
3074 | if (err) | |
3075 | return -1; | |
3076 | } | |
3b7a4475 JS |
3077 | if (write_extensions & WRITE_FSMONITOR_EXTENSION && |
3078 | istate->fsmonitor_last_update) { | |
883e248b BP |
3079 | struct strbuf sb = STRBUF_INIT; |
3080 | ||
3081 | write_fsmonitor_extension(&sb, istate); | |
410334ed DS |
3082 | err = write_index_ext_header(f, eoie_c, CACHE_EXT_FSMONITOR, sb.len) < 0; |
3083 | hashwrite(f, sb.buf, sb.len); | |
3b1d9e04 BP |
3084 | strbuf_release(&sb); |
3085 | if (err) | |
3086 | return -1; | |
3087 | } | |
cd42415f | 3088 | if (istate->sparse_index) { |
410334ed | 3089 | if (write_index_ext_header(f, eoie_c, CACHE_EXT_SPARSE_DIRECTORIES, 0) < 0) |
cd42415f DS |
3090 | return -1; |
3091 | } | |
3b1d9e04 BP |
3092 | |
3093 | /* | |
3094 | * CACHE_EXT_ENDOFINDEXENTRIES must be written as the last entry before the SHA1 | |
3095 | * so that it can be found and processed before all the index entries are | |
3096 | * read. Write it out regardless of the strip_extensions parameter as we need it | |
3097 | * when loading the shared index. | |
3098 | */ | |
410334ed | 3099 | if (eoie_c) { |
3b1d9e04 BP |
3100 | struct strbuf sb = STRBUF_INIT; |
3101 | ||
410334ed DS |
3102 | write_eoie_extension(&sb, eoie_c, offset); |
3103 | err = write_index_ext_header(f, NULL, CACHE_EXT_ENDOFINDEXENTRIES, sb.len) < 0; | |
3104 | hashwrite(f, sb.buf, sb.len); | |
883e248b BP |
3105 | strbuf_release(&sb); |
3106 | if (err) | |
3107 | return -1; | |
3108 | } | |
e1afca4f | 3109 | |
ba95e96d NS |
3110 | csum_fsync_flag = 0; |
3111 | if (!alternate_index_output && (flags & COMMIT_LOCK)) | |
3112 | csum_fsync_flag = CSUM_FSYNC; | |
3113 | ||
3114 | finalize_hashfile(f, istate->oid.hash, FSYNC_COMPONENT_INDEX, | |
3115 | CSUM_HASH_IN_STREAM | csum_fsync_flag); | |
3116 | ||
49bd0fc2 | 3117 | if (close_tempfile_gently(tempfile)) { |
6a8c89d0 | 3118 | error(_("could not close '%s'"), get_tempfile_path(tempfile)); |
49bd0fc2 JK |
3119 | return -1; |
3120 | } | |
6a8c89d0 | 3121 | if (stat(get_tempfile_path(tempfile), &st)) |
e1afca4f | 3122 | return -1; |
5bcf109c KB |
3123 | istate->timestamp.sec = (unsigned int)st.st_mtime; |
3124 | istate->timestamp.nsec = ST_MTIME_NSEC(st); | |
ca54d9ba | 3125 | trace_performance_since(start, "write index, changed mask = %x", istate->cache_changed); |
42fee7a3 JH |
3126 | |
3127 | /* | |
3128 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3129 | * that is associated with the given "istate". | |
3130 | */ | |
3131 | trace2_data_intmax("index", the_repository, "write/version", | |
3132 | istate->version); | |
3133 | trace2_data_intmax("index", the_repository, "write/cache_nr", | |
3134 | istate->cache_nr); | |
3135 | ||
e1afca4f | 3136 | return 0; |
197ee8c9 | 3137 | } |
e46bbcf6 | 3138 | |
626f35c8 NTND |
3139 | void set_alternate_index_output(const char *name) |
3140 | { | |
3141 | alternate_index_output = name; | |
3142 | } | |
3143 | ||
3144 | static int commit_locked_index(struct lock_file *lk) | |
3145 | { | |
751baced MH |
3146 | if (alternate_index_output) |
3147 | return commit_lock_file_to(lk, alternate_index_output); | |
3148 | else | |
626f35c8 | 3149 | return commit_lock_file(lk); |
626f35c8 NTND |
3150 | } |
3151 | ||
3b7a4475 JS |
3152 | static int do_write_locked_index(struct index_state *istate, |
3153 | struct lock_file *lock, | |
3154 | unsigned flags, | |
3155 | enum write_extensions write_extensions) | |
03b86647 | 3156 | { |
42fee7a3 | 3157 | int ret; |
9fadb373 | 3158 | int was_full = istate->sparse_index == INDEX_EXPANDED; |
6e773527 | 3159 | |
ce7a9f01 | 3160 | ret = convert_to_sparse(istate, 0); |
6e773527 DS |
3161 | |
3162 | if (ret) { | |
3163 | warning(_("failed to convert to a sparse-index")); | |
3164 | return ret; | |
3165 | } | |
42fee7a3 JH |
3166 | |
3167 | /* | |
3168 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3169 | * that is associated with the given "istate". | |
3170 | */ | |
3171 | trace2_region_enter_printf("index", "do_write_index", the_repository, | |
6a8c89d0 | 3172 | "%s", get_lock_file_path(lock)); |
3b7a4475 | 3173 | ret = do_write_index(istate, lock->tempfile, write_extensions, flags); |
42fee7a3 | 3174 | trace2_region_leave_printf("index", "do_write_index", the_repository, |
6a8c89d0 | 3175 | "%s", get_lock_file_path(lock)); |
42fee7a3 | 3176 | |
6e773527 DS |
3177 | if (was_full) |
3178 | ensure_full_index(istate); | |
3179 | ||
03b86647 NTND |
3180 | if (ret) |
3181 | return ret; | |
03b86647 | 3182 | if (flags & COMMIT_LOCK) |
1956ecd0 BP |
3183 | ret = commit_locked_index(lock); |
3184 | else | |
3185 | ret = close_lock_file_gently(lock); | |
3186 | ||
dbb1c613 | 3187 | run_hooks_l("post-index-change", |
1956ecd0 BP |
3188 | istate->updated_workdir ? "1" : "0", |
3189 | istate->updated_skipworktree ? "1" : "0", NULL); | |
3190 | istate->updated_workdir = 0; | |
3191 | istate->updated_skipworktree = 0; | |
3192 | ||
3193 | return ret; | |
03b86647 NTND |
3194 | } |
3195 | ||
5fc2fc8f NTND |
3196 | static int write_split_index(struct index_state *istate, |
3197 | struct lock_file *lock, | |
3198 | unsigned flags) | |
3199 | { | |
3200 | int ret; | |
3201 | prepare_to_write_split_index(istate); | |
3b7a4475 | 3202 | ret = do_write_locked_index(istate, lock, flags, WRITE_ALL_EXTENSIONS); |
5fc2fc8f NTND |
3203 | finish_writing_split_index(istate); |
3204 | return ret; | |
3205 | } | |
3206 | ||
b9683722 CC |
3207 | static const char *shared_index_expire = "2.weeks.ago"; |
3208 | ||
3209 | static unsigned long get_shared_index_expire_date(void) | |
3210 | { | |
3211 | static unsigned long shared_index_expire_date; | |
3212 | static int shared_index_expire_date_prepared; | |
3213 | ||
3214 | if (!shared_index_expire_date_prepared) { | |
3215 | git_config_get_expiry("splitindex.sharedindexexpire", | |
3216 | &shared_index_expire); | |
3217 | shared_index_expire_date = approxidate(shared_index_expire); | |
3218 | shared_index_expire_date_prepared = 1; | |
3219 | } | |
3220 | ||
3221 | return shared_index_expire_date; | |
3222 | } | |
3223 | ||
3224 | static int should_delete_shared_index(const char *shared_index_path) | |
3225 | { | |
3226 | struct stat st; | |
3227 | unsigned long expiration; | |
3228 | ||
3229 | /* Check timestamp */ | |
3230 | expiration = get_shared_index_expire_date(); | |
3231 | if (!expiration) | |
3232 | return 0; | |
3233 | if (stat(shared_index_path, &st)) | |
78bde923 | 3234 | return error_errno(_("could not stat '%s'"), shared_index_path); |
b9683722 CC |
3235 | if (st.st_mtime > expiration) |
3236 | return 0; | |
3237 | ||
3238 | return 1; | |
3239 | } | |
3240 | ||
3241 | static int clean_shared_index_files(const char *current_hex) | |
3242 | { | |
3243 | struct dirent *de; | |
3244 | DIR *dir = opendir(get_git_dir()); | |
3245 | ||
3246 | if (!dir) | |
3247 | return error_errno(_("unable to open git dir: %s"), get_git_dir()); | |
3248 | ||
3249 | while ((de = readdir(dir)) != NULL) { | |
3250 | const char *sha1_hex; | |
3251 | const char *shared_index_path; | |
3252 | if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex)) | |
3253 | continue; | |
3254 | if (!strcmp(sha1_hex, current_hex)) | |
3255 | continue; | |
3256 | shared_index_path = git_path("%s", de->d_name); | |
3257 | if (should_delete_shared_index(shared_index_path) > 0 && | |
3258 | unlink(shared_index_path)) | |
3259 | warning_errno(_("unable to unlink: %s"), shared_index_path); | |
3260 | } | |
3261 | closedir(dir); | |
3262 | ||
3263 | return 0; | |
3264 | } | |
3265 | ||
a0a96756 | 3266 | static int write_shared_index(struct index_state *istate, |
ba95e96d | 3267 | struct tempfile **temp, unsigned flags) |
c18b80a0 NTND |
3268 | { |
3269 | struct split_index *si = istate->split_index; | |
6e773527 | 3270 | int ret, was_full = !istate->sparse_index; |
c18b80a0 | 3271 | |
c18b80a0 | 3272 | move_cache_to_base_index(istate); |
ce7a9f01 | 3273 | convert_to_sparse(istate, 0); |
42fee7a3 JH |
3274 | |
3275 | trace2_region_enter_printf("index", "shared/do_write_index", | |
6a8c89d0 | 3276 | the_repository, "%s", get_tempfile_path(*temp)); |
3b7a4475 | 3277 | ret = do_write_index(si->base, *temp, WRITE_NO_EXTENSION, flags); |
c173542c | 3278 | trace2_region_leave_printf("index", "shared/do_write_index", |
6a8c89d0 | 3279 | the_repository, "%s", get_tempfile_path(*temp)); |
42fee7a3 | 3280 | |
6e773527 DS |
3281 | if (was_full) |
3282 | ensure_full_index(istate); | |
3283 | ||
59f9d2dd | 3284 | if (ret) |
c18b80a0 | 3285 | return ret; |
7db2d08c | 3286 | ret = adjust_shared_perm(get_tempfile_path(*temp)); |
df801f3f | 3287 | if (ret) { |
9d0a9e90 | 3288 | error(_("cannot fix permission bits on '%s'"), get_tempfile_path(*temp)); |
df801f3f CC |
3289 | return ret; |
3290 | } | |
7db2d08c | 3291 | ret = rename_tempfile(temp, |
75691ea3 | 3292 | git_path("sharedindex.%s", oid_to_hex(&si->base->oid))); |
b9683722 | 3293 | if (!ret) { |
75691ea3 | 3294 | oidcpy(&si->base_oid, &si->base->oid); |
3295 | clean_shared_index_files(oid_to_hex(&si->base->oid)); | |
b9683722 CC |
3296 | } |
3297 | ||
c18b80a0 NTND |
3298 | return ret; |
3299 | } | |
3300 | ||
e6a1dd77 CC |
3301 | static const int default_max_percent_split_change = 20; |
3302 | ||
3303 | static int too_many_not_shared_entries(struct index_state *istate) | |
3304 | { | |
3305 | int i, not_shared = 0; | |
3306 | int max_split = git_config_get_max_percent_split_change(); | |
3307 | ||
3308 | switch (max_split) { | |
3309 | case -1: | |
3310 | /* not or badly configured: use the default value */ | |
3311 | max_split = default_max_percent_split_change; | |
3312 | break; | |
3313 | case 0: | |
3314 | return 1; /* 0% means always write a new shared index */ | |
3315 | case 100: | |
3316 | return 0; /* 100% means never write a new shared index */ | |
3317 | default: | |
3318 | break; /* just use the configured value */ | |
3319 | } | |
3320 | ||
3321 | /* Count not shared entries */ | |
3322 | for (i = 0; i < istate->cache_nr; i++) { | |
3323 | struct cache_entry *ce = istate->cache[i]; | |
3324 | if (!ce->index) | |
3325 | not_shared++; | |
3326 | } | |
3327 | ||
3328 | return (int64_t)istate->cache_nr * max_split < (int64_t)not_shared * 100; | |
3329 | } | |
3330 | ||
03b86647 NTND |
3331 | int write_locked_index(struct index_state *istate, struct lock_file *lock, |
3332 | unsigned flags) | |
3333 | { | |
e8ffd034 | 3334 | int new_shared_index, ret, test_split_index_env; |
5fc2fc8f NTND |
3335 | struct split_index *si = istate->split_index; |
3336 | ||
4592e608 | 3337 | if (git_env_bool("GIT_TEST_CHECK_CACHE_TREE", 0)) |
c207e9e1 | 3338 | cache_tree_verify(the_repository, istate); |
4592e608 | 3339 | |
61000814 MÅ |
3340 | if ((flags & SKIP_IF_UNCHANGED) && !istate->cache_changed) { |
3341 | if (flags & COMMIT_LOCK) | |
3342 | rollback_lock_file(lock); | |
3343 | return 0; | |
3344 | } | |
3345 | ||
3bd28eb2 AV |
3346 | if (istate->fsmonitor_last_update) |
3347 | fill_fsmonitor_bitmap(istate); | |
3348 | ||
e8ffd034 SG |
3349 | test_split_index_env = git_env_bool("GIT_TEST_SPLIT_INDEX", 0); |
3350 | ||
3351 | if ((!si && !test_split_index_env) || | |
3352 | alternate_index_output || | |
5165dd59 | 3353 | (istate->cache_changed & ~EXTMASK)) { |
3b7a4475 JS |
3354 | ret = do_write_locked_index(istate, lock, flags, |
3355 | ~WRITE_SPLIT_INDEX_EXTENSION); | |
df60cf57 | 3356 | goto out; |
5fc2fc8f NTND |
3357 | } |
3358 | ||
e8ffd034 SG |
3359 | if (test_split_index_env) { |
3360 | if (!si) { | |
3361 | si = init_split_index(istate); | |
d6e3c181 | 3362 | istate->cache_changed |= SPLIT_INDEX_ORDERED; |
e8ffd034 SG |
3363 | } else { |
3364 | int v = si->base_oid.hash[0]; | |
3365 | if ((v & 15) < 6) | |
3366 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
3367 | } | |
d6e3c181 | 3368 | } |
e6a1dd77 CC |
3369 | if (too_many_not_shared_entries(istate)) |
3370 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
0d59ffb4 CC |
3371 | |
3372 | new_shared_index = istate->cache_changed & SPLIT_INDEX_ORDERED; | |
3373 | ||
3374 | if (new_shared_index) { | |
59f9d2dd NTND |
3375 | struct tempfile *temp; |
3376 | int saved_errno; | |
3377 | ||
c9d6c788 ÆAB |
3378 | /* Same initial permissions as the main .git/index file */ |
3379 | temp = mks_tempfile_sm(git_path("sharedindex_XXXXXX"), 0, 0666); | |
59f9d2dd | 3380 | if (!temp) { |
3b7a4475 JS |
3381 | ret = do_write_locked_index(istate, lock, flags, |
3382 | ~WRITE_SPLIT_INDEX_EXTENSION); | |
ef5b3a6c NTND |
3383 | goto out; |
3384 | } | |
ba95e96d | 3385 | ret = write_shared_index(istate, &temp, flags); |
59f9d2dd NTND |
3386 | |
3387 | saved_errno = errno; | |
3388 | if (is_tempfile_active(temp)) | |
3389 | delete_tempfile(&temp); | |
3390 | errno = saved_errno; | |
3391 | ||
c18b80a0 | 3392 | if (ret) |
df60cf57 | 3393 | goto out; |
c18b80a0 NTND |
3394 | } |
3395 | ||
0d59ffb4 CC |
3396 | ret = write_split_index(istate, lock, flags); |
3397 | ||
3398 | /* Freshen the shared index only if the split-index was written */ | |
6e37c8ed | 3399 | if (!ret && !new_shared_index && !is_null_oid(&si->base_oid)) { |
a125a223 | 3400 | const char *shared_index = git_path("sharedindex.%s", |
2182abd9 | 3401 | oid_to_hex(&si->base_oid)); |
a125a223 TG |
3402 | freshen_shared_index(shared_index, 1); |
3403 | } | |
0d59ffb4 | 3404 | |
df60cf57 MÅ |
3405 | out: |
3406 | if (flags & COMMIT_LOCK) | |
3407 | rollback_lock_file(lock); | |
0d59ffb4 | 3408 | return ret; |
03b86647 NTND |
3409 | } |
3410 | ||
e46bbcf6 MV |
3411 | /* |
3412 | * Read the index file that is potentially unmerged into given | |
ad376204 EN |
3413 | * index_state, dropping any unmerged entries to stage #0 (potentially |
3414 | * resulting in a path appearing as both a file and a directory in the | |
3415 | * index; the caller is responsible to clear out the extra entries | |
3416 | * before writing the index to a tree). Returns true if the index is | |
3417 | * unmerged. Callers who want to refuse to work from an unmerged | |
3418 | * state can call this and check its return value, instead of calling | |
3419 | * read_cache(). | |
e46bbcf6 | 3420 | */ |
e1ff0a32 | 3421 | int repo_read_index_unmerged(struct repository *repo) |
e46bbcf6 | 3422 | { |
e1ff0a32 | 3423 | struct index_state *istate; |
e46bbcf6 | 3424 | int i; |
d1a43f2a | 3425 | int unmerged = 0; |
e46bbcf6 | 3426 | |
e1ff0a32 NTND |
3427 | repo_read_index(repo); |
3428 | istate = repo->index; | |
e46bbcf6 MV |
3429 | for (i = 0; i < istate->cache_nr; i++) { |
3430 | struct cache_entry *ce = istate->cache[i]; | |
d1a43f2a | 3431 | struct cache_entry *new_ce; |
a849735b | 3432 | int len; |
d1a43f2a JH |
3433 | |
3434 | if (!ce_stage(ce)) | |
e46bbcf6 | 3435 | continue; |
d1a43f2a | 3436 | unmerged = 1; |
68c4f6a5 | 3437 | len = ce_namelen(ce); |
a849735b | 3438 | new_ce = make_empty_cache_entry(istate, len); |
d1a43f2a | 3439 | memcpy(new_ce->name, ce->name, len); |
b60e188c TG |
3440 | new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED; |
3441 | new_ce->ce_namelen = len; | |
d1a43f2a | 3442 | new_ce->ce_mode = ce->ce_mode; |
ad376204 | 3443 | if (add_index_entry(istate, new_ce, ADD_CACHE_SKIP_DFCHECK)) |
9d0a9e90 | 3444 | return error(_("%s: cannot drop to stage #0"), |
5699d17e | 3445 | new_ce->name); |
e46bbcf6 | 3446 | } |
d1a43f2a | 3447 | return unmerged; |
e46bbcf6 | 3448 | } |
041aee31 | 3449 | |
98fa4738 JK |
3450 | /* |
3451 | * Returns 1 if the path is an "other" path with respect to | |
3452 | * the index; that is, the path is not mentioned in the index at all, | |
3453 | * either as a file, a directory with some files in the index, | |
3454 | * or as an unmerged entry. | |
3455 | * | |
3456 | * We helpfully remove a trailing "/" from directories so that | |
3457 | * the output of read_directory can be used as-is. | |
3458 | */ | |
847a9e5d DS |
3459 | int index_name_is_other(struct index_state *istate, const char *name, |
3460 | int namelen) | |
98fa4738 JK |
3461 | { |
3462 | int pos; | |
3463 | if (namelen && name[namelen - 1] == '/') | |
3464 | namelen--; | |
3465 | pos = index_name_pos(istate, name, namelen); | |
3466 | if (0 <= pos) | |
3467 | return 0; /* exact match */ | |
3468 | pos = -pos - 1; | |
3469 | if (pos < istate->cache_nr) { | |
3470 | struct cache_entry *ce = istate->cache[pos]; | |
3471 | if (ce_namelen(ce) == namelen && | |
3472 | !memcmp(ce->name, name, namelen)) | |
3473 | return 0; /* Yup, this one exists unmerged */ | |
3474 | } | |
3475 | return 1; | |
3476 | } | |
29fb37b2 | 3477 | |
847a9e5d | 3478 | void *read_blob_data_from_index(struct index_state *istate, |
87542508 | 3479 | const char *path, unsigned long *size) |
29fb37b2 LF |
3480 | { |
3481 | int pos, len; | |
3482 | unsigned long sz; | |
3483 | enum object_type type; | |
3484 | void *data; | |
3485 | ||
3486 | len = strlen(path); | |
3487 | pos = index_name_pos(istate, path, len); | |
3488 | if (pos < 0) { | |
3489 | /* | |
3490 | * We might be in the middle of a merge, in which | |
3491 | * case we would read stage #2 (ours). | |
3492 | */ | |
3493 | int i; | |
3494 | for (i = -pos - 1; | |
3495 | (pos < 0 && i < istate->cache_nr && | |
3496 | !strcmp(istate->cache[i]->name, path)); | |
3497 | i++) | |
3498 | if (ce_stage(istate->cache[i]) == 2) | |
3499 | pos = i; | |
3500 | } | |
3501 | if (pos < 0) | |
3502 | return NULL; | |
bc726bd0 ÆAB |
3503 | data = repo_read_object_file(the_repository, &istate->cache[pos]->oid, |
3504 | &type, &sz); | |
29fb37b2 LF |
3505 | if (!data || type != OBJ_BLOB) { |
3506 | free(data); | |
3507 | return NULL; | |
3508 | } | |
ff366825 LF |
3509 | if (size) |
3510 | *size = sz; | |
29fb37b2 LF |
3511 | return data; |
3512 | } | |
38612532 | 3513 | |
edf3b905 DT |
3514 | void move_index_extensions(struct index_state *dst, struct index_state *src) |
3515 | { | |
3516 | dst->untracked = src->untracked; | |
3517 | src->untracked = NULL; | |
836ef2b6 NTND |
3518 | dst->cache_tree = src->cache_tree; |
3519 | src->cache_tree = NULL; | |
edf3b905 | 3520 | } |
a849735b | 3521 | |
8e72d675 JM |
3522 | struct cache_entry *dup_cache_entry(const struct cache_entry *ce, |
3523 | struct index_state *istate) | |
3524 | { | |
3525 | unsigned int size = ce_size(ce); | |
3526 | int mem_pool_allocated; | |
3527 | struct cache_entry *new_entry = make_empty_cache_entry(istate, ce_namelen(ce)); | |
3528 | mem_pool_allocated = new_entry->mem_pool_allocated; | |
3529 | ||
3530 | memcpy(new_entry, ce, size); | |
3531 | new_entry->mem_pool_allocated = mem_pool_allocated; | |
3532 | return new_entry; | |
3533 | } | |
3534 | ||
a849735b JM |
3535 | void discard_cache_entry(struct cache_entry *ce) |
3536 | { | |
8616a2d0 JM |
3537 | if (ce && should_validate_cache_entries()) |
3538 | memset(ce, 0xCD, cache_entry_size(ce->ce_namelen)); | |
3539 | ||
8e72d675 JM |
3540 | if (ce && ce->mem_pool_allocated) |
3541 | return; | |
3542 | ||
a849735b JM |
3543 | free(ce); |
3544 | } | |
8616a2d0 JM |
3545 | |
3546 | int should_validate_cache_entries(void) | |
3547 | { | |
3548 | static int validate_index_cache_entries = -1; | |
3549 | ||
3550 | if (validate_index_cache_entries < 0) { | |
3551 | if (getenv("GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES")) | |
3552 | validate_index_cache_entries = 1; | |
3553 | else | |
3554 | validate_index_cache_entries = 0; | |
3555 | } | |
3556 | ||
3557 | return validate_index_cache_entries; | |
3558 | } | |
3b1d9e04 BP |
3559 | |
3560 | #define EOIE_SIZE (4 + GIT_SHA1_RAWSZ) /* <4-byte offset> + <20-byte hash> */ | |
3561 | #define EOIE_SIZE_WITH_HEADER (4 + 4 + EOIE_SIZE) /* <4-byte signature> + <4-byte length> + EOIE_SIZE */ | |
3562 | ||
3563 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size) | |
3564 | { | |
3565 | /* | |
3566 | * The end of index entries (EOIE) extension is guaranteed to be last | |
3567 | * so that it can be found by scanning backwards from the EOF. | |
3568 | * | |
3569 | * "EOIE" | |
3570 | * <4-byte length> | |
3571 | * <4-byte offset> | |
3572 | * <20-byte hash> | |
3573 | */ | |
3574 | const char *index, *eoie; | |
3575 | uint32_t extsize; | |
3576 | size_t offset, src_offset; | |
3577 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3578 | git_hash_ctx c; | |
3579 | ||
3580 | /* ensure we have an index big enough to contain an EOIE extension */ | |
3581 | if (mmap_size < sizeof(struct cache_header) + EOIE_SIZE_WITH_HEADER + the_hash_algo->rawsz) | |
3582 | return 0; | |
3583 | ||
3584 | /* validate the extension signature */ | |
3585 | index = eoie = mmap + mmap_size - EOIE_SIZE_WITH_HEADER - the_hash_algo->rawsz; | |
3586 | if (CACHE_EXT(index) != CACHE_EXT_ENDOFINDEXENTRIES) | |
3587 | return 0; | |
3588 | index += sizeof(uint32_t); | |
3589 | ||
3590 | /* validate the extension size */ | |
3591 | extsize = get_be32(index); | |
3592 | if (extsize != EOIE_SIZE) | |
3593 | return 0; | |
3594 | index += sizeof(uint32_t); | |
3595 | ||
3596 | /* | |
3597 | * Validate the offset we're going to look for the first extension | |
3598 | * signature is after the index header and before the eoie extension. | |
3599 | */ | |
3600 | offset = get_be32(index); | |
3601 | if (mmap + offset < mmap + sizeof(struct cache_header)) | |
3602 | return 0; | |
3603 | if (mmap + offset >= eoie) | |
3604 | return 0; | |
3605 | index += sizeof(uint32_t); | |
3606 | ||
3607 | /* | |
3608 | * The hash is computed over extension types and their sizes (but not | |
3609 | * their contents). E.g. if we have "TREE" extension that is N-bytes | |
3610 | * long, "REUC" extension that is M-bytes long, followed by "EOIE", | |
3611 | * then the hash would be: | |
3612 | * | |
3613 | * SHA-1("TREE" + <binary representation of N> + | |
3614 | * "REUC" + <binary representation of M>) | |
3615 | */ | |
3616 | src_offset = offset; | |
3617 | the_hash_algo->init_fn(&c); | |
3618 | while (src_offset < mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) { | |
3619 | /* After an array of active_nr index entries, | |
3620 | * there can be arbitrary number of extended | |
3621 | * sections, each of which is prefixed with | |
3622 | * extension name (4-byte) and section length | |
3623 | * in 4-byte network byte order. | |
3624 | */ | |
3625 | uint32_t extsize; | |
3626 | memcpy(&extsize, mmap + src_offset + 4, 4); | |
3627 | extsize = ntohl(extsize); | |
3628 | ||
3629 | /* verify the extension size isn't so large it will wrap around */ | |
3630 | if (src_offset + 8 + extsize < src_offset) | |
3631 | return 0; | |
3632 | ||
3633 | the_hash_algo->update_fn(&c, mmap + src_offset, 8); | |
3634 | ||
3635 | src_offset += 8; | |
3636 | src_offset += extsize; | |
3637 | } | |
3638 | the_hash_algo->final_fn(hash, &c); | |
3639 | if (!hasheq(hash, (const unsigned char *)index)) | |
3640 | return 0; | |
3641 | ||
3642 | /* Validate that the extension offsets returned us back to the eoie extension. */ | |
3643 | if (src_offset != mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) | |
3644 | return 0; | |
3645 | ||
3646 | return offset; | |
3647 | } | |
3648 | ||
3649 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset) | |
3650 | { | |
3651 | uint32_t buffer; | |
3652 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3653 | ||
3654 | /* offset */ | |
3655 | put_be32(&buffer, offset); | |
3656 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3657 | ||
3658 | /* hash */ | |
3659 | the_hash_algo->final_fn(hash, eoie_context); | |
3660 | strbuf_add(sb, hash, the_hash_algo->rawsz); | |
3661 | } | |
3255089a | 3662 | |
3255089a BP |
3663 | #define IEOT_VERSION (1) |
3664 | ||
3665 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset) | |
3666 | { | |
ec36c42a NTND |
3667 | const char *index = NULL; |
3668 | uint32_t extsize, ext_version; | |
3669 | struct index_entry_offset_table *ieot; | |
3670 | int i, nr; | |
3671 | ||
3672 | /* find the IEOT extension */ | |
3673 | if (!offset) | |
3674 | return NULL; | |
3675 | while (offset <= mmap_size - the_hash_algo->rawsz - 8) { | |
3676 | extsize = get_be32(mmap + offset + 4); | |
3677 | if (CACHE_EXT((mmap + offset)) == CACHE_EXT_INDEXENTRYOFFSETTABLE) { | |
3678 | index = mmap + offset + 4 + 4; | |
3679 | break; | |
3680 | } | |
3681 | offset += 8; | |
3682 | offset += extsize; | |
3683 | } | |
3684 | if (!index) | |
3685 | return NULL; | |
3686 | ||
3687 | /* validate the version is IEOT_VERSION */ | |
3688 | ext_version = get_be32(index); | |
3689 | if (ext_version != IEOT_VERSION) { | |
3690 | error("invalid IEOT version %d", ext_version); | |
3691 | return NULL; | |
3692 | } | |
3693 | index += sizeof(uint32_t); | |
3694 | ||
3695 | /* extension size - version bytes / bytes per entry */ | |
3696 | nr = (extsize - sizeof(uint32_t)) / (sizeof(uint32_t) + sizeof(uint32_t)); | |
3697 | if (!nr) { | |
3698 | error("invalid number of IEOT entries %d", nr); | |
3699 | return NULL; | |
3700 | } | |
3701 | ieot = xmalloc(sizeof(struct index_entry_offset_table) | |
3702 | + (nr * sizeof(struct index_entry_offset))); | |
3703 | ieot->nr = nr; | |
3704 | for (i = 0; i < nr; i++) { | |
3705 | ieot->entries[i].offset = get_be32(index); | |
3706 | index += sizeof(uint32_t); | |
3707 | ieot->entries[i].nr = get_be32(index); | |
3708 | index += sizeof(uint32_t); | |
3709 | } | |
3710 | ||
3711 | return ieot; | |
3255089a BP |
3712 | } |
3713 | ||
3714 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot) | |
3715 | { | |
ec36c42a NTND |
3716 | uint32_t buffer; |
3717 | int i; | |
3255089a | 3718 | |
ec36c42a NTND |
3719 | /* version */ |
3720 | put_be32(&buffer, IEOT_VERSION); | |
3721 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3722 | |
ec36c42a NTND |
3723 | /* ieot */ |
3724 | for (i = 0; i < ieot->nr; i++) { | |
3255089a | 3725 | |
ec36c42a NTND |
3726 | /* offset */ |
3727 | put_be32(&buffer, ieot->entries[i].offset); | |
3728 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3729 | |
ec36c42a NTND |
3730 | /* count */ |
3731 | put_be32(&buffer, ieot->entries[i].nr); | |
3732 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3733 | } | |
3255089a | 3734 | } |
b2896d27 JT |
3735 | |
3736 | void prefetch_cache_entries(const struct index_state *istate, | |
3737 | must_prefetch_predicate must_prefetch) | |
3738 | { | |
3739 | int i; | |
3740 | struct oid_array to_fetch = OID_ARRAY_INIT; | |
3741 | ||
3742 | for (i = 0; i < istate->cache_nr; i++) { | |
3743 | struct cache_entry *ce = istate->cache[i]; | |
3744 | ||
3745 | if (S_ISGITLINK(ce->ce_mode) || !must_prefetch(ce)) | |
3746 | continue; | |
3747 | if (!oid_object_info_extended(the_repository, &ce->oid, | |
3748 | NULL, | |
3749 | OBJECT_INFO_FOR_PREFETCH)) | |
3750 | continue; | |
3751 | oid_array_append(&to_fetch, &ce->oid); | |
3752 | } | |
3753 | promisor_remote_get_direct(the_repository, | |
3754 | to_fetch.oid, to_fetch.nr); | |
3755 | oid_array_clear(&to_fetch); | |
3756 | } | |
1a40e7be EN |
3757 | |
3758 | static int read_one_entry_opt(struct index_state *istate, | |
3759 | const struct object_id *oid, | |
3760 | struct strbuf *base, | |
3761 | const char *pathname, | |
3762 | unsigned mode, int opt) | |
3763 | { | |
3764 | int len; | |
3765 | struct cache_entry *ce; | |
3766 | ||
3767 | if (S_ISDIR(mode)) | |
3768 | return READ_TREE_RECURSIVE; | |
3769 | ||
3770 | len = strlen(pathname); | |
3771 | ce = make_empty_cache_entry(istate, base->len + len); | |
3772 | ||
3773 | ce->ce_mode = create_ce_mode(mode); | |
3774 | ce->ce_flags = create_ce_flags(1); | |
3775 | ce->ce_namelen = base->len + len; | |
3776 | memcpy(ce->name, base->buf, base->len); | |
3777 | memcpy(ce->name + base->len, pathname, len+1); | |
3778 | oidcpy(&ce->oid, oid); | |
3779 | return add_index_entry(istate, ce, opt); | |
3780 | } | |
3781 | ||
3782 | static int read_one_entry(const struct object_id *oid, struct strbuf *base, | |
3783 | const char *pathname, unsigned mode, | |
3784 | void *context) | |
3785 | { | |
3786 | struct index_state *istate = context; | |
3787 | return read_one_entry_opt(istate, oid, base, pathname, | |
3788 | mode, | |
3789 | ADD_CACHE_OK_TO_ADD|ADD_CACHE_SKIP_DFCHECK); | |
3790 | } | |
3791 | ||
3792 | /* | |
3793 | * This is used when the caller knows there is no existing entries at | |
3794 | * the stage that will conflict with the entry being added. | |
3795 | */ | |
3796 | static int read_one_entry_quick(const struct object_id *oid, struct strbuf *base, | |
3797 | const char *pathname, unsigned mode, | |
3798 | void *context) | |
3799 | { | |
3800 | struct index_state *istate = context; | |
3801 | return read_one_entry_opt(istate, oid, base, pathname, | |
3802 | mode, ADD_CACHE_JUST_APPEND); | |
3803 | } | |
3804 | ||
3805 | /* | |
3806 | * Read the tree specified with --with-tree option | |
3807 | * (typically, HEAD) into stage #1 and then | |
3808 | * squash them down to stage #0. This is used for | |
3809 | * --error-unmatch to list and check the path patterns | |
3810 | * that were given from the command line. We are not | |
3811 | * going to write this index out. | |
3812 | */ | |
3813 | void overlay_tree_on_index(struct index_state *istate, | |
3814 | const char *tree_name, const char *prefix) | |
3815 | { | |
3816 | struct tree *tree; | |
3817 | struct object_id oid; | |
3818 | struct pathspec pathspec; | |
3819 | struct cache_entry *last_stage0 = NULL; | |
3820 | int i; | |
3821 | read_tree_fn_t fn = NULL; | |
3822 | int err; | |
3823 | ||
3824 | if (repo_get_oid(the_repository, tree_name, &oid)) | |
3825 | die("tree-ish %s not found.", tree_name); | |
3826 | tree = parse_tree_indirect(&oid); | |
3827 | if (!tree) | |
3828 | die("bad tree-ish %s", tree_name); | |
3829 | ||
3830 | /* Hoist the unmerged entries up to stage #3 to make room */ | |
3831 | /* TODO: audit for interaction with sparse-index. */ | |
3832 | ensure_full_index(istate); | |
3833 | for (i = 0; i < istate->cache_nr; i++) { | |
3834 | struct cache_entry *ce = istate->cache[i]; | |
3835 | if (!ce_stage(ce)) | |
3836 | continue; | |
3837 | ce->ce_flags |= CE_STAGEMASK; | |
3838 | } | |
3839 | ||
3840 | if (prefix) { | |
3841 | static const char *(matchbuf[1]); | |
3842 | matchbuf[0] = NULL; | |
3843 | parse_pathspec(&pathspec, PATHSPEC_ALL_MAGIC, | |
3844 | PATHSPEC_PREFER_CWD, prefix, matchbuf); | |
3845 | } else | |
3846 | memset(&pathspec, 0, sizeof(pathspec)); | |
3847 | ||
3848 | /* | |
3849 | * See if we have cache entry at the stage. If so, | |
3850 | * do it the original slow way, otherwise, append and then | |
3851 | * sort at the end. | |
3852 | */ | |
3853 | for (i = 0; !fn && i < istate->cache_nr; i++) { | |
3854 | const struct cache_entry *ce = istate->cache[i]; | |
3855 | if (ce_stage(ce) == 1) | |
3856 | fn = read_one_entry; | |
3857 | } | |
3858 | ||
3859 | if (!fn) | |
3860 | fn = read_one_entry_quick; | |
3861 | err = read_tree(the_repository, tree, &pathspec, fn, istate); | |
3862 | clear_pathspec(&pathspec); | |
3863 | if (err) | |
3864 | die("unable to read tree entries %s", tree_name); | |
3865 | ||
3866 | /* | |
3867 | * Sort the cache entry -- we need to nuke the cache tree, though. | |
3868 | */ | |
3869 | if (fn == read_one_entry_quick) { | |
3870 | cache_tree_free(&istate->cache_tree); | |
3871 | QSORT(istate->cache, istate->cache_nr, cmp_cache_name_compare); | |
3872 | } | |
3873 | ||
3874 | for (i = 0; i < istate->cache_nr; i++) { | |
3875 | struct cache_entry *ce = istate->cache[i]; | |
3876 | switch (ce_stage(ce)) { | |
3877 | case 0: | |
3878 | last_stage0 = ce; | |
3879 | /* fallthru */ | |
3880 | default: | |
3881 | continue; | |
3882 | case 1: | |
3883 | /* | |
3884 | * If there is stage #0 entry for this, we do not | |
3885 | * need to show it. We use CE_UPDATE bit to mark | |
3886 | * such an entry. | |
3887 | */ | |
3888 | if (last_stage0 && | |
3889 | !strcmp(last_stage0->name, ce->name)) | |
3890 | ce->ce_flags |= CE_UPDATE; | |
3891 | } | |
3892 | } | |
3893 | } | |
6cee5ebc EN |
3894 | |
3895 | struct update_callback_data { | |
3896 | struct index_state *index; | |
3897 | int include_sparse; | |
3898 | int flags; | |
3899 | int add_errors; | |
3900 | }; | |
3901 | ||
3902 | static int fix_unmerged_status(struct diff_filepair *p, | |
3903 | struct update_callback_data *data) | |
3904 | { | |
3905 | if (p->status != DIFF_STATUS_UNMERGED) | |
3906 | return p->status; | |
3907 | if (!(data->flags & ADD_CACHE_IGNORE_REMOVAL) && !p->two->mode) | |
3908 | /* | |
3909 | * This is not an explicit add request, and the | |
3910 | * path is missing from the working tree (deleted) | |
3911 | */ | |
3912 | return DIFF_STATUS_DELETED; | |
3913 | else | |
3914 | /* | |
3915 | * Either an explicit add request, or path exists | |
3916 | * in the working tree. An attempt to explicitly | |
3917 | * add a path that does not exist in the working tree | |
3918 | * will be caught as an error by the caller immediately. | |
3919 | */ | |
3920 | return DIFF_STATUS_MODIFIED; | |
3921 | } | |
3922 | ||
3923 | static void update_callback(struct diff_queue_struct *q, | |
3924 | struct diff_options *opt UNUSED, void *cbdata) | |
3925 | { | |
3926 | int i; | |
3927 | struct update_callback_data *data = cbdata; | |
3928 | ||
3929 | for (i = 0; i < q->nr; i++) { | |
3930 | struct diff_filepair *p = q->queue[i]; | |
3931 | const char *path = p->one->path; | |
3932 | ||
3933 | if (!data->include_sparse && | |
3934 | !path_in_sparse_checkout(path, data->index)) | |
3935 | continue; | |
3936 | ||
3937 | switch (fix_unmerged_status(p, data)) { | |
3938 | default: | |
3939 | die(_("unexpected diff status %c"), p->status); | |
3940 | case DIFF_STATUS_MODIFIED: | |
3941 | case DIFF_STATUS_TYPE_CHANGED: | |
3942 | if (add_file_to_index(data->index, path, data->flags)) { | |
3943 | if (!(data->flags & ADD_CACHE_IGNORE_ERRORS)) | |
3944 | die(_("updating files failed")); | |
3945 | data->add_errors++; | |
3946 | } | |
3947 | break; | |
3948 | case DIFF_STATUS_DELETED: | |
3949 | if (data->flags & ADD_CACHE_IGNORE_REMOVAL) | |
3950 | break; | |
3951 | if (!(data->flags & ADD_CACHE_PRETEND)) | |
3952 | remove_file_from_index(data->index, path); | |
3953 | if (data->flags & (ADD_CACHE_PRETEND|ADD_CACHE_VERBOSE)) | |
3954 | printf(_("remove '%s'\n"), path); | |
3955 | break; | |
3956 | } | |
3957 | } | |
3958 | } | |
3959 | ||
3960 | int add_files_to_cache(struct repository *repo, const char *prefix, | |
3961 | const struct pathspec *pathspec, int include_sparse, | |
3962 | int flags) | |
3963 | { | |
3964 | struct update_callback_data data; | |
3965 | struct rev_info rev; | |
3966 | ||
3967 | memset(&data, 0, sizeof(data)); | |
3968 | data.index = repo->index; | |
3969 | data.include_sparse = include_sparse; | |
3970 | data.flags = flags; | |
3971 | ||
3972 | repo_init_revisions(repo, &rev, prefix); | |
3973 | setup_revisions(0, NULL, &rev, NULL); | |
3974 | if (pathspec) | |
3975 | copy_pathspec(&rev.prune_data, pathspec); | |
3976 | rev.diffopt.output_format = DIFF_FORMAT_CALLBACK; | |
3977 | rev.diffopt.format_callback = update_callback; | |
3978 | rev.diffopt.format_callback_data = &data; | |
3979 | rev.diffopt.flags.override_submodule_config = 1; | |
3980 | rev.max_count = 0; /* do not compare unmerged paths with stage #2 */ | |
3981 | ||
3982 | /* | |
3983 | * Use an ODB transaction to optimize adding multiple objects. | |
3984 | * This function is invoked from commands other than 'add', which | |
3985 | * may not have their own transaction active. | |
3986 | */ | |
3987 | begin_odb_transaction(); | |
3988 | run_diff_files(&rev, DIFF_RACY_IS_MODIFIED); | |
3989 | end_odb_transaction(); | |
3990 | ||
3991 | release_revisions(&rev); | |
3992 | return !!data.add_errors; | |
3993 | } |