]>
Commit | Line | Data |
---|---|---|
8bc9a0c7 LT |
1 | /* |
2 | * GIT - The information manager from hell | |
3 | * | |
4 | * Copyright (C) Linus Torvalds, 2005 | |
5 | */ | |
e83c5163 | 6 | #include "cache.h" |
b2141fc1 | 7 | #include "config.h" |
cffbfad5 EN |
8 | #include "diff.h" |
9 | #include "diffcore.h" | |
f6ecc62d | 10 | #include "tempfile.h" |
697cc8ef | 11 | #include "lockfile.h" |
bad68ec9 | 12 | #include "cache-tree.h" |
f35a6d3b | 13 | #include "refs.h" |
d616813d | 14 | #include "dir.h" |
cbd53a21 | 15 | #include "object-store.h" |
041aee31 JH |
16 | #include "tree.h" |
17 | #include "commit.h" | |
39425819 | 18 | #include "blob.h" |
cfc5789a | 19 | #include "resolve-undo.h" |
1956ecd0 | 20 | #include "run-command.h" |
6c9cd161 JH |
21 | #include "strbuf.h" |
22 | #include "varint.h" | |
5fc2fc8f | 23 | #include "split-index.h" |
a42643aa | 24 | #include "utf8.h" |
883e248b | 25 | #include "fsmonitor.h" |
abb4bb83 | 26 | #include "thread-utils.h" |
ae9af122 | 27 | #include "progress.h" |
bad68ec9 | 28 | |
b60e188c TG |
29 | /* Mask for the name length in ce_flags in the on-disk index */ |
30 | ||
31 | #define CE_NAMEMASK (0x0fff) | |
32 | ||
bad68ec9 JH |
33 | /* Index extensions. |
34 | * | |
35 | * The first letter should be 'A'..'Z' for extensions that are not | |
36 | * necessary for a correct operation (i.e. optimization data). | |
37 | * When new extensions are added that _needs_ to be understood in | |
38 | * order to correctly interpret the index file, pick character that | |
39 | * is outside the range, to cause the reader to abort. | |
40 | */ | |
41 | ||
42 | #define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) ) | |
43 | #define CACHE_EXT_TREE 0x54524545 /* "TREE" */ | |
b659b49b | 44 | #define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */ |
5fc2fc8f | 45 | #define CACHE_EXT_LINK 0x6c696e6b /* "link" */ |
83c094ad | 46 | #define CACHE_EXT_UNTRACKED 0x554E5452 /* "UNTR" */ |
883e248b | 47 | #define CACHE_EXT_FSMONITOR 0x46534D4E /* "FSMN" */ |
3b1d9e04 | 48 | #define CACHE_EXT_ENDOFINDEXENTRIES 0x454F4945 /* "EOIE" */ |
3255089a | 49 | #define CACHE_EXT_INDEXENTRYOFFSETTABLE 0x49454F54 /* "IEOT" */ |
5fc2fc8f NTND |
50 | |
51 | /* changes that can be kept in $GIT_DIR/index (basically all extensions) */ | |
e0cf0d7d | 52 | #define EXTMASK (RESOLVE_UNDO_CHANGED | CACHE_TREE_CHANGED | \ |
c18b80a0 | 53 | CE_ENTRY_ADDED | CE_ENTRY_REMOVED | CE_ENTRY_CHANGED | \ |
883e248b | 54 | SPLIT_INDEX_ORDERED | UNTRACKED_CHANGED | FSMONITOR_CHANGED) |
e83c5163 | 55 | |
8e72d675 JM |
56 | |
57 | /* | |
58 | * This is an estimate of the pathname length in the index. We use | |
59 | * this for V4 index files to guess the un-deltafied size of the index | |
60 | * in memory because of pathname deltafication. This is not required | |
61 | * for V2/V3 index formats because their pathnames are not compressed. | |
62 | * If the initial amount of memory set aside is not sufficient, the | |
63 | * mem pool will allocate extra memory. | |
64 | */ | |
65 | #define CACHE_ENTRY_PATH_LENGTH 80 | |
66 | ||
67 | static inline struct cache_entry *mem_pool__ce_alloc(struct mem_pool *mem_pool, size_t len) | |
68 | { | |
69 | struct cache_entry *ce; | |
70 | ce = mem_pool_alloc(mem_pool, cache_entry_size(len)); | |
71 | ce->mem_pool_allocated = 1; | |
72 | return ce; | |
73 | } | |
74 | ||
75 | static inline struct cache_entry *mem_pool__ce_calloc(struct mem_pool *mem_pool, size_t len) | |
76 | { | |
77 | struct cache_entry * ce; | |
78 | ce = mem_pool_calloc(mem_pool, 1, cache_entry_size(len)); | |
79 | ce->mem_pool_allocated = 1; | |
80 | return ce; | |
81 | } | |
82 | ||
83 | static struct mem_pool *find_mem_pool(struct index_state *istate) | |
84 | { | |
85 | struct mem_pool **pool_ptr; | |
86 | ||
87 | if (istate->split_index && istate->split_index->base) | |
88 | pool_ptr = &istate->split_index->base->ce_mem_pool; | |
89 | else | |
90 | pool_ptr = &istate->ce_mem_pool; | |
91 | ||
92 | if (!*pool_ptr) | |
93 | mem_pool_init(pool_ptr, 0); | |
94 | ||
95 | return *pool_ptr; | |
96 | } | |
97 | ||
626f35c8 | 98 | static const char *alternate_index_output; |
8fd2cb40 | 99 | |
9cb76b8c JH |
100 | static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
101 | { | |
102 | istate->cache[nr] = ce; | |
96872bc2 | 103 | add_name_hash(istate, ce); |
9cb76b8c JH |
104 | } |
105 | ||
cf558704 LT |
106 | static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) |
107 | { | |
108 | struct cache_entry *old = istate->cache[nr]; | |
109 | ||
078a58e8 | 110 | replace_index_entry_in_base(istate, old, ce); |
2092678c | 111 | remove_name_hash(istate, old); |
a849735b | 112 | discard_cache_entry(old); |
0e267b7a | 113 | ce->ce_flags &= ~CE_HASHED; |
a22c6371 | 114 | set_index_entry(istate, nr, ce); |
078a58e8 | 115 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 116 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 117 | istate->cache_changed |= CE_ENTRY_CHANGED; |
cf558704 LT |
118 | } |
119 | ||
81dc2307 PB |
120 | void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name) |
121 | { | |
285c2e25 | 122 | struct cache_entry *old_entry = istate->cache[nr], *new_entry; |
81dc2307 PB |
123 | int namelen = strlen(new_name); |
124 | ||
a849735b | 125 | new_entry = make_empty_cache_entry(istate, namelen); |
285c2e25 BW |
126 | copy_cache_entry(new_entry, old_entry); |
127 | new_entry->ce_flags &= ~CE_HASHED; | |
128 | new_entry->ce_namelen = namelen; | |
129 | new_entry->index = 0; | |
130 | memcpy(new_entry->name, new_name, namelen + 1); | |
81dc2307 | 131 | |
285c2e25 BW |
132 | cache_tree_invalidate_path(istate, old_entry->name); |
133 | untracked_cache_remove_from_index(istate, old_entry->name); | |
81dc2307 | 134 | remove_index_entry_at(istate, nr); |
285c2e25 | 135 | add_index_entry(istate, new_entry, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); |
81dc2307 PB |
136 | } |
137 | ||
c21d39d7 MH |
138 | void fill_stat_data(struct stat_data *sd, struct stat *st) |
139 | { | |
140 | sd->sd_ctime.sec = (unsigned int)st->st_ctime; | |
141 | sd->sd_mtime.sec = (unsigned int)st->st_mtime; | |
142 | sd->sd_ctime.nsec = ST_CTIME_NSEC(*st); | |
143 | sd->sd_mtime.nsec = ST_MTIME_NSEC(*st); | |
144 | sd->sd_dev = st->st_dev; | |
145 | sd->sd_ino = st->st_ino; | |
146 | sd->sd_uid = st->st_uid; | |
147 | sd->sd_gid = st->st_gid; | |
148 | sd->sd_size = st->st_size; | |
149 | } | |
150 | ||
151 | int match_stat_data(const struct stat_data *sd, struct stat *st) | |
152 | { | |
153 | int changed = 0; | |
154 | ||
155 | if (sd->sd_mtime.sec != (unsigned int)st->st_mtime) | |
156 | changed |= MTIME_CHANGED; | |
157 | if (trust_ctime && check_stat && | |
158 | sd->sd_ctime.sec != (unsigned int)st->st_ctime) | |
159 | changed |= CTIME_CHANGED; | |
160 | ||
161 | #ifdef USE_NSEC | |
162 | if (check_stat && sd->sd_mtime.nsec != ST_MTIME_NSEC(*st)) | |
163 | changed |= MTIME_CHANGED; | |
164 | if (trust_ctime && check_stat && | |
165 | sd->sd_ctime.nsec != ST_CTIME_NSEC(*st)) | |
166 | changed |= CTIME_CHANGED; | |
167 | #endif | |
168 | ||
169 | if (check_stat) { | |
170 | if (sd->sd_uid != (unsigned int) st->st_uid || | |
171 | sd->sd_gid != (unsigned int) st->st_gid) | |
172 | changed |= OWNER_CHANGED; | |
173 | if (sd->sd_ino != (unsigned int) st->st_ino) | |
174 | changed |= INODE_CHANGED; | |
175 | } | |
176 | ||
177 | #ifdef USE_STDEV | |
178 | /* | |
179 | * st_dev breaks on network filesystems where different | |
180 | * clients will have different views of what "device" | |
181 | * the filesystem is on | |
182 | */ | |
183 | if (check_stat && sd->sd_dev != (unsigned int) st->st_dev) | |
184 | changed |= INODE_CHANGED; | |
185 | #endif | |
186 | ||
187 | if (sd->sd_size != (unsigned int) st->st_size) | |
188 | changed |= DATA_CHANGED; | |
189 | ||
190 | return changed; | |
191 | } | |
192 | ||
415e96c8 JH |
193 | /* |
194 | * This only updates the "non-critical" parts of the directory | |
195 | * cache, ie the parts that aren't tracked by GIT, and only used | |
196 | * to validate the cache. | |
197 | */ | |
198 | void fill_stat_cache_info(struct cache_entry *ce, struct stat *st) | |
199 | { | |
c21d39d7 | 200 | fill_stat_data(&ce->ce_stat_data, st); |
5f73076c JH |
201 | |
202 | if (assume_unchanged) | |
7a51ed66 | 203 | ce->ce_flags |= CE_VALID; |
eadb5831 | 204 | |
883e248b | 205 | if (S_ISREG(st->st_mode)) { |
eadb5831 | 206 | ce_mark_uptodate(ce); |
883e248b BP |
207 | mark_fsmonitor_valid(ce); |
208 | } | |
415e96c8 JH |
209 | } |
210 | ||
58bf2a4c NTND |
211 | static int ce_compare_data(struct index_state *istate, |
212 | const struct cache_entry *ce, | |
213 | struct stat *st) | |
29e4d363 JH |
214 | { |
215 | int match = -1; | |
1b8ac5ea | 216 | int fd = git_open_cloexec(ce->name, O_RDONLY); |
29e4d363 JH |
217 | |
218 | if (fd >= 0) { | |
bebfecb9 | 219 | struct object_id oid; |
58bf2a4c | 220 | if (!index_fd(istate, &oid, fd, st, OBJ_BLOB, ce->name, 0)) |
6a29d7b7 | 221 | match = !oideq(&oid, &ce->oid); |
7f8508e8 | 222 | /* index_fd() closed the file descriptor already */ |
29e4d363 JH |
223 | } |
224 | return match; | |
225 | } | |
226 | ||
21a6b9fa | 227 | static int ce_compare_link(const struct cache_entry *ce, size_t expected_size) |
29e4d363 JH |
228 | { |
229 | int match = -1; | |
29e4d363 JH |
230 | void *buffer; |
231 | unsigned long size; | |
21666f1a | 232 | enum object_type type; |
a60272b3 | 233 | struct strbuf sb = STRBUF_INIT; |
29e4d363 | 234 | |
a60272b3 | 235 | if (strbuf_readlink(&sb, ce->name, expected_size)) |
29e4d363 | 236 | return -1; |
a60272b3 | 237 | |
b4f5aca4 | 238 | buffer = read_object_file(&ce->oid, &type, &size); |
a60272b3 LT |
239 | if (buffer) { |
240 | if (size == sb.len) | |
241 | match = memcmp(buffer, sb.buf, size); | |
242 | free(buffer); | |
29e4d363 | 243 | } |
a60272b3 | 244 | strbuf_release(&sb); |
29e4d363 JH |
245 | return match; |
246 | } | |
247 | ||
21a6b9fa | 248 | static int ce_compare_gitlink(const struct cache_entry *ce) |
f35a6d3b | 249 | { |
1053fe82 | 250 | struct object_id oid; |
f35a6d3b LT |
251 | |
252 | /* | |
253 | * We don't actually require that the .git directory | |
302b9282 | 254 | * under GITLINK directory be a valid git directory. It |
f35a6d3b LT |
255 | * might even be missing (in case nobody populated that |
256 | * sub-project). | |
257 | * | |
258 | * If so, we consider it always to match. | |
259 | */ | |
a98e6101 | 260 | if (resolve_gitlink_ref(ce->name, "HEAD", &oid) < 0) |
f35a6d3b | 261 | return 0; |
6a29d7b7 | 262 | return !oideq(&oid, &ce->oid); |
f35a6d3b LT |
263 | } |
264 | ||
58bf2a4c NTND |
265 | static int ce_modified_check_fs(struct index_state *istate, |
266 | const struct cache_entry *ce, | |
267 | struct stat *st) | |
29e4d363 JH |
268 | { |
269 | switch (st->st_mode & S_IFMT) { | |
270 | case S_IFREG: | |
58bf2a4c | 271 | if (ce_compare_data(istate, ce, st)) |
29e4d363 JH |
272 | return DATA_CHANGED; |
273 | break; | |
274 | case S_IFLNK: | |
dc49cd76 | 275 | if (ce_compare_link(ce, xsize_t(st->st_size))) |
29e4d363 JH |
276 | return DATA_CHANGED; |
277 | break; | |
a8ee75bc | 278 | case S_IFDIR: |
7a51ed66 | 279 | if (S_ISGITLINK(ce->ce_mode)) |
c70115b4 | 280 | return ce_compare_gitlink(ce) ? DATA_CHANGED : 0; |
1cf01a34 | 281 | /* else fallthrough */ |
29e4d363 JH |
282 | default: |
283 | return TYPE_CHANGED; | |
284 | } | |
285 | return 0; | |
286 | } | |
287 | ||
21a6b9fa | 288 | static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st) |
734aab75 LT |
289 | { |
290 | unsigned int changed = 0; | |
291 | ||
7a51ed66 LT |
292 | if (ce->ce_flags & CE_REMOVE) |
293 | return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED; | |
294 | ||
295 | switch (ce->ce_mode & S_IFMT) { | |
8ae0a8c5 KS |
296 | case S_IFREG: |
297 | changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0; | |
3e09cdfd JH |
298 | /* We consider only the owner x bit to be relevant for |
299 | * "mode changes" | |
300 | */ | |
301 | if (trust_executable_bit && | |
7a51ed66 | 302 | (0100 & (ce->ce_mode ^ st->st_mode))) |
ffbe1add | 303 | changed |= MODE_CHANGED; |
8ae0a8c5 KS |
304 | break; |
305 | case S_IFLNK: | |
78a8d641 JS |
306 | if (!S_ISLNK(st->st_mode) && |
307 | (has_symlinks || !S_ISREG(st->st_mode))) | |
308 | changed |= TYPE_CHANGED; | |
8ae0a8c5 | 309 | break; |
302b9282 | 310 | case S_IFGITLINK: |
c70115b4 | 311 | /* We ignore most of the st_xxx fields for gitlinks */ |
f35a6d3b LT |
312 | if (!S_ISDIR(st->st_mode)) |
313 | changed |= TYPE_CHANGED; | |
314 | else if (ce_compare_gitlink(ce)) | |
315 | changed |= DATA_CHANGED; | |
a8ee75bc | 316 | return changed; |
8ae0a8c5 | 317 | default: |
391408e5 | 318 | BUG("unsupported ce_mode: %o", ce->ce_mode); |
8ae0a8c5 | 319 | } |
2cb45e95 | 320 | |
c21d39d7 | 321 | changed |= match_stat_data(&ce->ce_stat_data, st); |
b0391890 | 322 | |
f49c2c22 | 323 | /* Racily smudged entry? */ |
c21d39d7 | 324 | if (!ce->ce_stat_data.sd_size) { |
99d1a986 | 325 | if (!is_empty_blob_sha1(ce->oid.hash)) |
f49c2c22 LT |
326 | changed |= DATA_CHANGED; |
327 | } | |
328 | ||
407c8eb0 JH |
329 | return changed; |
330 | } | |
331 | ||
2bb4cda1 NTND |
332 | static int is_racy_stat(const struct index_state *istate, |
333 | const struct stat_data *sd) | |
6d91da6d | 334 | { |
2bb4cda1 | 335 | return (istate->timestamp.sec && |
fba2f38a KB |
336 | #ifdef USE_NSEC |
337 | /* nanosecond timestamped files can also be racy! */ | |
2bb4cda1 NTND |
338 | (istate->timestamp.sec < sd->sd_mtime.sec || |
339 | (istate->timestamp.sec == sd->sd_mtime.sec && | |
340 | istate->timestamp.nsec <= sd->sd_mtime.nsec)) | |
fba2f38a | 341 | #else |
2bb4cda1 | 342 | istate->timestamp.sec <= sd->sd_mtime.sec |
fba2f38a | 343 | #endif |
2bb4cda1 NTND |
344 | ); |
345 | } | |
346 | ||
5581a019 | 347 | int is_racy_timestamp(const struct index_state *istate, |
2bb4cda1 NTND |
348 | const struct cache_entry *ce) |
349 | { | |
350 | return (!S_ISGITLINK(ce->ce_mode) && | |
351 | is_racy_stat(istate, &ce->ce_stat_data)); | |
6d91da6d JH |
352 | } |
353 | ||
ed4efab1 NTND |
354 | int match_stat_data_racy(const struct index_state *istate, |
355 | const struct stat_data *sd, struct stat *st) | |
356 | { | |
357 | if (is_racy_stat(istate, sd)) | |
358 | return MTIME_CHANGED; | |
359 | return match_stat_data(sd, st); | |
6d91da6d JH |
360 | } |
361 | ||
883e248b | 362 | int ie_match_stat(struct index_state *istate, |
21a6b9fa | 363 | const struct cache_entry *ce, struct stat *st, |
4bd5b7da | 364 | unsigned int options) |
407c8eb0 | 365 | { |
5f73076c | 366 | unsigned int changed; |
4bd5b7da | 367 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 368 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
4bd5b7da | 369 | int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY; |
883e248b | 370 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
5f73076c | 371 | |
883e248b BP |
372 | if (!ignore_fsmonitor) |
373 | refresh_fsmonitor(istate); | |
5f73076c JH |
374 | /* |
375 | * If it's marked as always valid in the index, it's | |
376 | * valid whatever the checked-out copy says. | |
56cac48c NTND |
377 | * |
378 | * skip-worktree has the same effect with higher precedence | |
5f73076c | 379 | */ |
56cac48c NTND |
380 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) |
381 | return 0; | |
7a51ed66 | 382 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) |
5f73076c | 383 | return 0; |
883e248b BP |
384 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) |
385 | return 0; | |
5f73076c | 386 | |
331fcb59 JH |
387 | /* |
388 | * Intent-to-add entries have not been added, so the index entry | |
389 | * by definition never matches what is in the work tree until it | |
390 | * actually gets added. | |
391 | */ | |
895ff3b2 | 392 | if (ce_intent_to_add(ce)) |
331fcb59 JH |
393 | return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED; |
394 | ||
5f73076c | 395 | changed = ce_match_stat_basic(ce, st); |
407c8eb0 | 396 | |
29e4d363 JH |
397 | /* |
398 | * Within 1 second of this sequence: | |
399 | * echo xyzzy >file && git-update-index --add file | |
400 | * running this command: | |
401 | * echo frotz >file | |
402 | * would give a falsely clean cache entry. The mtime and | |
403 | * length match the cache, and other stat fields do not change. | |
404 | * | |
405 | * We could detect this at update-index time (the cache entry | |
406 | * being registered/updated records the same time as "now") | |
407 | * and delay the return from git-update-index, but that would | |
408 | * effectively mean we can make at most one commit per second, | |
409 | * which is not acceptable. Instead, we check cache entries | |
410 | * whose mtime are the same as the index file timestamp more | |
5f73076c | 411 | * carefully than others. |
29e4d363 | 412 | */ |
6d91da6d | 413 | if (!changed && is_racy_timestamp(istate, ce)) { |
42f77406 JH |
414 | if (assume_racy_is_modified) |
415 | changed |= DATA_CHANGED; | |
416 | else | |
58bf2a4c | 417 | changed |= ce_modified_check_fs(istate, ce, st); |
42f77406 | 418 | } |
b0391890 | 419 | |
29e4d363 | 420 | return changed; |
b0391890 JH |
421 | } |
422 | ||
883e248b | 423 | int ie_modified(struct index_state *istate, |
21a6b9fa RS |
424 | const struct cache_entry *ce, |
425 | struct stat *st, unsigned int options) | |
b0391890 | 426 | { |
29e4d363 | 427 | int changed, changed_fs; |
4bd5b7da JH |
428 | |
429 | changed = ie_match_stat(istate, ce, st, options); | |
b0391890 JH |
430 | if (!changed) |
431 | return 0; | |
b0391890 JH |
432 | /* |
433 | * If the mode or type has changed, there's no point in trying | |
434 | * to refresh the entry - it's not going to match | |
435 | */ | |
436 | if (changed & (MODE_CHANGED | TYPE_CHANGED)) | |
437 | return changed; | |
438 | ||
c70115b4 JH |
439 | /* |
440 | * Immediately after read-tree or update-index --cacheinfo, | |
441 | * the length field is zero, as we have never even read the | |
442 | * lstat(2) information once, and we cannot trust DATA_CHANGED | |
443 | * returned by ie_match_stat() which in turn was returned by | |
444 | * ce_match_stat_basic() to signal that the filesize of the | |
445 | * blob changed. We have to actually go to the filesystem to | |
446 | * see if the contents match, and if so, should answer "unchanged". | |
447 | * | |
448 | * The logic does not apply to gitlinks, as ce_match_stat_basic() | |
449 | * already has checked the actual HEAD from the filesystem in the | |
450 | * subproject. If ie_match_stat() already said it is different, | |
451 | * then we know it is. | |
b0391890 | 452 | */ |
c70115b4 | 453 | if ((changed & DATA_CHANGED) && |
c21d39d7 | 454 | (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0)) |
b0391890 JH |
455 | return changed; |
456 | ||
58bf2a4c | 457 | changed_fs = ce_modified_check_fs(istate, ce, st); |
29e4d363 JH |
458 | if (changed_fs) |
459 | return changed | changed_fs; | |
b0391890 JH |
460 | return 0; |
461 | } | |
462 | ||
958ba6c9 LT |
463 | int base_name_compare(const char *name1, int len1, int mode1, |
464 | const char *name2, int len2, int mode2) | |
465 | { | |
466 | unsigned char c1, c2; | |
467 | int len = len1 < len2 ? len1 : len2; | |
468 | int cmp; | |
469 | ||
470 | cmp = memcmp(name1, name2, len); | |
471 | if (cmp) | |
472 | return cmp; | |
473 | c1 = name1[len]; | |
474 | c2 = name2[len]; | |
1833a925 | 475 | if (!c1 && S_ISDIR(mode1)) |
958ba6c9 | 476 | c1 = '/'; |
1833a925 | 477 | if (!c2 && S_ISDIR(mode2)) |
958ba6c9 LT |
478 | c2 = '/'; |
479 | return (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0; | |
480 | } | |
481 | ||
0ab9e1e8 LT |
482 | /* |
483 | * df_name_compare() is identical to base_name_compare(), except it | |
484 | * compares conflicting directory/file entries as equal. Note that | |
485 | * while a directory name compares as equal to a regular file, they | |
486 | * then individually compare _differently_ to a filename that has | |
487 | * a dot after the basename (because '\0' < '.' < '/'). | |
488 | * | |
489 | * This is used by routines that want to traverse the git namespace | |
490 | * but then handle conflicting entries together when possible. | |
491 | */ | |
492 | int df_name_compare(const char *name1, int len1, int mode1, | |
493 | const char *name2, int len2, int mode2) | |
494 | { | |
495 | int len = len1 < len2 ? len1 : len2, cmp; | |
496 | unsigned char c1, c2; | |
497 | ||
498 | cmp = memcmp(name1, name2, len); | |
499 | if (cmp) | |
500 | return cmp; | |
501 | /* Directories and files compare equal (same length, same name) */ | |
502 | if (len1 == len2) | |
503 | return 0; | |
504 | c1 = name1[len]; | |
505 | if (!c1 && S_ISDIR(mode1)) | |
506 | c1 = '/'; | |
507 | c2 = name2[len]; | |
508 | if (!c2 && S_ISDIR(mode2)) | |
509 | c2 = '/'; | |
510 | if (c1 == '/' && !c2) | |
511 | return 0; | |
512 | if (c2 == '/' && !c1) | |
513 | return 0; | |
514 | return c1 - c2; | |
515 | } | |
516 | ||
ccdd4a0f | 517 | int name_compare(const char *name1, size_t len1, const char *name2, size_t len2) |
eb38c22f | 518 | { |
ccdd4a0f JM |
519 | size_t min_len = (len1 < len2) ? len1 : len2; |
520 | int cmp = memcmp(name1, name2, min_len); | |
eb38c22f LT |
521 | if (cmp) |
522 | return cmp; | |
523 | if (len1 < len2) | |
524 | return -1; | |
525 | if (len1 > len2) | |
526 | return 1; | |
ccdd4a0f JM |
527 | return 0; |
528 | } | |
529 | ||
530 | int cache_name_stage_compare(const char *name1, int len1, int stage1, const char *name2, int len2, int stage2) | |
531 | { | |
532 | int cmp; | |
533 | ||
534 | cmp = name_compare(name1, len1, name2, len2); | |
535 | if (cmp) | |
536 | return cmp; | |
5f73076c | 537 | |
b60e188c | 538 | if (stage1 < stage2) |
95fd5bf8 | 539 | return -1; |
b60e188c | 540 | if (stage1 > stage2) |
95fd5bf8 | 541 | return 1; |
eb38c22f LT |
542 | return 0; |
543 | } | |
544 | ||
357e9c69 | 545 | static int index_name_stage_pos(const struct index_state *istate, const char *name, int namelen, int stage) |
eb38c22f LT |
546 | { |
547 | int first, last; | |
548 | ||
549 | first = 0; | |
4aab5b46 | 550 | last = istate->cache_nr; |
eb38c22f LT |
551 | while (last > first) { |
552 | int next = (last + first) >> 1; | |
4aab5b46 | 553 | struct cache_entry *ce = istate->cache[next]; |
b60e188c | 554 | int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce)); |
eb38c22f | 555 | if (!cmp) |
76e7f4ec | 556 | return next; |
eb38c22f LT |
557 | if (cmp < 0) { |
558 | last = next; | |
559 | continue; | |
560 | } | |
561 | first = next+1; | |
562 | } | |
76e7f4ec | 563 | return -first-1; |
eb38c22f LT |
564 | } |
565 | ||
b60e188c TG |
566 | int index_name_pos(const struct index_state *istate, const char *name, int namelen) |
567 | { | |
568 | return index_name_stage_pos(istate, name, namelen, 0); | |
569 | } | |
570 | ||
4aab5b46 | 571 | int remove_index_entry_at(struct index_state *istate, int pos) |
7b937ca3 | 572 | { |
cf558704 LT |
573 | struct cache_entry *ce = istate->cache[pos]; |
574 | ||
cfc5789a | 575 | record_resolve_undo(istate, ce); |
2092678c | 576 | remove_name_hash(istate, ce); |
045113a5 | 577 | save_or_free_index_entry(istate, ce); |
e636a7b4 | 578 | istate->cache_changed |= CE_ENTRY_REMOVED; |
4aab5b46 JH |
579 | istate->cache_nr--; |
580 | if (pos >= istate->cache_nr) | |
7b937ca3 | 581 | return 0; |
f331ab9d RS |
582 | MOVE_ARRAY(istate->cache + pos, istate->cache + pos + 1, |
583 | istate->cache_nr - pos); | |
7b937ca3 LT |
584 | return 1; |
585 | } | |
586 | ||
36419c8e | 587 | /* |
98e023de | 588 | * Remove all cache entries marked for removal, that is where |
36419c8e KB |
589 | * CE_REMOVE is set in ce_flags. This is much more effective than |
590 | * calling remove_index_entry_at() for each entry to be removed. | |
591 | */ | |
6fdc2057 | 592 | void remove_marked_cache_entries(struct index_state *istate, int invalidate) |
36419c8e KB |
593 | { |
594 | struct cache_entry **ce_array = istate->cache; | |
595 | unsigned int i, j; | |
596 | ||
597 | for (i = j = 0; i < istate->cache_nr; i++) { | |
5699d17e | 598 | if (ce_array[i]->ce_flags & CE_REMOVE) { |
6fdc2057 TG |
599 | if (invalidate) { |
600 | cache_tree_invalidate_path(istate, | |
601 | ce_array[i]->name); | |
602 | untracked_cache_remove_from_index(istate, | |
603 | ce_array[i]->name); | |
604 | } | |
2092678c | 605 | remove_name_hash(istate, ce_array[i]); |
045113a5 | 606 | save_or_free_index_entry(istate, ce_array[i]); |
5699d17e | 607 | } |
36419c8e KB |
608 | else |
609 | ce_array[j++] = ce_array[i]; | |
610 | } | |
ad837d9e NTND |
611 | if (j == istate->cache_nr) |
612 | return; | |
e636a7b4 | 613 | istate->cache_changed |= CE_ENTRY_REMOVED; |
36419c8e KB |
614 | istate->cache_nr = j; |
615 | } | |
616 | ||
4aab5b46 | 617 | int remove_file_from_index(struct index_state *istate, const char *path) |
197ee8c9 | 618 | { |
4aab5b46 | 619 | int pos = index_name_pos(istate, path, strlen(path)); |
c4e3cca1 JH |
620 | if (pos < 0) |
621 | pos = -pos-1; | |
a5400efe | 622 | cache_tree_invalidate_path(istate, path); |
e931371a | 623 | untracked_cache_remove_from_index(istate, path); |
4aab5b46 JH |
624 | while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path)) |
625 | remove_index_entry_at(istate, pos); | |
197ee8c9 LT |
626 | return 0; |
627 | } | |
628 | ||
20314271 JS |
629 | static int compare_name(struct cache_entry *ce, const char *path, int namelen) |
630 | { | |
631 | return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen); | |
632 | } | |
633 | ||
634 | static int index_name_pos_also_unmerged(struct index_state *istate, | |
635 | const char *path, int namelen) | |
636 | { | |
637 | int pos = index_name_pos(istate, path, namelen); | |
638 | struct cache_entry *ce; | |
639 | ||
640 | if (pos >= 0) | |
641 | return pos; | |
642 | ||
643 | /* maybe unmerged? */ | |
644 | pos = -1 - pos; | |
645 | if (pos >= istate->cache_nr || | |
646 | compare_name((ce = istate->cache[pos]), path, namelen)) | |
647 | return -1; | |
648 | ||
649 | /* order of preference: stage 2, 1, 3 */ | |
650 | if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr && | |
651 | ce_stage((ce = istate->cache[pos + 1])) == 2 && | |
652 | !compare_name(ce, path, namelen)) | |
653 | pos++; | |
654 | return pos; | |
655 | } | |
656 | ||
1102952b LT |
657 | static int different_name(struct cache_entry *ce, struct cache_entry *alias) |
658 | { | |
659 | int len = ce_namelen(ce); | |
660 | return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len); | |
661 | } | |
662 | ||
663 | /* | |
664 | * If we add a filename that aliases in the cache, we will use the | |
665 | * name that we already have - but we don't want to update the same | |
666 | * alias twice, because that implies that there were actually two | |
667 | * different files with aliasing names! | |
668 | * | |
669 | * So we use the CE_ADDED flag to verify that the alias was an old | |
670 | * one before we accept it as | |
671 | */ | |
045113a5 NTND |
672 | static struct cache_entry *create_alias_ce(struct index_state *istate, |
673 | struct cache_entry *ce, | |
674 | struct cache_entry *alias) | |
1102952b LT |
675 | { |
676 | int len; | |
285c2e25 | 677 | struct cache_entry *new_entry; |
1102952b LT |
678 | |
679 | if (alias->ce_flags & CE_ADDED) | |
9d0a9e90 NTND |
680 | die(_("will not add file alias '%s' ('%s' already exists in index)"), |
681 | ce->name, alias->name); | |
1102952b LT |
682 | |
683 | /* Ok, create the new entry using the name of the existing alias */ | |
684 | len = ce_namelen(alias); | |
a849735b | 685 | new_entry = make_empty_cache_entry(istate, len); |
285c2e25 BW |
686 | memcpy(new_entry->name, alias->name, len); |
687 | copy_cache_entry(new_entry, ce); | |
045113a5 | 688 | save_or_free_index_entry(istate, ce); |
285c2e25 | 689 | return new_entry; |
1102952b LT |
690 | } |
691 | ||
b4b313f9 | 692 | void set_object_name_for_intent_to_add_entry(struct cache_entry *ce) |
39425819 | 693 | { |
a09c985e PO |
694 | struct object_id oid; |
695 | if (write_object_file("", 0, blob_type, &oid)) | |
9d0a9e90 | 696 | die(_("cannot create an empty blob in the object database")); |
a09c985e | 697 | oidcpy(&ce->oid, &oid); |
39425819 JH |
698 | } |
699 | ||
610d55af | 700 | int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags) |
11be42a4 | 701 | { |
a849735b | 702 | int namelen, was_same; |
d177cab0 | 703 | mode_t st_mode = st->st_mode; |
9472935d | 704 | struct cache_entry *ce, *alias = NULL; |
56cac48c | 705 | unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY; |
38ed1d89 JH |
706 | int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND); |
707 | int pretend = flags & ADD_CACHE_PRETEND; | |
39425819 JH |
708 | int intent_only = flags & ADD_CACHE_INTENT; |
709 | int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE| | |
710 | (intent_only ? ADD_CACHE_NEW_ONLY : 0)); | |
9e5da3d0 | 711 | int hash_flags = HASH_WRITE_OBJECT; |
9472935d | 712 | |
9e5da3d0 JK |
713 | if (flags & ADD_CACHE_RENORMALIZE) |
714 | hash_flags |= HASH_RENORMALIZE; | |
11be42a4 | 715 | |
d177cab0 | 716 | if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode)) |
9d0a9e90 | 717 | return error(_("%s: can only add regular files, symbolic links or git-directories"), path); |
11be42a4 JS |
718 | |
719 | namelen = strlen(path); | |
d177cab0 | 720 | if (S_ISDIR(st_mode)) { |
09595258 LT |
721 | while (namelen && path[namelen-1] == '/') |
722 | namelen--; | |
723 | } | |
a849735b | 724 | ce = make_empty_cache_entry(istate, namelen); |
11be42a4 | 725 | memcpy(ce->name, path, namelen); |
b60e188c | 726 | ce->ce_namelen = namelen; |
39425819 JH |
727 | if (!intent_only) |
728 | fill_stat_cache_info(ce, st); | |
388b2acd JH |
729 | else |
730 | ce->ce_flags |= CE_INTENT_TO_ADD; | |
11be42a4 | 731 | |
610d55af TG |
732 | |
733 | if (trust_executable_bit && has_symlinks) { | |
d177cab0 | 734 | ce->ce_mode = create_ce_mode(st_mode); |
610d55af | 735 | } else { |
78a8d641 JS |
736 | /* If there is an existing entry, pick the mode bits and type |
737 | * from it, otherwise assume unexecutable regular file. | |
11be42a4 | 738 | */ |
185c975f | 739 | struct cache_entry *ent; |
20314271 | 740 | int pos = index_name_pos_also_unmerged(istate, path, namelen); |
185c975f | 741 | |
4aab5b46 | 742 | ent = (0 <= pos) ? istate->cache[pos] : NULL; |
d177cab0 | 743 | ce->ce_mode = ce_mode_from_stat(ent, st_mode); |
11be42a4 JS |
744 | } |
745 | ||
dc1ae704 JJ |
746 | /* When core.ignorecase=true, determine if a directory of the same name but differing |
747 | * case already exists within the Git repository. If it does, ensure the directory | |
748 | * case of the file being added to the repository matches (is folded into) the existing | |
749 | * entry's directory case. | |
750 | */ | |
751 | if (ignore_case) { | |
41284eb0 | 752 | adjust_dirname_case(istate, ce->name); |
dc1ae704 | 753 | } |
e2c2a375 | 754 | if (!(flags & ADD_CACHE_RENORMALIZE)) { |
9472935d TB |
755 | alias = index_file_exists(istate, ce->name, |
756 | ce_namelen(ce), ignore_case); | |
757 | if (alias && | |
758 | !ce_stage(alias) && | |
759 | !ie_match_stat(istate, alias, st, ce_option)) { | |
760 | /* Nothing changed, really */ | |
761 | if (!S_ISGITLINK(alias->ce_mode)) | |
762 | ce_mark_uptodate(alias); | |
763 | alias->ce_flags |= CE_ADDED; | |
dc1ae704 | 764 | |
a849735b | 765 | discard_cache_entry(ce); |
9472935d TB |
766 | return 0; |
767 | } | |
0781b8a9 | 768 | } |
39425819 | 769 | if (!intent_only) { |
1c418243 | 770 | if (index_path(istate, &ce->oid, path, st, hash_flags)) { |
a849735b | 771 | discard_cache_entry(ce); |
9d0a9e90 | 772 | return error(_("unable to index file '%s'"), path); |
2d9426b0 | 773 | } |
39425819 | 774 | } else |
b4b313f9 | 775 | set_object_name_for_intent_to_add_entry(ce); |
39425819 | 776 | |
1102952b | 777 | if (ignore_case && alias && different_name(ce, alias)) |
045113a5 | 778 | ce = create_alias_ce(istate, ce, alias); |
1102952b | 779 | ce->ce_flags |= CE_ADDED; |
38ed1d89 | 780 | |
3bf0dd1f | 781 | /* It was suspected to be racily clean, but it turns out to be Ok */ |
38ed1d89 JH |
782 | was_same = (alias && |
783 | !ce_stage(alias) && | |
4a7e27e9 | 784 | oideq(&alias->oid, &ce->oid) && |
38ed1d89 JH |
785 | ce->ce_mode == alias->ce_mode); |
786 | ||
787 | if (pretend) | |
a849735b | 788 | discard_cache_entry(ce); |
067178ed | 789 | else if (add_index_entry(istate, ce, add_option)) { |
a849735b | 790 | discard_cache_entry(ce); |
9d0a9e90 | 791 | return error(_("unable to add '%s' to index"), path); |
067178ed | 792 | } |
38ed1d89 | 793 | if (verbose && !was_same) |
11be42a4 | 794 | printf("add '%s'\n", path); |
11be42a4 JS |
795 | return 0; |
796 | } | |
797 | ||
610d55af | 798 | int add_file_to_index(struct index_state *istate, const char *path, int flags) |
d177cab0 LT |
799 | { |
800 | struct stat st; | |
801 | if (lstat(path, &st)) | |
9d0a9e90 | 802 | die_errno(_("unable to stat '%s'"), path); |
610d55af | 803 | return add_to_index(istate, path, &st, flags); |
d177cab0 LT |
804 | } |
805 | ||
a849735b JM |
806 | struct cache_entry *make_empty_cache_entry(struct index_state *istate, size_t len) |
807 | { | |
8e72d675 | 808 | return mem_pool__ce_calloc(find_mem_pool(istate), len); |
a849735b JM |
809 | } |
810 | ||
811 | struct cache_entry *make_empty_transient_cache_entry(size_t len) | |
812 | { | |
813 | return xcalloc(1, cache_entry_size(len)); | |
814 | } | |
815 | ||
816 | struct cache_entry *make_cache_entry(struct index_state *istate, | |
817 | unsigned int mode, | |
825ed4d9 JM |
818 | const struct object_id *oid, |
819 | const char *path, | |
820 | int stage, | |
821 | unsigned int refresh_options) | |
6640f881 | 822 | { |
bc1c2caa | 823 | struct cache_entry *ce, *ret; |
a849735b | 824 | int len; |
6640f881 | 825 | |
10ecfa76 | 826 | if (!verify_path(path, mode)) { |
9d0a9e90 | 827 | error(_("invalid path '%s'"), path); |
6640f881 | 828 | return NULL; |
7e7abea9 | 829 | } |
6640f881 CR |
830 | |
831 | len = strlen(path); | |
a849735b | 832 | ce = make_empty_cache_entry(istate, len); |
6640f881 | 833 | |
825ed4d9 | 834 | oidcpy(&ce->oid, oid); |
6640f881 | 835 | memcpy(ce->name, path, len); |
b60e188c TG |
836 | ce->ce_flags = create_ce_flags(stage); |
837 | ce->ce_namelen = len; | |
6640f881 CR |
838 | ce->ce_mode = create_ce_mode(mode); |
839 | ||
d7b665c3 | 840 | ret = refresh_cache_entry(istate, ce, refresh_options); |
915e44c6 | 841 | if (ret != ce) |
a849735b | 842 | discard_cache_entry(ce); |
915e44c6 | 843 | return ret; |
6640f881 CR |
844 | } |
845 | ||
a849735b JM |
846 | struct cache_entry *make_transient_cache_entry(unsigned int mode, const struct object_id *oid, |
847 | const char *path, int stage) | |
848 | { | |
849 | struct cache_entry *ce; | |
850 | int len; | |
851 | ||
852 | if (!verify_path(path, mode)) { | |
9d0a9e90 | 853 | error(_("invalid path '%s'"), path); |
a849735b JM |
854 | return NULL; |
855 | } | |
856 | ||
857 | len = strlen(path); | |
858 | ce = make_empty_transient_cache_entry(len); | |
859 | ||
860 | oidcpy(&ce->oid, oid); | |
861 | memcpy(ce->name, path, len); | |
862 | ce->ce_flags = create_ce_flags(stage); | |
863 | ce->ce_namelen = len; | |
864 | ce->ce_mode = create_ce_mode(mode); | |
865 | ||
866 | return ce; | |
867 | } | |
868 | ||
d9d70966 TG |
869 | /* |
870 | * Chmod an index entry with either +x or -x. | |
871 | * | |
872 | * Returns -1 if the chmod for the particular cache entry failed (if it's | |
873 | * not a regular file), -2 if an invalid flip argument is passed in, 0 | |
874 | * otherwise. | |
875 | */ | |
876 | int chmod_index_entry(struct index_state *istate, struct cache_entry *ce, | |
877 | char flip) | |
878 | { | |
879 | if (!S_ISREG(ce->ce_mode)) | |
880 | return -1; | |
881 | switch (flip) { | |
882 | case '+': | |
883 | ce->ce_mode |= 0111; | |
884 | break; | |
885 | case '-': | |
886 | ce->ce_mode &= ~0111; | |
887 | break; | |
888 | default: | |
889 | return -2; | |
890 | } | |
891 | cache_tree_invalidate_path(istate, ce->name); | |
892 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
883e248b | 893 | mark_fsmonitor_invalid(istate, ce); |
d9d70966 TG |
894 | istate->cache_changed |= CE_ENTRY_CHANGED; |
895 | ||
896 | return 0; | |
897 | } | |
898 | ||
9c5e6c80 | 899 | int ce_same_name(const struct cache_entry *a, const struct cache_entry *b) |
7b937ca3 LT |
900 | { |
901 | int len = ce_namelen(a); | |
902 | return ce_namelen(b) == len && !memcmp(a->name, b->name, len); | |
903 | } | |
904 | ||
8dcf39c4 LT |
905 | /* |
906 | * We fundamentally don't like some paths: we don't want | |
907 | * dot or dot-dot anywhere, and for obvious reasons don't | |
908 | * want to recurse into ".git" either. | |
909 | * | |
910 | * Also, we don't want double slashes or slashes at the | |
911 | * end that can make pathnames ambiguous. | |
912 | */ | |
10ecfa76 | 913 | static int verify_dotfile(const char *rest, unsigned mode) |
8dcf39c4 LT |
914 | { |
915 | /* | |
916 | * The first character was '.', but that | |
917 | * has already been discarded, we now test | |
918 | * the rest. | |
919 | */ | |
e0f530ff | 920 | |
8dcf39c4 | 921 | /* "." is not allowed */ |
e0f530ff | 922 | if (*rest == '\0' || is_dir_sep(*rest)) |
8dcf39c4 LT |
923 | return 0; |
924 | ||
e0f530ff | 925 | switch (*rest) { |
8dcf39c4 | 926 | /* |
641084b6 JK |
927 | * ".git" followed by NUL or slash is bad. Note that we match |
928 | * case-insensitively here, even if ignore_case is not set. | |
929 | * This outlaws ".GIT" everywhere out of an abundance of caution, | |
930 | * since there's really no good reason to allow it. | |
10ecfa76 JK |
931 | * |
932 | * Once we've seen ".git", we can also find ".gitmodules", etc (also | |
933 | * case-insensitively). | |
8dcf39c4 LT |
934 | */ |
935 | case 'g': | |
cc2fc7c2 JK |
936 | case 'G': |
937 | if (rest[1] != 'i' && rest[1] != 'I') | |
8dcf39c4 | 938 | break; |
cc2fc7c2 | 939 | if (rest[2] != 't' && rest[2] != 'T') |
8dcf39c4 | 940 | break; |
e19e5e66 JK |
941 | if (rest[3] == '\0' || is_dir_sep(rest[3])) |
942 | return 0; | |
10ecfa76 JK |
943 | if (S_ISLNK(mode)) { |
944 | rest += 3; | |
945 | if (skip_iprefix(rest, "modules", &rest) && | |
946 | (*rest == '\0' || is_dir_sep(*rest))) | |
947 | return 0; | |
948 | } | |
e19e5e66 | 949 | break; |
8dcf39c4 | 950 | case '.': |
e0f530ff | 951 | if (rest[1] == '\0' || is_dir_sep(rest[1])) |
8dcf39c4 LT |
952 | return 0; |
953 | } | |
954 | return 1; | |
955 | } | |
956 | ||
10ecfa76 | 957 | int verify_path(const char *path, unsigned mode) |
8dcf39c4 LT |
958 | { |
959 | char c; | |
960 | ||
56948cb6 EFL |
961 | if (has_dos_drive_prefix(path)) |
962 | return 0; | |
963 | ||
8dcf39c4 LT |
964 | goto inside; |
965 | for (;;) { | |
966 | if (!c) | |
967 | return 1; | |
56948cb6 | 968 | if (is_dir_sep(c)) { |
8dcf39c4 | 969 | inside: |
10ecfa76 JK |
970 | if (protect_hfs) { |
971 | if (is_hfs_dotgit(path)) | |
972 | return 0; | |
973 | if (S_ISLNK(mode)) { | |
974 | if (is_hfs_dotgitmodules(path)) | |
975 | return 0; | |
976 | } | |
977 | } | |
978 | if (protect_ntfs) { | |
979 | if (is_ntfs_dotgit(path)) | |
980 | return 0; | |
981 | if (S_ISLNK(mode)) { | |
982 | if (is_ntfs_dotgitmodules(path)) | |
983 | return 0; | |
984 | } | |
985 | } | |
986 | ||
8dcf39c4 | 987 | c = *path++; |
10ecfa76 | 988 | if ((c == '.' && !verify_dotfile(path, mode)) || |
3bdf09c7 JH |
989 | is_dir_sep(c) || c == '\0') |
990 | return 0; | |
8dcf39c4 LT |
991 | } |
992 | c = *path++; | |
993 | } | |
994 | } | |
995 | ||
12676608 LT |
996 | /* |
997 | * Do we have another file that has the beginning components being a | |
998 | * proper superset of the name we're trying to add? | |
0f1e4f04 | 999 | */ |
4aab5b46 JH |
1000 | static int has_file_name(struct index_state *istate, |
1001 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
0f1e4f04 | 1002 | { |
12676608 LT |
1003 | int retval = 0; |
1004 | int len = ce_namelen(ce); | |
b155725d | 1005 | int stage = ce_stage(ce); |
12676608 | 1006 | const char *name = ce->name; |
0f1e4f04 | 1007 | |
4aab5b46 JH |
1008 | while (pos < istate->cache_nr) { |
1009 | struct cache_entry *p = istate->cache[pos++]; | |
0f1e4f04 | 1010 | |
12676608 | 1011 | if (len >= ce_namelen(p)) |
0f1e4f04 | 1012 | break; |
12676608 LT |
1013 | if (memcmp(name, p->name, len)) |
1014 | break; | |
b155725d JH |
1015 | if (ce_stage(p) != stage) |
1016 | continue; | |
12676608 LT |
1017 | if (p->name[len] != '/') |
1018 | continue; | |
7a51ed66 | 1019 | if (p->ce_flags & CE_REMOVE) |
21cd8d00 | 1020 | continue; |
12676608 LT |
1021 | retval = -1; |
1022 | if (!ok_to_replace) | |
1023 | break; | |
4aab5b46 | 1024 | remove_index_entry_at(istate, --pos); |
0f1e4f04 | 1025 | } |
12676608 LT |
1026 | return retval; |
1027 | } | |
0f1e4f04 | 1028 | |
a6db3fbb JH |
1029 | |
1030 | /* | |
1031 | * Like strcmp(), but also return the offset of the first change. | |
1032 | * If strings are equal, return the length. | |
1033 | */ | |
1034 | int strcmp_offset(const char *s1, const char *s2, size_t *first_change) | |
1035 | { | |
1036 | size_t k; | |
1037 | ||
1038 | if (!first_change) | |
1039 | return strcmp(s1, s2); | |
1040 | ||
1041 | for (k = 0; s1[k] == s2[k]; k++) | |
1042 | if (s1[k] == '\0') | |
1043 | break; | |
1044 | ||
1045 | *first_change = k; | |
1046 | return (unsigned char)s1[k] - (unsigned char)s2[k]; | |
1047 | } | |
1048 | ||
12676608 LT |
1049 | /* |
1050 | * Do we have another file with a pathname that is a proper | |
1051 | * subset of the name we're trying to add? | |
06b6d81b JH |
1052 | * |
1053 | * That is, is there another file in the index with a path | |
1054 | * that matches a sub-directory in the given entry? | |
12676608 | 1055 | */ |
4aab5b46 JH |
1056 | static int has_dir_name(struct index_state *istate, |
1057 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
12676608 LT |
1058 | { |
1059 | int retval = 0; | |
b155725d | 1060 | int stage = ce_stage(ce); |
12676608 LT |
1061 | const char *name = ce->name; |
1062 | const char *slash = name + ce_namelen(ce); | |
06b6d81b JH |
1063 | size_t len_eq_last; |
1064 | int cmp_last = 0; | |
1065 | ||
1066 | /* | |
1067 | * We are frequently called during an iteration on a sorted | |
1068 | * list of pathnames and while building a new index. Therefore, | |
1069 | * there is a high probability that this entry will eventually | |
1070 | * be appended to the index, rather than inserted in the middle. | |
1071 | * If we can confirm that, we can avoid binary searches on the | |
1072 | * components of the pathname. | |
1073 | * | |
1074 | * Compare the entry's full path with the last path in the index. | |
1075 | */ | |
1076 | if (istate->cache_nr > 0) { | |
1077 | cmp_last = strcmp_offset(name, | |
1078 | istate->cache[istate->cache_nr - 1]->name, | |
1079 | &len_eq_last); | |
1080 | if (cmp_last > 0) { | |
1081 | if (len_eq_last == 0) { | |
1082 | /* | |
1083 | * The entry sorts AFTER the last one in the | |
1084 | * index and their paths have no common prefix, | |
1085 | * so there cannot be a F/D conflict. | |
1086 | */ | |
1087 | return retval; | |
1088 | } else { | |
1089 | /* | |
1090 | * The entry sorts AFTER the last one in the | |
1091 | * index, but has a common prefix. Fall through | |
1092 | * to the loop below to disect the entry's path | |
1093 | * and see where the difference is. | |
1094 | */ | |
1095 | } | |
1096 | } else if (cmp_last == 0) { | |
1097 | /* | |
1098 | * The entry exactly matches the last one in the | |
1099 | * index, but because of multiple stage and CE_REMOVE | |
1100 | * items, we fall through and let the regular search | |
1101 | * code handle it. | |
1102 | */ | |
1103 | } | |
1104 | } | |
0f1e4f04 | 1105 | |
12676608 | 1106 | for (;;) { |
b986df5c | 1107 | size_t len; |
0f1e4f04 | 1108 | |
12676608 LT |
1109 | for (;;) { |
1110 | if (*--slash == '/') | |
1111 | break; | |
1112 | if (slash <= ce->name) | |
1113 | return retval; | |
1114 | } | |
1115 | len = slash - name; | |
0f1e4f04 | 1116 | |
b986df5c JH |
1117 | if (cmp_last > 0) { |
1118 | /* | |
1119 | * (len + 1) is a directory boundary (including | |
1120 | * the trailing slash). And since the loop is | |
1121 | * decrementing "slash", the first iteration is | |
1122 | * the longest directory prefix; subsequent | |
1123 | * iterations consider parent directories. | |
1124 | */ | |
1125 | ||
1126 | if (len + 1 <= len_eq_last) { | |
1127 | /* | |
1128 | * The directory prefix (including the trailing | |
1129 | * slash) also appears as a prefix in the last | |
1130 | * entry, so the remainder cannot collide (because | |
1131 | * strcmp said the whole path was greater). | |
1132 | * | |
1133 | * EQ: last: xxx/A | |
1134 | * this: xxx/B | |
1135 | * | |
1136 | * LT: last: xxx/file_A | |
1137 | * this: xxx/file_B | |
1138 | */ | |
1139 | return retval; | |
1140 | } | |
1141 | ||
1142 | if (len > len_eq_last) { | |
1143 | /* | |
1144 | * This part of the directory prefix (excluding | |
1145 | * the trailing slash) is longer than the known | |
1146 | * equal portions, so this sub-directory cannot | |
1147 | * collide with a file. | |
1148 | * | |
1149 | * GT: last: xxxA | |
1150 | * this: xxxB/file | |
1151 | */ | |
1152 | return retval; | |
1153 | } | |
1154 | ||
1155 | if (istate->cache_nr > 0 && | |
1156 | ce_namelen(istate->cache[istate->cache_nr - 1]) > len) { | |
1157 | /* | |
1158 | * The directory prefix lines up with part of | |
1159 | * a longer file or directory name, but sorts | |
1160 | * after it, so this sub-directory cannot | |
1161 | * collide with a file. | |
1162 | * | |
1163 | * last: xxx/yy-file (because '-' sorts before '/') | |
1164 | * this: xxx/yy/abc | |
1165 | */ | |
1166 | return retval; | |
1167 | } | |
1168 | ||
1169 | /* | |
1170 | * This is a possible collision. Fall through and | |
1171 | * let the regular search code handle it. | |
1172 | * | |
1173 | * last: xxx | |
1174 | * this: xxx/file | |
1175 | */ | |
1176 | } | |
1177 | ||
b60e188c | 1178 | pos = index_name_stage_pos(istate, name, len, stage); |
12676608 | 1179 | if (pos >= 0) { |
21cd8d00 JH |
1180 | /* |
1181 | * Found one, but not so fast. This could | |
1182 | * be a marker that says "I was here, but | |
1183 | * I am being removed". Such an entry is | |
1184 | * not a part of the resulting tree, and | |
1185 | * it is Ok to have a directory at the same | |
1186 | * path. | |
1187 | */ | |
077c48df | 1188 | if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) { |
21cd8d00 JH |
1189 | retval = -1; |
1190 | if (!ok_to_replace) | |
1191 | break; | |
4aab5b46 | 1192 | remove_index_entry_at(istate, pos); |
21cd8d00 JH |
1193 | continue; |
1194 | } | |
12676608 | 1195 | } |
21cd8d00 JH |
1196 | else |
1197 | pos = -pos-1; | |
12676608 LT |
1198 | |
1199 | /* | |
1200 | * Trivial optimization: if we find an entry that | |
1201 | * already matches the sub-directory, then we know | |
b155725d | 1202 | * we're ok, and we can exit. |
12676608 | 1203 | */ |
4aab5b46 JH |
1204 | while (pos < istate->cache_nr) { |
1205 | struct cache_entry *p = istate->cache[pos]; | |
b155725d JH |
1206 | if ((ce_namelen(p) <= len) || |
1207 | (p->name[len] != '/') || | |
1208 | memcmp(p->name, name, len)) | |
1209 | break; /* not our subdirectory */ | |
077c48df JH |
1210 | if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE)) |
1211 | /* | |
1212 | * p is at the same stage as our entry, and | |
b155725d JH |
1213 | * is a subdirectory of what we are looking |
1214 | * at, so we cannot have conflicts at our | |
1215 | * level or anything shorter. | |
1216 | */ | |
1217 | return retval; | |
1218 | pos++; | |
192268c1 | 1219 | } |
0f1e4f04 | 1220 | } |
12676608 LT |
1221 | return retval; |
1222 | } | |
1223 | ||
1224 | /* We may be in a situation where we already have path/file and path | |
1225 | * is being added, or we already have path and path/file is being | |
1226 | * added. Either one would result in a nonsense tree that has path | |
1227 | * twice when git-write-tree tries to write it out. Prevent it. | |
a6080a0a | 1228 | * |
12676608 LT |
1229 | * If ok-to-replace is specified, we remove the conflicting entries |
1230 | * from the cache so the caller should recompute the insert position. | |
1231 | * When this happens, we return non-zero. | |
1232 | */ | |
4aab5b46 JH |
1233 | static int check_file_directory_conflict(struct index_state *istate, |
1234 | const struct cache_entry *ce, | |
1235 | int pos, int ok_to_replace) | |
12676608 | 1236 | { |
21cd8d00 JH |
1237 | int retval; |
1238 | ||
1239 | /* | |
1240 | * When ce is an "I am going away" entry, we allow it to be added | |
1241 | */ | |
7a51ed66 | 1242 | if (ce->ce_flags & CE_REMOVE) |
21cd8d00 JH |
1243 | return 0; |
1244 | ||
12676608 LT |
1245 | /* |
1246 | * We check if the path is a sub-path of a subsequent pathname | |
1247 | * first, since removing those will not change the position | |
21cd8d00 | 1248 | * in the array. |
12676608 | 1249 | */ |
4aab5b46 | 1250 | retval = has_file_name(istate, ce, pos, ok_to_replace); |
21cd8d00 | 1251 | |
12676608 LT |
1252 | /* |
1253 | * Then check if the path might have a clashing sub-directory | |
1254 | * before it. | |
1255 | */ | |
4aab5b46 | 1256 | return retval + has_dir_name(istate, ce, pos, ok_to_replace); |
0f1e4f04 JH |
1257 | } |
1258 | ||
af3785dc | 1259 | static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option) |
197ee8c9 LT |
1260 | { |
1261 | int pos; | |
192268c1 JH |
1262 | int ok_to_add = option & ADD_CACHE_OK_TO_ADD; |
1263 | int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE; | |
b155725d | 1264 | int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK; |
39425819 | 1265 | int new_only = option & ADD_CACHE_NEW_ONLY; |
5f73076c | 1266 | |
ce7c614b NTND |
1267 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1268 | cache_tree_invalidate_path(istate, ce->name); | |
e5494631 JH |
1269 | |
1270 | /* | |
1271 | * If this entry's path sorts after the last entry in the index, | |
1272 | * we can avoid searching for it. | |
1273 | */ | |
1274 | if (istate->cache_nr > 0 && | |
1275 | strcmp(ce->name, istate->cache[istate->cache_nr - 1]->name) > 0) | |
1276 | pos = -istate->cache_nr - 1; | |
1277 | else | |
1278 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce)); | |
197ee8c9 | 1279 | |
3e09cdfd | 1280 | /* existing match? Just replace it. */ |
76e7f4ec | 1281 | if (pos >= 0) { |
39425819 JH |
1282 | if (!new_only) |
1283 | replace_index_entry(istate, pos, ce); | |
197ee8c9 LT |
1284 | return 0; |
1285 | } | |
76e7f4ec | 1286 | pos = -pos-1; |
197ee8c9 | 1287 | |
ffcc9ba7 NTND |
1288 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) |
1289 | untracked_cache_add_to_index(istate, ce->name); | |
e931371a | 1290 | |
7b937ca3 LT |
1291 | /* |
1292 | * Inserting a merged entry ("stage 0") into the index | |
1293 | * will always replace all non-merged entries.. | |
1294 | */ | |
4aab5b46 JH |
1295 | if (pos < istate->cache_nr && ce_stage(ce) == 0) { |
1296 | while (ce_same_name(istate->cache[pos], ce)) { | |
7b937ca3 | 1297 | ok_to_add = 1; |
4aab5b46 | 1298 | if (!remove_index_entry_at(istate, pos)) |
7b937ca3 LT |
1299 | break; |
1300 | } | |
1301 | } | |
1302 | ||
121481ab LT |
1303 | if (!ok_to_add) |
1304 | return -1; | |
10ecfa76 | 1305 | if (!verify_path(ce->name, ce->ce_mode)) |
9d0a9e90 | 1306 | return error(_("invalid path '%s'"), ce->name); |
121481ab | 1307 | |
3e09cdfd | 1308 | if (!skip_df_check && |
4aab5b46 | 1309 | check_file_directory_conflict(istate, ce, pos, ok_to_replace)) { |
192268c1 | 1310 | if (!ok_to_replace) |
9d0a9e90 | 1311 | return error(_("'%s' appears as both a file and as a directory"), |
4aab5b46 | 1312 | ce->name); |
b60e188c | 1313 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce)); |
192268c1 JH |
1314 | pos = -pos-1; |
1315 | } | |
af3785dc JH |
1316 | return pos + 1; |
1317 | } | |
1318 | ||
1319 | int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option) | |
1320 | { | |
1321 | int pos; | |
1322 | ||
1323 | if (option & ADD_CACHE_JUST_APPEND) | |
1324 | pos = istate->cache_nr; | |
1325 | else { | |
1326 | int ret; | |
1327 | ret = add_index_entry_with_check(istate, ce, option); | |
1328 | if (ret <= 0) | |
1329 | return ret; | |
1330 | pos = ret - 1; | |
1331 | } | |
0f1e4f04 | 1332 | |
197ee8c9 | 1333 | /* Make sure the array is big enough .. */ |
999f5660 | 1334 | ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc); |
197ee8c9 LT |
1335 | |
1336 | /* Add it in.. */ | |
4aab5b46 | 1337 | istate->cache_nr++; |
af3785dc | 1338 | if (istate->cache_nr > pos + 1) |
f919ffeb SG |
1339 | MOVE_ARRAY(istate->cache + pos + 1, istate->cache + pos, |
1340 | istate->cache_nr - pos - 1); | |
cf558704 | 1341 | set_index_entry(istate, pos, ce); |
e636a7b4 | 1342 | istate->cache_changed |= CE_ENTRY_ADDED; |
197ee8c9 LT |
1343 | return 0; |
1344 | } | |
1345 | ||
405e5b2f LT |
1346 | /* |
1347 | * "refresh" does not calculate a new sha1 file or bring the | |
1348 | * cache up-to-date for mode/content changes. But what it | |
1349 | * _does_ do is to "re-match" the stat information of a file | |
1350 | * with the cache, so that you can refresh the cache for a | |
1351 | * file that hasn't been changed but where the stat entry is | |
1352 | * out of date. | |
1353 | * | |
1354 | * For example, you'd want to do this after doing a "git-read-tree", | |
1355 | * to link up the stat cache details with the proper files. | |
1356 | */ | |
4aab5b46 | 1357 | static struct cache_entry *refresh_cache_ent(struct index_state *istate, |
4bd5b7da | 1358 | struct cache_entry *ce, |
d05e6970 JK |
1359 | unsigned int options, int *err, |
1360 | int *changed_ret) | |
405e5b2f LT |
1361 | { |
1362 | struct stat st; | |
1363 | struct cache_entry *updated; | |
a849735b | 1364 | int changed; |
25762726 | 1365 | int refresh = options & CE_MATCH_REFRESH; |
4bd5b7da | 1366 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; |
56cac48c | 1367 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; |
2e2e7ec1 | 1368 | int ignore_missing = options & CE_MATCH_IGNORE_MISSING; |
883e248b | 1369 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; |
405e5b2f | 1370 | |
25762726 | 1371 | if (!refresh || ce_uptodate(ce)) |
eadb5831 JH |
1372 | return ce; |
1373 | ||
883e248b BP |
1374 | if (!ignore_fsmonitor) |
1375 | refresh_fsmonitor(istate); | |
aa9349d4 | 1376 | /* |
56cac48c NTND |
1377 | * CE_VALID or CE_SKIP_WORKTREE means the user promised us |
1378 | * that the change to the work tree does not matter and told | |
1379 | * us not to worry. | |
aa9349d4 | 1380 | */ |
56cac48c NTND |
1381 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) { |
1382 | ce_mark_uptodate(ce); | |
1383 | return ce; | |
1384 | } | |
aa9349d4 MSO |
1385 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) { |
1386 | ce_mark_uptodate(ce); | |
1387 | return ce; | |
1388 | } | |
883e248b BP |
1389 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) { |
1390 | ce_mark_uptodate(ce); | |
1391 | return ce; | |
1392 | } | |
aa9349d4 | 1393 | |
ccad42d4 RS |
1394 | if (has_symlink_leading_path(ce->name, ce_namelen(ce))) { |
1395 | if (ignore_missing) | |
1396 | return ce; | |
1397 | if (err) | |
1398 | *err = ENOENT; | |
1399 | return NULL; | |
1400 | } | |
1401 | ||
8fd2cb40 | 1402 | if (lstat(ce->name, &st) < 0) { |
2e2e7ec1 BK |
1403 | if (ignore_missing && errno == ENOENT) |
1404 | return ce; | |
ec0cc704 JH |
1405 | if (err) |
1406 | *err = errno; | |
8fd2cb40 JS |
1407 | return NULL; |
1408 | } | |
405e5b2f | 1409 | |
4bd5b7da | 1410 | changed = ie_match_stat(istate, ce, &st, options); |
d05e6970 JK |
1411 | if (changed_ret) |
1412 | *changed_ret = changed; | |
405e5b2f | 1413 | if (!changed) { |
4bd5b7da JH |
1414 | /* |
1415 | * The path is unchanged. If we were told to ignore | |
1416 | * valid bit, then we did the actual stat check and | |
1417 | * found that the entry is unmodified. If the entry | |
1418 | * is not marked VALID, this is the place to mark it | |
1419 | * valid again, under "assume unchanged" mode. | |
1420 | */ | |
1421 | if (ignore_valid && assume_unchanged && | |
7a51ed66 | 1422 | !(ce->ce_flags & CE_VALID)) |
405e5b2f | 1423 | ; /* mark this one VALID again */ |
eadb5831 JH |
1424 | else { |
1425 | /* | |
1426 | * We do not mark the index itself "modified" | |
1427 | * because CE_UPTODATE flag is in-core only; | |
1428 | * we are not going to write this change out. | |
1429 | */ | |
883e248b | 1430 | if (!S_ISGITLINK(ce->ce_mode)) { |
125fd984 | 1431 | ce_mark_uptodate(ce); |
883e248b BP |
1432 | mark_fsmonitor_valid(ce); |
1433 | } | |
8fd2cb40 | 1434 | return ce; |
eadb5831 | 1435 | } |
405e5b2f LT |
1436 | } |
1437 | ||
4bd5b7da | 1438 | if (ie_modified(istate, ce, &st, options)) { |
ec0cc704 JH |
1439 | if (err) |
1440 | *err = EINVAL; | |
8fd2cb40 JS |
1441 | return NULL; |
1442 | } | |
405e5b2f | 1443 | |
a849735b | 1444 | updated = make_empty_cache_entry(istate, ce_namelen(ce)); |
0e267b7a BP |
1445 | copy_cache_entry(updated, ce); |
1446 | memcpy(updated->name, ce->name, ce->ce_namelen + 1); | |
405e5b2f | 1447 | fill_stat_cache_info(updated, &st); |
4bd5b7da JH |
1448 | /* |
1449 | * If ignore_valid is not set, we should leave CE_VALID bit | |
1450 | * alone. Otherwise, paths marked with --no-assume-unchanged | |
1451 | * (i.e. things to be edited) will reacquire CE_VALID bit | |
1452 | * automatically, which is not really what we want. | |
405e5b2f | 1453 | */ |
4bd5b7da | 1454 | if (!ignore_valid && assume_unchanged && |
7a51ed66 LT |
1455 | !(ce->ce_flags & CE_VALID)) |
1456 | updated->ce_flags &= ~CE_VALID; | |
405e5b2f | 1457 | |
e636a7b4 | 1458 | /* istate->cache_changed is updated in the caller */ |
405e5b2f LT |
1459 | return updated; |
1460 | } | |
1461 | ||
3deffc52 | 1462 | static void show_file(const char * fmt, const char * name, int in_porcelain, |
046613c5 | 1463 | int * first, const char *header_msg) |
3deffc52 MM |
1464 | { |
1465 | if (in_porcelain && *first && header_msg) { | |
1466 | printf("%s\n", header_msg); | |
cd2b8ae9 | 1467 | *first = 0; |
3deffc52 MM |
1468 | } |
1469 | printf(fmt, name); | |
1470 | } | |
1471 | ||
9b2d6149 NTND |
1472 | int refresh_index(struct index_state *istate, unsigned int flags, |
1473 | const struct pathspec *pathspec, | |
046613c5 | 1474 | char *seen, const char *header_msg) |
405e5b2f LT |
1475 | { |
1476 | int i; | |
1477 | int has_errors = 0; | |
1478 | int really = (flags & REFRESH_REALLY) != 0; | |
1479 | int allow_unmerged = (flags & REFRESH_UNMERGED) != 0; | |
1480 | int quiet = (flags & REFRESH_QUIET) != 0; | |
1481 | int not_new = (flags & REFRESH_IGNORE_MISSING) != 0; | |
5fdeacb0 | 1482 | int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0; |
3deffc52 MM |
1483 | int first = 1; |
1484 | int in_porcelain = (flags & REFRESH_IN_PORCELAIN); | |
25762726 BK |
1485 | unsigned int options = (CE_MATCH_REFRESH | |
1486 | (really ? CE_MATCH_IGNORE_VALID : 0) | | |
2e2e7ec1 | 1487 | (not_new ? CE_MATCH_IGNORE_MISSING : 0)); |
4bd4e730 | 1488 | const char *modified_fmt; |
73b7eae6 JK |
1489 | const char *deleted_fmt; |
1490 | const char *typechange_fmt; | |
1491 | const char *added_fmt; | |
4bd4e730 | 1492 | const char *unmerged_fmt; |
ae9af122 NTND |
1493 | struct progress *progress = NULL; |
1494 | ||
1495 | if (flags & REFRESH_PROGRESS && isatty(2)) | |
1496 | progress = start_delayed_progress(_("Refresh index"), | |
1497 | istate->cache_nr); | |
405e5b2f | 1498 | |
c46c406a | 1499 | trace_performance_enter(); |
a71806a7 NTND |
1500 | modified_fmt = in_porcelain ? "M\t%s\n" : "%s: needs update\n"; |
1501 | deleted_fmt = in_porcelain ? "D\t%s\n" : "%s: needs update\n"; | |
1502 | typechange_fmt = in_porcelain ? "T\t%s\n" : "%s: needs update\n"; | |
1503 | added_fmt = in_porcelain ? "A\t%s\n" : "%s: needs update\n"; | |
1504 | unmerged_fmt = in_porcelain ? "U\t%s\n" : "%s: needs merge\n"; | |
99ce720c BP |
1505 | /* |
1506 | * Use the multi-threaded preload_index() to refresh most of the | |
1507 | * cache entries quickly then in the single threaded loop below, | |
1508 | * we only have to do the special cases that are left. | |
1509 | */ | |
1510 | preload_index(istate, pathspec, 0); | |
4aab5b46 | 1511 | for (i = 0; i < istate->cache_nr; i++) { |
285c2e25 | 1512 | struct cache_entry *ce, *new_entry; |
ec0cc704 | 1513 | int cache_errno = 0; |
73b7eae6 | 1514 | int changed = 0; |
3d1f148c | 1515 | int filtered = 0; |
ec0cc704 | 1516 | |
4aab5b46 | 1517 | ce = istate->cache[i]; |
5fdeacb0 JS |
1518 | if (ignore_submodules && S_ISGITLINK(ce->ce_mode)) |
1519 | continue; | |
1520 | ||
d7b665c3 | 1521 | if (pathspec && !ce_path_match(istate, ce, pathspec, seen)) |
3d1f148c JH |
1522 | filtered = 1; |
1523 | ||
405e5b2f | 1524 | if (ce_stage(ce)) { |
4aab5b46 JH |
1525 | while ((i < istate->cache_nr) && |
1526 | ! strcmp(istate->cache[i]->name, ce->name)) | |
405e5b2f LT |
1527 | i++; |
1528 | i--; | |
1529 | if (allow_unmerged) | |
1530 | continue; | |
3d1f148c JH |
1531 | if (!filtered) |
1532 | show_file(unmerged_fmt, ce->name, in_porcelain, | |
1533 | &first, header_msg); | |
405e5b2f LT |
1534 | has_errors = 1; |
1535 | continue; | |
1536 | } | |
1537 | ||
3d1f148c | 1538 | if (filtered) |
d616813d AJ |
1539 | continue; |
1540 | ||
285c2e25 BW |
1541 | new_entry = refresh_cache_ent(istate, ce, options, &cache_errno, &changed); |
1542 | if (new_entry == ce) | |
405e5b2f | 1543 | continue; |
ae9af122 NTND |
1544 | if (progress) |
1545 | display_progress(progress, i); | |
285c2e25 | 1546 | if (!new_entry) { |
73b7eae6 JK |
1547 | const char *fmt; |
1548 | ||
8fd2cb40 | 1549 | if (really && cache_errno == EINVAL) { |
405e5b2f LT |
1550 | /* If we are doing --really-refresh that |
1551 | * means the index is not valid anymore. | |
1552 | */ | |
7a51ed66 | 1553 | ce->ce_flags &= ~CE_VALID; |
078a58e8 | 1554 | ce->ce_flags |= CE_UPDATE_IN_BASE; |
883e248b | 1555 | mark_fsmonitor_invalid(istate, ce); |
e636a7b4 | 1556 | istate->cache_changed |= CE_ENTRY_CHANGED; |
405e5b2f LT |
1557 | } |
1558 | if (quiet) | |
1559 | continue; | |
73b7eae6 JK |
1560 | |
1561 | if (cache_errno == ENOENT) | |
1562 | fmt = deleted_fmt; | |
895ff3b2 | 1563 | else if (ce_intent_to_add(ce)) |
73b7eae6 JK |
1564 | fmt = added_fmt; /* must be before other checks */ |
1565 | else if (changed & TYPE_CHANGED) | |
1566 | fmt = typechange_fmt; | |
1567 | else | |
1568 | fmt = modified_fmt; | |
1569 | show_file(fmt, | |
1570 | ce->name, in_porcelain, &first, header_msg); | |
405e5b2f LT |
1571 | has_errors = 1; |
1572 | continue; | |
1573 | } | |
cf558704 | 1574 | |
285c2e25 | 1575 | replace_index_entry(istate, i, new_entry); |
405e5b2f | 1576 | } |
ae9af122 NTND |
1577 | if (progress) { |
1578 | display_progress(progress, istate->cache_nr); | |
1579 | stop_progress(&progress); | |
1580 | } | |
c46c406a | 1581 | trace_performance_leave("refresh index"); |
405e5b2f LT |
1582 | return has_errors; |
1583 | } | |
1584 | ||
768d7965 JM |
1585 | struct cache_entry *refresh_cache_entry(struct index_state *istate, |
1586 | struct cache_entry *ce, | |
1587 | unsigned int options) | |
ec0cc704 | 1588 | { |
768d7965 | 1589 | return refresh_cache_ent(istate, ce, options, NULL, NULL); |
ec0cc704 JH |
1590 | } |
1591 | ||
db3b313c JH |
1592 | |
1593 | /***************************************************************** | |
1594 | * Index File I/O | |
1595 | *****************************************************************/ | |
1596 | ||
9d227781 JH |
1597 | #define INDEX_FORMAT_DEFAULT 3 |
1598 | ||
136347d7 TG |
1599 | static unsigned int get_index_format_default(void) |
1600 | { | |
1601 | char *envversion = getenv("GIT_INDEX_VERSION"); | |
3c09d684 | 1602 | char *endp; |
b27a5720 | 1603 | int value; |
3c09d684 TG |
1604 | unsigned int version = INDEX_FORMAT_DEFAULT; |
1605 | ||
136347d7 | 1606 | if (!envversion) { |
b27a5720 TA |
1607 | if (!git_config_get_int("index.version", &value)) |
1608 | version = value; | |
3c09d684 TG |
1609 | if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { |
1610 | warning(_("index.version set, but the value is invalid.\n" | |
136347d7 | 1611 | "Using version %i"), INDEX_FORMAT_DEFAULT); |
3c09d684 | 1612 | return INDEX_FORMAT_DEFAULT; |
136347d7 TG |
1613 | } |
1614 | return version; | |
1615 | } | |
3c09d684 TG |
1616 | |
1617 | version = strtoul(envversion, &endp, 10); | |
1618 | if (*endp || | |
1619 | version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { | |
1620 | warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n" | |
1621 | "Using version %i"), INDEX_FORMAT_DEFAULT); | |
1622 | version = INDEX_FORMAT_DEFAULT; | |
1623 | } | |
1624 | return version; | |
136347d7 TG |
1625 | } |
1626 | ||
db3b313c JH |
1627 | /* |
1628 | * dev/ino/uid/gid/size are also just tracked to the low 32 bits | |
1629 | * Again - this is just a (very strong in practice) heuristic that | |
1630 | * the inode hasn't changed. | |
1631 | * | |
1632 | * We save the fields in big-endian order to allow using the | |
1633 | * index file over NFS transparently. | |
1634 | */ | |
1635 | struct ondisk_cache_entry { | |
1636 | struct cache_time ctime; | |
1637 | struct cache_time mtime; | |
7800c1eb TG |
1638 | uint32_t dev; |
1639 | uint32_t ino; | |
1640 | uint32_t mode; | |
1641 | uint32_t uid; | |
1642 | uint32_t gid; | |
1643 | uint32_t size; | |
575fa8a3 | 1644 | /* |
1645 | * unsigned char hash[hashsz]; | |
1646 | * uint16_t flags; | |
1647 | * if (flags & CE_EXTENDED) | |
1648 | * uint16_t flags2; | |
1649 | */ | |
1650 | unsigned char data[GIT_MAX_RAWSZ + 2 * sizeof(uint16_t)]; | |
1651 | char name[FLEX_ARRAY]; | |
db3b313c JH |
1652 | }; |
1653 | ||
6c9cd161 | 1654 | /* These are only used for v3 or lower */ |
ce012deb | 1655 | #define align_padding_size(size, len) ((size + (len) + 8) & ~7) - (size + len) |
575fa8a3 | 1656 | #define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,data) + (len) + 8) & ~7) |
db3b313c | 1657 | #define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len) |
575fa8a3 | 1658 | #define ondisk_data_size(flags, len) (the_hash_algo->rawsz + \ |
1659 | ((flags & CE_EXTENDED) ? 2 : 1) * sizeof(uint16_t) + len) | |
1660 | #define ondisk_data_size_max(len) (ondisk_data_size(CE_EXTENDED, len)) | |
1661 | #define ondisk_ce_size(ce) (ondisk_cache_entry_size(ondisk_data_size((ce)->ce_flags, ce_namelen(ce)))) | |
db3b313c | 1662 | |
a33fc72f JH |
1663 | /* Allow fsck to force verification of the index checksum. */ |
1664 | int verify_index_checksum; | |
1665 | ||
00ec50e5 BP |
1666 | /* Allow fsck to force verification of the cache entry order. */ |
1667 | int verify_ce_order; | |
1668 | ||
371ed0de | 1669 | static int verify_hdr(const struct cache_header *hdr, unsigned long size) |
e83c5163 | 1670 | { |
aab61359 | 1671 | git_hash_ctx c; |
1672 | unsigned char hash[GIT_MAX_RAWSZ]; | |
0136bac9 | 1673 | int hdr_version; |
e83c5163 | 1674 | |
ccc4feb5 | 1675 | if (hdr->hdr_signature != htonl(CACHE_SIGNATURE)) |
9d0a9e90 | 1676 | return error(_("bad signature 0x%08x"), hdr->hdr_signature); |
0136bac9 | 1677 | hdr_version = ntohl(hdr->hdr_version); |
b82a7b5b | 1678 | if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version) |
9d0a9e90 | 1679 | return error(_("bad index version %d"), hdr_version); |
a33fc72f JH |
1680 | |
1681 | if (!verify_index_checksum) | |
1682 | return 0; | |
1683 | ||
aab61359 | 1684 | the_hash_algo->init_fn(&c); |
1685 | the_hash_algo->update_fn(&c, hdr, size - the_hash_algo->rawsz); | |
1686 | the_hash_algo->final_fn(hash, &c); | |
67947c34 | 1687 | if (!hasheq(hash, (unsigned char *)hdr + size - the_hash_algo->rawsz)) |
9d0a9e90 | 1688 | return error(_("bad index file sha1 signature")); |
e83c5163 LT |
1689 | return 0; |
1690 | } | |
1691 | ||
4aab5b46 | 1692 | static int read_index_extension(struct index_state *istate, |
371ed0de | 1693 | const char *ext, const char *data, unsigned long sz) |
bad68ec9 JH |
1694 | { |
1695 | switch (CACHE_EXT(ext)) { | |
1696 | case CACHE_EXT_TREE: | |
4aab5b46 | 1697 | istate->cache_tree = cache_tree_read(data, sz); |
bad68ec9 | 1698 | break; |
cfc5789a JH |
1699 | case CACHE_EXT_RESOLVE_UNDO: |
1700 | istate->resolve_undo = resolve_undo_read(data, sz); | |
1701 | break; | |
5fc2fc8f NTND |
1702 | case CACHE_EXT_LINK: |
1703 | if (read_link_extension(istate, data, sz)) | |
1704 | return -1; | |
1705 | break; | |
f9e6c649 NTND |
1706 | case CACHE_EXT_UNTRACKED: |
1707 | istate->untracked = read_untracked_extension(data, sz); | |
1708 | break; | |
883e248b BP |
1709 | case CACHE_EXT_FSMONITOR: |
1710 | read_fsmonitor_extension(istate, data, sz); | |
1711 | break; | |
3b1d9e04 | 1712 | case CACHE_EXT_ENDOFINDEXENTRIES: |
3255089a | 1713 | case CACHE_EXT_INDEXENTRYOFFSETTABLE: |
3b1d9e04 BP |
1714 | /* already handled in do_read_index() */ |
1715 | break; | |
bad68ec9 JH |
1716 | default: |
1717 | if (*ext < 'A' || 'Z' < *ext) | |
9d0a9e90 | 1718 | return error(_("index uses %.4s extension, which we do not understand"), |
bad68ec9 | 1719 | ext); |
9d0a9e90 | 1720 | fprintf_ln(stderr, _("ignoring %.4s extension"), ext); |
bad68ec9 JH |
1721 | break; |
1722 | } | |
1723 | return 0; | |
1724 | } | |
1725 | ||
77ff1127 BP |
1726 | static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool, |
1727 | unsigned int version, | |
a849735b | 1728 | struct ondisk_cache_entry *ondisk, |
6c9cd161 | 1729 | unsigned long *ent_size, |
252d079c | 1730 | const struct cache_entry *previous_ce) |
7a51ed66 | 1731 | { |
debed2a6 | 1732 | struct cache_entry *ce; |
7fec10b7 | 1733 | size_t len; |
06aaaa0b | 1734 | const char *name; |
575fa8a3 | 1735 | const unsigned hashsz = the_hash_algo->rawsz; |
1736 | const uint16_t *flagsp = (const uint16_t *)(ondisk->data + hashsz); | |
debed2a6 | 1737 | unsigned int flags; |
f5c4a9af | 1738 | size_t copy_len = 0; |
252d079c NTND |
1739 | /* |
1740 | * Adjacent cache entries tend to share the leading paths, so it makes | |
1741 | * sense to only store the differences in later entries. In the v4 | |
1742 | * on-disk format of the index, each on-disk cache entry stores the | |
1743 | * number of bytes to be stripped from the end of the previous name, | |
1744 | * and the bytes to append to the result, to come up with its name. | |
1745 | */ | |
77ff1127 | 1746 | int expand_name_field = version == 4; |
7fec10b7 | 1747 | |
7a51ed66 | 1748 | /* On-disk flags are just 16 bits */ |
575fa8a3 | 1749 | flags = get_be16(flagsp); |
debed2a6 | 1750 | len = flags & CE_NAMEMASK; |
7fec10b7 | 1751 | |
debed2a6 | 1752 | if (flags & CE_EXTENDED) { |
06aaaa0b | 1753 | int extended_flags; |
575fa8a3 | 1754 | extended_flags = get_be16(flagsp + 1) << 16; |
06aaaa0b NTND |
1755 | /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */ |
1756 | if (extended_flags & ~CE_EXTENDED_FLAGS) | |
9d0a9e90 | 1757 | die(_("unknown index entry format 0x%08x"), extended_flags); |
debed2a6 | 1758 | flags |= extended_flags; |
575fa8a3 | 1759 | name = (const char *)(flagsp + 2); |
06aaaa0b NTND |
1760 | } |
1761 | else | |
575fa8a3 | 1762 | name = (const char *)(flagsp + 1); |
06aaaa0b | 1763 | |
252d079c NTND |
1764 | if (expand_name_field) { |
1765 | const unsigned char *cp = (const unsigned char *)name; | |
1766 | size_t strip_len, previous_len; | |
6c9cd161 | 1767 | |
77ff1127 | 1768 | /* If we're at the begining of a block, ignore the previous name */ |
252d079c | 1769 | strip_len = decode_varint(&cp); |
77ff1127 BP |
1770 | if (previous_ce) { |
1771 | previous_len = previous_ce->ce_namelen; | |
1772 | if (previous_len < strip_len) | |
252d079c | 1773 | die(_("malformed name field in the index, near path '%s'"), |
77ff1127 BP |
1774 | previous_ce->name); |
1775 | copy_len = previous_len - strip_len; | |
252d079c | 1776 | } |
252d079c NTND |
1777 | name = (const char *)cp; |
1778 | } | |
6c9cd161 | 1779 | |
252d079c NTND |
1780 | if (len == CE_NAMEMASK) { |
1781 | len = strlen(name); | |
1782 | if (expand_name_field) | |
1783 | len += copy_len; | |
1784 | } | |
6c9cd161 | 1785 | |
77ff1127 | 1786 | ce = mem_pool__ce_alloc(ce_mem_pool, len); |
252d079c NTND |
1787 | |
1788 | ce->ce_stat_data.sd_ctime.sec = get_be32(&ondisk->ctime.sec); | |
1789 | ce->ce_stat_data.sd_mtime.sec = get_be32(&ondisk->mtime.sec); | |
1790 | ce->ce_stat_data.sd_ctime.nsec = get_be32(&ondisk->ctime.nsec); | |
1791 | ce->ce_stat_data.sd_mtime.nsec = get_be32(&ondisk->mtime.nsec); | |
1792 | ce->ce_stat_data.sd_dev = get_be32(&ondisk->dev); | |
1793 | ce->ce_stat_data.sd_ino = get_be32(&ondisk->ino); | |
1794 | ce->ce_mode = get_be32(&ondisk->mode); | |
1795 | ce->ce_stat_data.sd_uid = get_be32(&ondisk->uid); | |
1796 | ce->ce_stat_data.sd_gid = get_be32(&ondisk->gid); | |
1797 | ce->ce_stat_data.sd_size = get_be32(&ondisk->size); | |
1798 | ce->ce_flags = flags & ~CE_NAMEMASK; | |
1799 | ce->ce_namelen = len; | |
1800 | ce->index = 0; | |
575fa8a3 | 1801 | hashcpy(ce->oid.hash, ondisk->data); |
1802 | memcpy(ce->name, name, len); | |
1803 | ce->name[len] = '\0'; | |
6c9cd161 | 1804 | |
252d079c NTND |
1805 | if (expand_name_field) { |
1806 | if (copy_len) | |
1807 | memcpy(ce->name, previous_ce->name, copy_len); | |
1808 | memcpy(ce->name + copy_len, name, len + 1 - copy_len); | |
1809 | *ent_size = (name - ((char *)ondisk)) + len + 1 - copy_len; | |
1810 | } else { | |
1811 | memcpy(ce->name, name, len + 1); | |
1812 | *ent_size = ondisk_ce_size(ce); | |
6c9cd161 | 1813 | } |
debed2a6 | 1814 | return ce; |
cf558704 LT |
1815 | } |
1816 | ||
03f15a79 | 1817 | static void check_ce_order(struct index_state *istate) |
15999d0b | 1818 | { |
03f15a79 TG |
1819 | unsigned int i; |
1820 | ||
00ec50e5 BP |
1821 | if (!verify_ce_order) |
1822 | return; | |
1823 | ||
03f15a79 TG |
1824 | for (i = 1; i < istate->cache_nr; i++) { |
1825 | struct cache_entry *ce = istate->cache[i - 1]; | |
1826 | struct cache_entry *next_ce = istate->cache[i]; | |
1827 | int name_compare = strcmp(ce->name, next_ce->name); | |
1828 | ||
1829 | if (0 < name_compare) | |
9d0a9e90 | 1830 | die(_("unordered stage entries in index")); |
03f15a79 TG |
1831 | if (!name_compare) { |
1832 | if (!ce_stage(ce)) | |
9d0a9e90 | 1833 | die(_("multiple stage entries for merged file '%s'"), |
03f15a79 TG |
1834 | ce->name); |
1835 | if (ce_stage(ce) > ce_stage(next_ce)) | |
9d0a9e90 | 1836 | die(_("unordered stage entries for '%s'"), |
03f15a79 TG |
1837 | ce->name); |
1838 | } | |
15999d0b JSP |
1839 | } |
1840 | } | |
1841 | ||
435ec090 CC |
1842 | static void tweak_untracked_cache(struct index_state *istate) |
1843 | { | |
1844 | switch (git_config_get_untracked_cache()) { | |
1845 | case -1: /* keep: do nothing */ | |
1846 | break; | |
1847 | case 0: /* false */ | |
1848 | remove_untracked_cache(istate); | |
1849 | break; | |
1850 | case 1: /* true */ | |
1851 | add_untracked_cache(istate); | |
1852 | break; | |
1853 | default: /* unknown value: do nothing */ | |
1854 | break; | |
1855 | } | |
1856 | } | |
1857 | ||
43925312 CC |
1858 | static void tweak_split_index(struct index_state *istate) |
1859 | { | |
1860 | switch (git_config_get_split_index()) { | |
1861 | case -1: /* unset: do nothing */ | |
1862 | break; | |
1863 | case 0: /* false */ | |
1864 | remove_split_index(istate); | |
1865 | break; | |
1866 | case 1: /* true */ | |
1867 | add_split_index(istate); | |
1868 | break; | |
1869 | default: /* unknown value: do nothing */ | |
1870 | break; | |
1871 | } | |
1872 | } | |
1873 | ||
435ec090 CC |
1874 | static void post_read_index_from(struct index_state *istate) |
1875 | { | |
1876 | check_ce_order(istate); | |
1877 | tweak_untracked_cache(istate); | |
43925312 | 1878 | tweak_split_index(istate); |
883e248b | 1879 | tweak_fsmonitor(istate); |
435ec090 CC |
1880 | } |
1881 | ||
8e72d675 JM |
1882 | static size_t estimate_cache_size_from_compressed(unsigned int entries) |
1883 | { | |
1884 | return entries * (sizeof(struct cache_entry) + CACHE_ENTRY_PATH_LENGTH); | |
1885 | } | |
1886 | ||
1887 | static size_t estimate_cache_size(size_t ondisk_size, unsigned int entries) | |
1888 | { | |
1889 | long per_entry = sizeof(struct cache_entry) - sizeof(struct ondisk_cache_entry); | |
1890 | ||
1891 | /* | |
1892 | * Account for potential alignment differences. | |
1893 | */ | |
1894 | per_entry += align_padding_size(sizeof(struct cache_entry), -sizeof(struct ondisk_cache_entry)); | |
1895 | return ondisk_size + entries * per_entry; | |
1896 | } | |
1897 | ||
3255089a BP |
1898 | struct index_entry_offset |
1899 | { | |
1900 | /* starting byte offset into index file, count of index entries in this block */ | |
1901 | int offset, nr; | |
1902 | }; | |
1903 | ||
1904 | struct index_entry_offset_table | |
1905 | { | |
1906 | int nr; | |
1907 | struct index_entry_offset entries[FLEX_ARRAY]; | |
1908 | }; | |
1909 | ||
3255089a BP |
1910 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset); |
1911 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot); | |
3255089a | 1912 | |
3b1d9e04 BP |
1913 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size); |
1914 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset); | |
1915 | ||
abb4bb83 BP |
1916 | struct load_index_extensions |
1917 | { | |
abb4bb83 | 1918 | pthread_t pthread; |
abb4bb83 BP |
1919 | struct index_state *istate; |
1920 | const char *mmap; | |
1921 | size_t mmap_size; | |
1922 | unsigned long src_offset; | |
1923 | }; | |
1924 | ||
1925 | static void *load_index_extensions(void *_data) | |
1926 | { | |
1927 | struct load_index_extensions *p = _data; | |
1928 | unsigned long src_offset = p->src_offset; | |
1929 | ||
1930 | while (src_offset <= p->mmap_size - the_hash_algo->rawsz - 8) { | |
1931 | /* After an array of active_nr index entries, | |
1932 | * there can be arbitrary number of extended | |
1933 | * sections, each of which is prefixed with | |
1934 | * extension name (4-byte) and section length | |
1935 | * in 4-byte network byte order. | |
1936 | */ | |
1937 | uint32_t extsize = get_be32(p->mmap + src_offset + 4); | |
1938 | if (read_index_extension(p->istate, | |
1939 | p->mmap + src_offset, | |
1940 | p->mmap + src_offset + 8, | |
1941 | extsize) < 0) { | |
1942 | munmap((void *)p->mmap, p->mmap_size); | |
1943 | die(_("index file corrupt")); | |
1944 | } | |
1945 | src_offset += 8; | |
1946 | src_offset += extsize; | |
1947 | } | |
1948 | ||
1949 | return NULL; | |
1950 | } | |
1951 | ||
77ff1127 BP |
1952 | /* |
1953 | * A helper function that will load the specified range of cache entries | |
1954 | * from the memory mapped file and add them to the given index. | |
1955 | */ | |
1956 | static unsigned long load_cache_entry_block(struct index_state *istate, | |
1957 | struct mem_pool *ce_mem_pool, int offset, int nr, const char *mmap, | |
1958 | unsigned long start_offset, const struct cache_entry *previous_ce) | |
1959 | { | |
1960 | int i; | |
1961 | unsigned long src_offset = start_offset; | |
1962 | ||
1963 | for (i = offset; i < offset + nr; i++) { | |
1964 | struct ondisk_cache_entry *disk_ce; | |
1965 | struct cache_entry *ce; | |
1966 | unsigned long consumed; | |
1967 | ||
1968 | disk_ce = (struct ondisk_cache_entry *)(mmap + src_offset); | |
1969 | ce = create_from_disk(ce_mem_pool, istate->version, disk_ce, &consumed, previous_ce); | |
1970 | set_index_entry(istate, i, ce); | |
1971 | ||
1972 | src_offset += consumed; | |
1973 | previous_ce = ce; | |
1974 | } | |
1975 | return src_offset - start_offset; | |
1976 | } | |
1977 | ||
1978 | static unsigned long load_all_cache_entries(struct index_state *istate, | |
1979 | const char *mmap, size_t mmap_size, unsigned long src_offset) | |
1980 | { | |
1981 | unsigned long consumed; | |
1982 | ||
1983 | if (istate->version == 4) { | |
1984 | mem_pool_init(&istate->ce_mem_pool, | |
1985 | estimate_cache_size_from_compressed(istate->cache_nr)); | |
1986 | } else { | |
1987 | mem_pool_init(&istate->ce_mem_pool, | |
1988 | estimate_cache_size(mmap_size, istate->cache_nr)); | |
1989 | } | |
1990 | ||
1991 | consumed = load_cache_entry_block(istate, istate->ce_mem_pool, | |
1992 | 0, istate->cache_nr, mmap, src_offset, NULL); | |
1993 | return consumed; | |
1994 | } | |
1995 | ||
3255089a BP |
1996 | /* |
1997 | * Mostly randomly chosen maximum thread counts: we | |
1998 | * cap the parallelism to online_cpus() threads, and we want | |
1999 | * to have at least 10000 cache entries per thread for it to | |
2000 | * be worth starting a thread. | |
2001 | */ | |
2002 | ||
2003 | #define THREAD_COST (10000) | |
2004 | ||
77ff1127 BP |
2005 | struct load_cache_entries_thread_data |
2006 | { | |
2007 | pthread_t pthread; | |
2008 | struct index_state *istate; | |
2009 | struct mem_pool *ce_mem_pool; | |
2010 | int offset; | |
2011 | const char *mmap; | |
2012 | struct index_entry_offset_table *ieot; | |
2013 | int ieot_start; /* starting index into the ieot array */ | |
2014 | int ieot_blocks; /* count of ieot entries to process */ | |
2015 | unsigned long consumed; /* return # of bytes in index file processed */ | |
2016 | }; | |
2017 | ||
2018 | /* | |
2019 | * A thread proc to run the load_cache_entries() computation | |
2020 | * across multiple background threads. | |
2021 | */ | |
2022 | static void *load_cache_entries_thread(void *_data) | |
2023 | { | |
2024 | struct load_cache_entries_thread_data *p = _data; | |
2025 | int i; | |
2026 | ||
2027 | /* iterate across all ieot blocks assigned to this thread */ | |
2028 | for (i = p->ieot_start; i < p->ieot_start + p->ieot_blocks; i++) { | |
2029 | p->consumed += load_cache_entry_block(p->istate, p->ce_mem_pool, | |
2030 | p->offset, p->ieot->entries[i].nr, p->mmap, p->ieot->entries[i].offset, NULL); | |
2031 | p->offset += p->ieot->entries[i].nr; | |
2032 | } | |
2033 | return NULL; | |
2034 | } | |
2035 | ||
2036 | static unsigned long load_cache_entries_threaded(struct index_state *istate, const char *mmap, size_t mmap_size, | |
2037 | unsigned long src_offset, int nr_threads, struct index_entry_offset_table *ieot) | |
2038 | { | |
2039 | int i, offset, ieot_blocks, ieot_start, err; | |
2040 | struct load_cache_entries_thread_data *data; | |
2041 | unsigned long consumed = 0; | |
2042 | ||
2043 | /* a little sanity checking */ | |
2044 | if (istate->name_hash_initialized) | |
2045 | BUG("the name hash isn't thread safe"); | |
2046 | ||
2047 | mem_pool_init(&istate->ce_mem_pool, 0); | |
2048 | ||
2049 | /* ensure we have no more threads than we have blocks to process */ | |
2050 | if (nr_threads > ieot->nr) | |
2051 | nr_threads = ieot->nr; | |
2052 | data = xcalloc(nr_threads, sizeof(*data)); | |
2053 | ||
2054 | offset = ieot_start = 0; | |
2055 | ieot_blocks = DIV_ROUND_UP(ieot->nr, nr_threads); | |
2056 | for (i = 0; i < nr_threads; i++) { | |
2057 | struct load_cache_entries_thread_data *p = &data[i]; | |
2058 | int nr, j; | |
2059 | ||
2060 | if (ieot_start + ieot_blocks > ieot->nr) | |
2061 | ieot_blocks = ieot->nr - ieot_start; | |
2062 | ||
2063 | p->istate = istate; | |
2064 | p->offset = offset; | |
2065 | p->mmap = mmap; | |
2066 | p->ieot = ieot; | |
2067 | p->ieot_start = ieot_start; | |
2068 | p->ieot_blocks = ieot_blocks; | |
2069 | ||
2070 | /* create a mem_pool for each thread */ | |
2071 | nr = 0; | |
2072 | for (j = p->ieot_start; j < p->ieot_start + p->ieot_blocks; j++) | |
2073 | nr += p->ieot->entries[j].nr; | |
2074 | if (istate->version == 4) { | |
2075 | mem_pool_init(&p->ce_mem_pool, | |
2076 | estimate_cache_size_from_compressed(nr)); | |
2077 | } else { | |
2078 | mem_pool_init(&p->ce_mem_pool, | |
2079 | estimate_cache_size(mmap_size, nr)); | |
2080 | } | |
2081 | ||
2082 | err = pthread_create(&p->pthread, NULL, load_cache_entries_thread, p); | |
2083 | if (err) | |
2084 | die(_("unable to create load_cache_entries thread: %s"), strerror(err)); | |
2085 | ||
2086 | /* increment by the number of cache entries in the ieot block being processed */ | |
2087 | for (j = 0; j < ieot_blocks; j++) | |
2088 | offset += ieot->entries[ieot_start + j].nr; | |
2089 | ieot_start += ieot_blocks; | |
2090 | } | |
2091 | ||
2092 | for (i = 0; i < nr_threads; i++) { | |
2093 | struct load_cache_entries_thread_data *p = &data[i]; | |
2094 | ||
2095 | err = pthread_join(p->pthread, NULL); | |
2096 | if (err) | |
2097 | die(_("unable to join load_cache_entries thread: %s"), strerror(err)); | |
2098 | mem_pool_combine(istate->ce_mem_pool, p->ce_mem_pool); | |
2099 | consumed += p->consumed; | |
2100 | } | |
2101 | ||
2102 | free(data); | |
2103 | ||
2104 | return consumed; | |
2105 | } | |
77ff1127 | 2106 | |
8fd2cb40 | 2107 | /* remember to discard_cache() before reading a different cache! */ |
3e52f70b | 2108 | int do_read_index(struct index_state *istate, const char *path, int must_exist) |
e83c5163 | 2109 | { |
77ff1127 | 2110 | int fd; |
e83c5163 | 2111 | struct stat st; |
debed2a6 | 2112 | unsigned long src_offset; |
371ed0de BP |
2113 | const struct cache_header *hdr; |
2114 | const char *mmap; | |
7a51ed66 | 2115 | size_t mmap_size; |
abb4bb83 BP |
2116 | struct load_index_extensions p; |
2117 | size_t extension_offset = 0; | |
77ff1127 BP |
2118 | int nr_threads, cpus; |
2119 | struct index_entry_offset_table *ieot = NULL; | |
e83c5163 | 2120 | |
913e0e99 | 2121 | if (istate->initialized) |
4aab5b46 | 2122 | return istate->cache_nr; |
5d1a5c02 | 2123 | |
fba2f38a KB |
2124 | istate->timestamp.sec = 0; |
2125 | istate->timestamp.nsec = 0; | |
8fd2cb40 | 2126 | fd = open(path, O_RDONLY); |
5d1a5c02 | 2127 | if (fd < 0) { |
5fc2fc8f | 2128 | if (!must_exist && errno == ENOENT) |
5d1a5c02 | 2129 | return 0; |
9d0a9e90 | 2130 | die_errno(_("%s: index file open failed"), path); |
5d1a5c02 | 2131 | } |
e83c5163 | 2132 | |
3511a377 | 2133 | if (fstat(fd, &st)) |
9d0a9e90 | 2134 | die_errno(_("%s: cannot stat the open index"), path); |
3511a377 | 2135 | |
7a51ed66 | 2136 | mmap_size = xsize_t(st.st_size); |
aab61359 | 2137 | if (mmap_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
9d0a9e90 | 2138 | die(_("%s: index file smaller than expected"), path); |
3511a377 | 2139 | |
a1293ef7 | 2140 | mmap = xmmap(NULL, mmap_size, PROT_READ, MAP_PRIVATE, fd, 0); |
7a51ed66 | 2141 | if (mmap == MAP_FAILED) |
9d0a9e90 | 2142 | die_errno(_("%s: unable to map index file"), path); |
57d84f8d | 2143 | close(fd); |
e83c5163 | 2144 | |
371ed0de | 2145 | hdr = (const struct cache_header *)mmap; |
7a51ed66 | 2146 | if (verify_hdr(hdr, mmap_size) < 0) |
e83c5163 LT |
2147 | goto unmap; |
2148 | ||
75691ea3 | 2149 | hashcpy(istate->oid.hash, (const unsigned char *)hdr + mmap_size - the_hash_algo->rawsz); |
9d227781 | 2150 | istate->version = ntohl(hdr->hdr_version); |
4aab5b46 JH |
2151 | istate->cache_nr = ntohl(hdr->hdr_entries); |
2152 | istate->cache_alloc = alloc_nr(istate->cache_nr); | |
c4aa3167 | 2153 | istate->cache = xcalloc(istate->cache_alloc, sizeof(*istate->cache)); |
913e0e99 | 2154 | istate->initialized = 1; |
7a51ed66 | 2155 | |
abb4bb83 BP |
2156 | p.istate = istate; |
2157 | p.mmap = mmap; | |
2158 | p.mmap_size = mmap_size; | |
6c9cd161 | 2159 | |
7a51ed66 | 2160 | src_offset = sizeof(*hdr); |
4aab5b46 | 2161 | |
2a9dedef JN |
2162 | if (git_config_get_index_threads(&nr_threads)) |
2163 | nr_threads = 1; | |
7a51ed66 | 2164 | |
77ff1127 BP |
2165 | /* TODO: does creating more threads than cores help? */ |
2166 | if (!nr_threads) { | |
2167 | nr_threads = istate->cache_nr / THREAD_COST; | |
2168 | cpus = online_cpus(); | |
2169 | if (nr_threads > cpus) | |
2170 | nr_threads = cpus; | |
e83c5163 | 2171 | } |
abb4bb83 | 2172 | |
88168b9b NTND |
2173 | if (!HAVE_THREADS) |
2174 | nr_threads = 1; | |
2175 | ||
abb4bb83 BP |
2176 | if (nr_threads > 1) { |
2177 | extension_offset = read_eoie_extension(mmap, mmap_size); | |
2178 | if (extension_offset) { | |
2179 | int err; | |
2180 | ||
2181 | p.src_offset = extension_offset; | |
2182 | err = pthread_create(&p.pthread, NULL, load_index_extensions, &p); | |
2183 | if (err) | |
2184 | die(_("unable to create load_index_extensions thread: %s"), strerror(err)); | |
2185 | ||
2186 | nr_threads--; | |
2187 | } | |
2188 | } | |
abb4bb83 | 2189 | |
77ff1127 BP |
2190 | /* |
2191 | * Locate and read the index entry offset table so that we can use it | |
2192 | * to multi-thread the reading of the cache entries. | |
2193 | */ | |
2194 | if (extension_offset && nr_threads > 1) | |
2195 | ieot = read_ieot_extension(mmap, mmap_size, extension_offset); | |
2196 | ||
2197 | if (ieot) { | |
2198 | src_offset += load_cache_entries_threaded(istate, mmap, mmap_size, src_offset, nr_threads, ieot); | |
2199 | free(ieot); | |
8e72d675 | 2200 | } else { |
77ff1127 | 2201 | src_offset += load_all_cache_entries(istate, mmap, mmap_size, src_offset); |
8e72d675 | 2202 | } |
6c9cd161 | 2203 | |
fba2f38a | 2204 | istate->timestamp.sec = st.st_mtime; |
c06ff490 | 2205 | istate->timestamp.nsec = ST_MTIME_NSEC(st); |
fba2f38a | 2206 | |
abb4bb83 | 2207 | /* if we created a thread, join it otherwise load the extensions on the primary thread */ |
abb4bb83 BP |
2208 | if (extension_offset) { |
2209 | int ret = pthread_join(p.pthread, NULL); | |
2210 | if (ret) | |
2211 | die(_("unable to join load_index_extensions thread: %s"), strerror(ret)); | |
88168b9b | 2212 | } else { |
abb4bb83 BP |
2213 | p.src_offset = src_offset; |
2214 | load_index_extensions(&p); | |
bad68ec9 | 2215 | } |
371ed0de | 2216 | munmap((void *)mmap, mmap_size); |
42fee7a3 JH |
2217 | |
2218 | /* | |
2219 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2220 | * that is associated with the given "istate". | |
2221 | */ | |
2222 | trace2_data_intmax("index", the_repository, "read/version", | |
2223 | istate->version); | |
2224 | trace2_data_intmax("index", the_repository, "read/cache_nr", | |
2225 | istate->cache_nr); | |
2226 | ||
4aab5b46 | 2227 | return istate->cache_nr; |
e83c5163 LT |
2228 | |
2229 | unmap: | |
371ed0de | 2230 | munmap((void *)mmap, mmap_size); |
9d0a9e90 | 2231 | die(_("index file corrupt")); |
e83c5163 LT |
2232 | } |
2233 | ||
0d59ffb4 CC |
2234 | /* |
2235 | * Signal that the shared index is used by updating its mtime. | |
2236 | * | |
2237 | * This way, shared index can be removed if they have not been used | |
2238 | * for some time. | |
2239 | */ | |
a125a223 | 2240 | static void freshen_shared_index(const char *shared_index, int warn) |
0d59ffb4 | 2241 | { |
0d59ffb4 | 2242 | if (!check_and_freshen_file(shared_index, 1) && warn) |
9d0a9e90 | 2243 | warning(_("could not freshen shared index '%s'"), shared_index); |
0d59ffb4 CC |
2244 | } |
2245 | ||
a125a223 TG |
2246 | int read_index_from(struct index_state *istate, const char *path, |
2247 | const char *gitdir) | |
5fc2fc8f NTND |
2248 | { |
2249 | struct split_index *split_index; | |
2250 | int ret; | |
2182abd9 | 2251 | char *base_oid_hex; |
a125a223 | 2252 | char *base_path; |
5fc2fc8f NTND |
2253 | |
2254 | /* istate->initialized covers both .git/index and .git/sharedindex.xxx */ | |
2255 | if (istate->initialized) | |
2256 | return istate->cache_nr; | |
2257 | ||
42fee7a3 JH |
2258 | /* |
2259 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2260 | * that is associated with the given "istate". | |
2261 | */ | |
2262 | trace2_region_enter_printf("index", "do_read_index", the_repository, | |
2263 | "%s", path); | |
c46c406a | 2264 | trace_performance_enter(); |
5fc2fc8f | 2265 | ret = do_read_index(istate, path, 0); |
c46c406a | 2266 | trace_performance_leave("read cache %s", path); |
42fee7a3 JH |
2267 | trace2_region_leave_printf("index", "do_read_index", the_repository, |
2268 | "%s", path); | |
435ec090 | 2269 | |
5fc2fc8f | 2270 | split_index = istate->split_index; |
2182abd9 | 2271 | if (!split_index || is_null_oid(&split_index->base_oid)) { |
435ec090 | 2272 | post_read_index_from(istate); |
5fc2fc8f | 2273 | return ret; |
03f15a79 | 2274 | } |
5fc2fc8f | 2275 | |
c46c406a | 2276 | trace_performance_enter(); |
5fc2fc8f NTND |
2277 | if (split_index->base) |
2278 | discard_index(split_index->base); | |
2279 | else | |
2280 | split_index->base = xcalloc(1, sizeof(*split_index->base)); | |
de6ae5f9 | 2281 | |
2182abd9 | 2282 | base_oid_hex = oid_to_hex(&split_index->base_oid); |
2283 | base_path = xstrfmt("%s/sharedindex.%s", gitdir, base_oid_hex); | |
42fee7a3 JH |
2284 | trace2_region_enter_printf("index", "shared/do_read_index", |
2285 | the_repository, "%s", base_path); | |
de6ae5f9 | 2286 | ret = do_read_index(split_index->base, base_path, 1); |
42fee7a3 JH |
2287 | trace2_region_leave_printf("index", "shared/do_read_index", |
2288 | the_repository, "%s", base_path); | |
9001dc2a | 2289 | if (!oideq(&split_index->base_oid, &split_index->base->oid)) |
9d0a9e90 | 2290 | die(_("broken index, expect %s in %s, got %s"), |
2182abd9 | 2291 | base_oid_hex, base_path, |
75691ea3 | 2292 | oid_to_hex(&split_index->base->oid)); |
de6ae5f9 | 2293 | |
a125a223 | 2294 | freshen_shared_index(base_path, 0); |
5fc2fc8f | 2295 | merge_base_index(istate); |
435ec090 | 2296 | post_read_index_from(istate); |
c46c406a | 2297 | trace_performance_leave("read cache %s", base_path); |
b42ad7d5 | 2298 | free(base_path); |
5fc2fc8f NTND |
2299 | return ret; |
2300 | } | |
2301 | ||
fa7b3c2f JH |
2302 | int is_index_unborn(struct index_state *istate) |
2303 | { | |
debed2a6 | 2304 | return (!istate->cache_nr && !istate->timestamp.sec); |
fa7b3c2f JH |
2305 | } |
2306 | ||
4aab5b46 | 2307 | int discard_index(struct index_state *istate) |
6d297f81 | 2308 | { |
8e72d675 JM |
2309 | /* |
2310 | * Cache entries in istate->cache[] should have been allocated | |
2311 | * from the memory pool associated with this index, or from an | |
2312 | * associated split_index. There is no need to free individual | |
8616a2d0 JM |
2313 | * cache entries. validate_cache_entries can detect when this |
2314 | * assertion does not hold. | |
8e72d675 | 2315 | */ |
8616a2d0 | 2316 | validate_cache_entries(istate); |
debed2a6 | 2317 | |
cfc5789a | 2318 | resolve_undo_clear_index(istate); |
4aab5b46 JH |
2319 | istate->cache_nr = 0; |
2320 | istate->cache_changed = 0; | |
fba2f38a KB |
2321 | istate->timestamp.sec = 0; |
2322 | istate->timestamp.nsec = 0; | |
2092678c | 2323 | free_name_hash(istate); |
4aab5b46 | 2324 | cache_tree_free(&(istate->cache_tree)); |
913e0e99 | 2325 | istate->initialized = 0; |
6a83d902 | 2326 | FREE_AND_NULL(istate->cache); |
a0fc4db0 | 2327 | istate->cache_alloc = 0; |
5fc2fc8f | 2328 | discard_split_index(istate); |
f9e6c649 NTND |
2329 | free_untracked_cache(istate->untracked); |
2330 | istate->untracked = NULL; | |
8e72d675 JM |
2331 | |
2332 | if (istate->ce_mem_pool) { | |
8616a2d0 | 2333 | mem_pool_discard(istate->ce_mem_pool, should_validate_cache_entries()); |
8e72d675 JM |
2334 | istate->ce_mem_pool = NULL; |
2335 | } | |
2336 | ||
7a51ed66 | 2337 | return 0; |
6d297f81 JS |
2338 | } |
2339 | ||
8616a2d0 JM |
2340 | /* |
2341 | * Validate the cache entries of this index. | |
2342 | * All cache entries associated with this index | |
2343 | * should have been allocated by the memory pool | |
2344 | * associated with this index, or by a referenced | |
2345 | * split index. | |
2346 | */ | |
2347 | void validate_cache_entries(const struct index_state *istate) | |
2348 | { | |
2349 | int i; | |
2350 | ||
2351 | if (!should_validate_cache_entries() ||!istate || !istate->initialized) | |
2352 | return; | |
2353 | ||
2354 | for (i = 0; i < istate->cache_nr; i++) { | |
2355 | if (!istate) { | |
391408e5 | 2356 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2357 | } else if (!istate->ce_mem_pool || |
2358 | !mem_pool_contains(istate->ce_mem_pool, istate->cache[i])) { | |
2359 | if (!istate->split_index || | |
2360 | !istate->split_index->base || | |
2361 | !istate->split_index->base->ce_mem_pool || | |
2362 | !mem_pool_contains(istate->split_index->base->ce_mem_pool, istate->cache[i])) { | |
391408e5 | 2363 | BUG("cache entry is not allocated from expected memory pool"); |
8616a2d0 JM |
2364 | } |
2365 | } | |
2366 | } | |
2367 | ||
2368 | if (istate->split_index) | |
2369 | validate_cache_entries(istate->split_index->base); | |
2370 | } | |
2371 | ||
d1f128b0 | 2372 | int unmerged_index(const struct index_state *istate) |
94a5728c DB |
2373 | { |
2374 | int i; | |
2375 | for (i = 0; i < istate->cache_nr; i++) { | |
2376 | if (ce_stage(istate->cache[i])) | |
2377 | return 1; | |
2378 | } | |
2379 | return 0; | |
2380 | } | |
2381 | ||
150fe065 NTND |
2382 | int repo_index_has_changes(struct repository *repo, |
2383 | struct tree *tree, | |
2384 | struct strbuf *sb) | |
cffbfad5 | 2385 | { |
150fe065 | 2386 | struct index_state *istate = repo->index; |
e1f8694f | 2387 | struct object_id cmp; |
cffbfad5 EN |
2388 | int i; |
2389 | ||
e1f8694f EN |
2390 | if (tree) |
2391 | cmp = tree->object.oid; | |
2392 | if (tree || !get_oid_tree("HEAD", &cmp)) { | |
cffbfad5 EN |
2393 | struct diff_options opt; |
2394 | ||
150fe065 | 2395 | repo_diff_setup(repo, &opt); |
cffbfad5 EN |
2396 | opt.flags.exit_with_status = 1; |
2397 | if (!sb) | |
2398 | opt.flags.quick = 1; | |
e1f8694f | 2399 | do_diff_cache(&cmp, &opt); |
cffbfad5 EN |
2400 | diffcore_std(&opt); |
2401 | for (i = 0; sb && i < diff_queued_diff.nr; i++) { | |
2402 | if (i) | |
2403 | strbuf_addch(sb, ' '); | |
2404 | strbuf_addstr(sb, diff_queued_diff.queue[i]->two->path); | |
2405 | } | |
2406 | diff_flush(&opt); | |
2407 | return opt.flags.has_changes != 0; | |
2408 | } else { | |
1b9fbefb | 2409 | for (i = 0; sb && i < istate->cache_nr; i++) { |
cffbfad5 EN |
2410 | if (i) |
2411 | strbuf_addch(sb, ' '); | |
1b9fbefb | 2412 | strbuf_addstr(sb, istate->cache[i]->name); |
cffbfad5 | 2413 | } |
1b9fbefb | 2414 | return !!istate->cache_nr; |
cffbfad5 EN |
2415 | } |
2416 | } | |
2417 | ||
4990aadc | 2418 | #define WRITE_BUFFER_SIZE 8192 |
bf0f910d | 2419 | static unsigned char write_buffer[WRITE_BUFFER_SIZE]; |
4990aadc LT |
2420 | static unsigned long write_buffer_len; |
2421 | ||
aab61359 | 2422 | static int ce_write_flush(git_hash_ctx *context, int fd) |
6015c28b JH |
2423 | { |
2424 | unsigned int buffered = write_buffer_len; | |
2425 | if (buffered) { | |
aab61359 | 2426 | the_hash_algo->update_fn(context, write_buffer, buffered); |
06f46f23 | 2427 | if (write_in_full(fd, write_buffer, buffered) < 0) |
6015c28b JH |
2428 | return -1; |
2429 | write_buffer_len = 0; | |
2430 | } | |
2431 | return 0; | |
2432 | } | |
2433 | ||
aab61359 | 2434 | static int ce_write(git_hash_ctx *context, int fd, void *data, unsigned int len) |
4990aadc LT |
2435 | { |
2436 | while (len) { | |
2437 | unsigned int buffered = write_buffer_len; | |
2438 | unsigned int partial = WRITE_BUFFER_SIZE - buffered; | |
2439 | if (partial > len) | |
2440 | partial = len; | |
2441 | memcpy(write_buffer + buffered, data, partial); | |
2442 | buffered += partial; | |
2443 | if (buffered == WRITE_BUFFER_SIZE) { | |
6015c28b JH |
2444 | write_buffer_len = buffered; |
2445 | if (ce_write_flush(context, fd)) | |
4990aadc LT |
2446 | return -1; |
2447 | buffered = 0; | |
2448 | } | |
2449 | write_buffer_len = buffered; | |
2450 | len -= partial; | |
1d7f171c | 2451 | data = (char *) data + partial; |
a6080a0a JH |
2452 | } |
2453 | return 0; | |
4990aadc LT |
2454 | } |
2455 | ||
3b1d9e04 BP |
2456 | static int write_index_ext_header(git_hash_ctx *context, git_hash_ctx *eoie_context, |
2457 | int fd, unsigned int ext, unsigned int sz) | |
bad68ec9 JH |
2458 | { |
2459 | ext = htonl(ext); | |
2460 | sz = htonl(sz); | |
3b1d9e04 BP |
2461 | if (eoie_context) { |
2462 | the_hash_algo->update_fn(eoie_context, &ext, 4); | |
2463 | the_hash_algo->update_fn(eoie_context, &sz, 4); | |
2464 | } | |
968a1d65 DR |
2465 | return ((ce_write(context, fd, &ext, 4) < 0) || |
2466 | (ce_write(context, fd, &sz, 4) < 0)) ? -1 : 0; | |
bad68ec9 JH |
2467 | } |
2468 | ||
aab61359 | 2469 | static int ce_flush(git_hash_ctx *context, int fd, unsigned char *hash) |
4990aadc LT |
2470 | { |
2471 | unsigned int left = write_buffer_len; | |
ca9be054 | 2472 | |
4990aadc LT |
2473 | if (left) { |
2474 | write_buffer_len = 0; | |
aab61359 | 2475 | the_hash_algo->update_fn(context, write_buffer, left); |
4990aadc | 2476 | } |
ca9be054 | 2477 | |
aab61359 | 2478 | /* Flush first if not enough space for hash signature */ |
2479 | if (left + the_hash_algo->rawsz > WRITE_BUFFER_SIZE) { | |
06f46f23 | 2480 | if (write_in_full(fd, write_buffer, left) < 0) |
2c865d9a QH |
2481 | return -1; |
2482 | left = 0; | |
2483 | } | |
2484 | ||
aab61359 | 2485 | /* Append the hash signature at the end */ |
2486 | the_hash_algo->final_fn(write_buffer + left, context); | |
2487 | hashcpy(hash, write_buffer + left); | |
2488 | left += the_hash_algo->rawsz; | |
06f46f23 | 2489 | return (write_in_full(fd, write_buffer, left) < 0) ? -1 : 0; |
4990aadc LT |
2490 | } |
2491 | ||
58bf2a4c NTND |
2492 | static void ce_smudge_racily_clean_entry(struct index_state *istate, |
2493 | struct cache_entry *ce) | |
407c8eb0 JH |
2494 | { |
2495 | /* | |
2496 | * The only thing we care about in this function is to smudge the | |
2497 | * falsely clean entry due to touch-update-touch race, so we leave | |
2498 | * everything else as they are. We are called for entries whose | |
c21d39d7 | 2499 | * ce_stat_data.sd_mtime match the index file mtime. |
c70115b4 JH |
2500 | * |
2501 | * Note that this actually does not do much for gitlinks, for | |
2502 | * which ce_match_stat_basic() always goes to the actual | |
2503 | * contents. The caller checks with is_racy_timestamp() which | |
2504 | * always says "no" for gitlinks, so we are not called for them ;-) | |
407c8eb0 JH |
2505 | */ |
2506 | struct stat st; | |
2507 | ||
2508 | if (lstat(ce->name, &st) < 0) | |
2509 | return; | |
2510 | if (ce_match_stat_basic(ce, &st)) | |
2511 | return; | |
58bf2a4c | 2512 | if (ce_modified_check_fs(istate, ce, &st)) { |
4b3511b0 JH |
2513 | /* This is "racily clean"; smudge it. Note that this |
2514 | * is a tricky code. At first glance, it may appear | |
2515 | * that it can break with this sequence: | |
2516 | * | |
2517 | * $ echo xyzzy >frotz | |
2518 | * $ git-update-index --add frotz | |
2519 | * $ : >frotz | |
2520 | * $ sleep 3 | |
2521 | * $ echo filfre >nitfol | |
2522 | * $ git-update-index --add nitfol | |
2523 | * | |
b7e58b17 | 2524 | * but it does not. When the second update-index runs, |
4b3511b0 JH |
2525 | * it notices that the entry "frotz" has the same timestamp |
2526 | * as index, and if we were to smudge it by resetting its | |
2527 | * size to zero here, then the object name recorded | |
2528 | * in index is the 6-byte file but the cached stat information | |
2529 | * becomes zero --- which would then match what we would | |
a6080a0a | 2530 | * obtain from the filesystem next time we stat("frotz"). |
4b3511b0 JH |
2531 | * |
2532 | * However, the second update-index, before calling | |
2533 | * this function, notices that the cached size is 6 | |
2534 | * bytes and what is on the filesystem is an empty | |
2535 | * file, and never calls us, so the cached size information | |
2536 | * for "frotz" stays 6 which does not match the filesystem. | |
2537 | */ | |
c21d39d7 | 2538 | ce->ce_stat_data.sd_size = 0; |
407c8eb0 JH |
2539 | } |
2540 | } | |
2541 | ||
f136f7bf | 2542 | /* Copy miscellaneous fields but not the name */ |
ce012deb | 2543 | static void copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk, |
f136f7bf | 2544 | struct cache_entry *ce) |
7a51ed66 | 2545 | { |
b60e188c | 2546 | short flags; |
575fa8a3 | 2547 | const unsigned hashsz = the_hash_algo->rawsz; |
2548 | uint16_t *flagsp = (uint16_t *)(ondisk->data + hashsz); | |
b60e188c | 2549 | |
c21d39d7 MH |
2550 | ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec); |
2551 | ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec); | |
2552 | ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec); | |
2553 | ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec); | |
2554 | ondisk->dev = htonl(ce->ce_stat_data.sd_dev); | |
2555 | ondisk->ino = htonl(ce->ce_stat_data.sd_ino); | |
7a51ed66 | 2556 | ondisk->mode = htonl(ce->ce_mode); |
c21d39d7 MH |
2557 | ondisk->uid = htonl(ce->ce_stat_data.sd_uid); |
2558 | ondisk->gid = htonl(ce->ce_stat_data.sd_gid); | |
2559 | ondisk->size = htonl(ce->ce_stat_data.sd_size); | |
575fa8a3 | 2560 | hashcpy(ondisk->data, ce->oid.hash); |
b60e188c | 2561 | |
ce51bf09 | 2562 | flags = ce->ce_flags & ~CE_NAMEMASK; |
b60e188c | 2563 | flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce)); |
575fa8a3 | 2564 | flagsp[0] = htons(flags); |
06aaaa0b | 2565 | if (ce->ce_flags & CE_EXTENDED) { |
575fa8a3 | 2566 | flagsp[1] = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16); |
f136f7bf JH |
2567 | } |
2568 | } | |
2569 | ||
aab61359 | 2570 | static int ce_write_entry(git_hash_ctx *c, int fd, struct cache_entry *ce, |
ce012deb | 2571 | struct strbuf *previous_name, struct ondisk_cache_entry *ondisk) |
f136f7bf | 2572 | { |
9d227781 | 2573 | int size; |
f136f7bf | 2574 | int result; |
00a4b035 RJ |
2575 | unsigned int saved_namelen; |
2576 | int stripped_name = 0; | |
ce012deb | 2577 | static unsigned char padding[8] = { 0x00 }; |
f136f7bf | 2578 | |
b3c96fb1 NTND |
2579 | if (ce->ce_flags & CE_STRIP_NAME) { |
2580 | saved_namelen = ce_namelen(ce); | |
2581 | ce->ce_namelen = 0; | |
00a4b035 | 2582 | stripped_name = 1; |
b3c96fb1 NTND |
2583 | } |
2584 | ||
575fa8a3 | 2585 | size = offsetof(struct ondisk_cache_entry,data) + ondisk_data_size(ce->ce_flags, 0); |
ce012deb | 2586 | |
9d227781 | 2587 | if (!previous_name) { |
ce012deb KW |
2588 | int len = ce_namelen(ce); |
2589 | copy_cache_entry_to_ondisk(ondisk, ce); | |
2590 | result = ce_write(c, fd, ondisk, size); | |
2591 | if (!result) | |
2592 | result = ce_write(c, fd, ce->name, len); | |
2593 | if (!result) | |
2594 | result = ce_write(c, fd, padding, align_padding_size(size, len)); | |
9d227781 JH |
2595 | } else { |
2596 | int common, to_remove, prefix_size; | |
2597 | unsigned char to_remove_vi[16]; | |
2598 | for (common = 0; | |
2599 | (ce->name[common] && | |
2600 | common < previous_name->len && | |
2601 | ce->name[common] == previous_name->buf[common]); | |
2602 | common++) | |
2603 | ; /* still matching */ | |
2604 | to_remove = previous_name->len - common; | |
2605 | prefix_size = encode_varint(to_remove, to_remove_vi); | |
2606 | ||
ce012deb KW |
2607 | copy_cache_entry_to_ondisk(ondisk, ce); |
2608 | result = ce_write(c, fd, ondisk, size); | |
2609 | if (!result) | |
2610 | result = ce_write(c, fd, to_remove_vi, prefix_size); | |
2611 | if (!result) | |
0b90b881 TG |
2612 | result = ce_write(c, fd, ce->name + common, ce_namelen(ce) - common); |
2613 | if (!result) | |
2614 | result = ce_write(c, fd, padding, 1); | |
9d227781 JH |
2615 | |
2616 | strbuf_splice(previous_name, common, to_remove, | |
2617 | ce->name + common, ce_namelen(ce) - common); | |
06aaaa0b | 2618 | } |
00a4b035 | 2619 | if (stripped_name) { |
b3c96fb1 NTND |
2620 | ce->ce_namelen = saved_namelen; |
2621 | ce->ce_flags &= ~CE_STRIP_NAME; | |
2622 | } | |
7a51ed66 | 2623 | |
59efba64 | 2624 | return result; |
7a51ed66 LT |
2625 | } |
2626 | ||
426ddeea YM |
2627 | /* |
2628 | * This function verifies if index_state has the correct sha1 of the | |
2629 | * index file. Don't die if we have any other failure, just return 0. | |
2630 | */ | |
2631 | static int verify_index_from(const struct index_state *istate, const char *path) | |
2632 | { | |
2633 | int fd; | |
2634 | ssize_t n; | |
2635 | struct stat st; | |
aab61359 | 2636 | unsigned char hash[GIT_MAX_RAWSZ]; |
426ddeea YM |
2637 | |
2638 | if (!istate->initialized) | |
2639 | return 0; | |
2640 | ||
2641 | fd = open(path, O_RDONLY); | |
2642 | if (fd < 0) | |
2643 | return 0; | |
2644 | ||
2645 | if (fstat(fd, &st)) | |
2646 | goto out; | |
2647 | ||
aab61359 | 2648 | if (st.st_size < sizeof(struct cache_header) + the_hash_algo->rawsz) |
426ddeea YM |
2649 | goto out; |
2650 | ||
aab61359 | 2651 | n = pread_in_full(fd, hash, the_hash_algo->rawsz, st.st_size - the_hash_algo->rawsz); |
2652 | if (n != the_hash_algo->rawsz) | |
426ddeea YM |
2653 | goto out; |
2654 | ||
67947c34 | 2655 | if (!hasheq(istate->oid.hash, hash)) |
426ddeea YM |
2656 | goto out; |
2657 | ||
2658 | close(fd); | |
2659 | return 1; | |
2660 | ||
2661 | out: | |
2662 | close(fd); | |
2663 | return 0; | |
2664 | } | |
2665 | ||
1b0d968b | 2666 | static int repo_verify_index(struct repository *repo) |
426ddeea | 2667 | { |
1b0d968b | 2668 | return verify_index_from(repo->index, repo->index_file); |
426ddeea YM |
2669 | } |
2670 | ||
483fbe2b JH |
2671 | static int has_racy_timestamp(struct index_state *istate) |
2672 | { | |
2673 | int entries = istate->cache_nr; | |
2674 | int i; | |
2675 | ||
2676 | for (i = 0; i < entries; i++) { | |
2677 | struct cache_entry *ce = istate->cache[i]; | |
2678 | if (is_racy_timestamp(istate, ce)) | |
2679 | return 1; | |
2680 | } | |
2681 | return 0; | |
2682 | } | |
2683 | ||
1b0d968b NTND |
2684 | void repo_update_index_if_able(struct repository *repo, |
2685 | struct lock_file *lockfile) | |
ccdc4ec3 | 2686 | { |
1b0d968b NTND |
2687 | if ((repo->index->cache_changed || |
2688 | has_racy_timestamp(repo->index)) && | |
2689 | repo_verify_index(repo)) | |
2690 | write_locked_index(repo->index, lockfile, COMMIT_LOCK); | |
b74c90fb MÅ |
2691 | else |
2692 | rollback_lock_file(lockfile); | |
ccdc4ec3 JH |
2693 | } |
2694 | ||
d8465500 JN |
2695 | static int record_eoie(void) |
2696 | { | |
2697 | int val; | |
2698 | ||
2699 | if (!git_config_get_bool("index.recordendofindexentries", &val)) | |
2700 | return val; | |
2a9dedef JN |
2701 | |
2702 | /* | |
2703 | * As a convenience, the end of index entries extension | |
2704 | * used for threading is written by default if the user | |
2705 | * explicitly requested threaded index reads. | |
2706 | */ | |
2707 | return !git_config_get_index_threads(&val) && val != 1; | |
d8465500 JN |
2708 | } |
2709 | ||
42916054 JN |
2710 | static int record_ieot(void) |
2711 | { | |
2712 | int val; | |
2713 | ||
2714 | if (!git_config_get_bool("index.recordoffsettable", &val)) | |
2715 | return val; | |
2a9dedef JN |
2716 | |
2717 | /* | |
2718 | * As a convenience, the offset table used for threading is | |
2719 | * written by default if the user explicitly requested | |
2720 | * threaded index reads. | |
2721 | */ | |
2722 | return !git_config_get_index_threads(&val) && val != 1; | |
42916054 JN |
2723 | } |
2724 | ||
812d6b00 MÅ |
2725 | /* |
2726 | * On success, `tempfile` is closed. If it is the temporary file | |
2727 | * of a `struct lock_file`, we will therefore effectively perform | |
2728 | * a 'close_lock_file_gently()`. Since that is an implementation | |
2729 | * detail of lockfiles, callers of `do_write_index()` should not | |
2730 | * rely on it. | |
2731 | */ | |
9f41c7a6 | 2732 | static int do_write_index(struct index_state *istate, struct tempfile *tempfile, |
c18b80a0 | 2733 | int strip_extensions) |
197ee8c9 | 2734 | { |
ca54d9ba | 2735 | uint64_t start = getnanotime(); |
9f41c7a6 | 2736 | int newfd = tempfile->fd; |
3b1d9e04 | 2737 | git_hash_ctx c, eoie_c; |
197ee8c9 | 2738 | struct cache_header hdr; |
b50386c7 | 2739 | int i, err = 0, removed, extended, hdr_version; |
4aab5b46 JH |
2740 | struct cache_entry **cache = istate->cache; |
2741 | int entries = istate->cache_nr; | |
e1afca4f | 2742 | struct stat st; |
575fa8a3 | 2743 | struct ondisk_cache_entry ondisk; |
9d227781 | 2744 | struct strbuf previous_name_buf = STRBUF_INIT, *previous_name; |
4bddd983 | 2745 | int drop_cache_tree = istate->drop_cache_tree; |
3b1d9e04 | 2746 | off_t offset; |
77ff1127 | 2747 | int ieot_entries = 1; |
3255089a BP |
2748 | struct index_entry_offset_table *ieot = NULL; |
2749 | int nr, nr_threads; | |
025a0709 | 2750 | |
06aaaa0b | 2751 | for (i = removed = extended = 0; i < entries; i++) { |
7a51ed66 | 2752 | if (cache[i]->ce_flags & CE_REMOVE) |
025a0709 | 2753 | removed++; |
197ee8c9 | 2754 | |
06aaaa0b NTND |
2755 | /* reduce extended entries if possible */ |
2756 | cache[i]->ce_flags &= ~CE_EXTENDED; | |
2757 | if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) { | |
2758 | extended++; | |
2759 | cache[i]->ce_flags |= CE_EXTENDED; | |
2760 | } | |
2761 | } | |
2762 | ||
d6e3c181 | 2763 | if (!istate->version) { |
136347d7 | 2764 | istate->version = get_index_format_default(); |
4c2db938 | 2765 | if (git_env_bool("GIT_TEST_SPLIT_INDEX", 0)) |
d6e3c181 NTND |
2766 | init_split_index(istate); |
2767 | } | |
9d227781 JH |
2768 | |
2769 | /* demote version 3 to version 2 when the latter suffices */ | |
2770 | if (istate->version == 3 || istate->version == 2) | |
2771 | istate->version = extended ? 3 : 2; | |
2772 | ||
2773 | hdr_version = istate->version; | |
2774 | ||
ccc4feb5 | 2775 | hdr.hdr_signature = htonl(CACHE_SIGNATURE); |
9d227781 | 2776 | hdr.hdr_version = htonl(hdr_version); |
025a0709 | 2777 | hdr.hdr_entries = htonl(entries - removed); |
197ee8c9 | 2778 | |
aab61359 | 2779 | the_hash_algo->init_fn(&c); |
ca9be054 | 2780 | if (ce_write(&c, newfd, &hdr, sizeof(hdr)) < 0) |
197ee8c9 LT |
2781 | return -1; |
2782 | ||
2a9dedef | 2783 | if (!HAVE_THREADS || git_config_get_index_threads(&nr_threads)) |
88168b9b | 2784 | nr_threads = 1; |
62e5ee81 | 2785 | |
42916054 | 2786 | if (nr_threads != 1 && record_ieot()) { |
3255089a BP |
2787 | int ieot_blocks, cpus; |
2788 | ||
2789 | /* | |
2790 | * ensure default number of ieot blocks maps evenly to the | |
2791 | * default number of threads that will process them leaving | |
2792 | * room for the thread to load the index extensions. | |
2793 | */ | |
2794 | if (!nr_threads) { | |
2795 | ieot_blocks = istate->cache_nr / THREAD_COST; | |
2796 | cpus = online_cpus(); | |
2797 | if (ieot_blocks > cpus - 1) | |
2798 | ieot_blocks = cpus - 1; | |
2799 | } else { | |
2800 | ieot_blocks = nr_threads; | |
77ff1127 BP |
2801 | if (ieot_blocks > istate->cache_nr) |
2802 | ieot_blocks = istate->cache_nr; | |
3255089a BP |
2803 | } |
2804 | ||
2805 | /* | |
2806 | * no reason to write out the IEOT extension if we don't | |
2807 | * have enough blocks to utilize multi-threading | |
2808 | */ | |
2809 | if (ieot_blocks > 1) { | |
2810 | ieot = xcalloc(1, sizeof(struct index_entry_offset_table) | |
2811 | + (ieot_blocks * sizeof(struct index_entry_offset))); | |
77ff1127 | 2812 | ieot_entries = DIV_ROUND_UP(entries, ieot_blocks); |
3255089a BP |
2813 | } |
2814 | } | |
3255089a | 2815 | |
3b1d9e04 | 2816 | offset = lseek(newfd, 0, SEEK_CUR); |
3255089a BP |
2817 | if (offset < 0) { |
2818 | free(ieot); | |
3b1d9e04 | 2819 | return -1; |
3255089a | 2820 | } |
3b1d9e04 | 2821 | offset += write_buffer_len; |
3255089a | 2822 | nr = 0; |
9d227781 | 2823 | previous_name = (hdr_version == 4) ? &previous_name_buf : NULL; |
ce012deb | 2824 | |
197ee8c9 LT |
2825 | for (i = 0; i < entries; i++) { |
2826 | struct cache_entry *ce = cache[i]; | |
7a51ed66 | 2827 | if (ce->ce_flags & CE_REMOVE) |
aa16021e | 2828 | continue; |
e06c43c7 | 2829 | if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce)) |
58bf2a4c | 2830 | ce_smudge_racily_clean_entry(istate, ce); |
99d1a986 | 2831 | if (is_null_oid(&ce->oid)) { |
83bd7437 JK |
2832 | static const char msg[] = "cache entry has null sha1: %s"; |
2833 | static int allow = -1; | |
2834 | ||
2835 | if (allow < 0) | |
2836 | allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0); | |
2837 | if (allow) | |
2838 | warning(msg, ce->name); | |
2839 | else | |
b50386c7 | 2840 | err = error(msg, ce->name); |
a96d3cc3 JK |
2841 | |
2842 | drop_cache_tree = 1; | |
83bd7437 | 2843 | } |
77ff1127 | 2844 | if (ieot && i && (i % ieot_entries == 0)) { |
3255089a BP |
2845 | ieot->entries[ieot->nr].nr = nr; |
2846 | ieot->entries[ieot->nr].offset = offset; | |
2847 | ieot->nr++; | |
2848 | /* | |
2849 | * If we have a V4 index, set the first byte to an invalid | |
2850 | * character to ensure there is nothing common with the previous | |
2851 | * entry | |
2852 | */ | |
2853 | if (previous_name) | |
2854 | previous_name->buf[0] = 0; | |
2855 | nr = 0; | |
2856 | offset = lseek(newfd, 0, SEEK_CUR); | |
2857 | if (offset < 0) { | |
2858 | free(ieot); | |
2859 | return -1; | |
2860 | } | |
2861 | offset += write_buffer_len; | |
2862 | } | |
ce012deb | 2863 | if (ce_write_entry(&c, newfd, ce, previous_name, (struct ondisk_cache_entry *)&ondisk) < 0) |
b50386c7 KW |
2864 | err = -1; |
2865 | ||
2866 | if (err) | |
2867 | break; | |
3255089a BP |
2868 | nr++; |
2869 | } | |
2870 | if (ieot && nr) { | |
2871 | ieot->entries[ieot->nr].nr = nr; | |
2872 | ieot->entries[ieot->nr].offset = offset; | |
2873 | ieot->nr++; | |
197ee8c9 | 2874 | } |
9d227781 | 2875 | strbuf_release(&previous_name_buf); |
1af1c2b6 | 2876 | |
3255089a BP |
2877 | if (err) { |
2878 | free(ieot); | |
b50386c7 | 2879 | return err; |
3255089a | 2880 | } |
b50386c7 | 2881 | |
bad68ec9 | 2882 | /* Write extension data here */ |
3b1d9e04 | 2883 | offset = lseek(newfd, 0, SEEK_CUR); |
3255089a BP |
2884 | if (offset < 0) { |
2885 | free(ieot); | |
3b1d9e04 | 2886 | return -1; |
3255089a | 2887 | } |
3b1d9e04 BP |
2888 | offset += write_buffer_len; |
2889 | the_hash_algo->init_fn(&eoie_c); | |
2890 | ||
3255089a BP |
2891 | /* |
2892 | * Lets write out CACHE_EXT_INDEXENTRYOFFSETTABLE first so that we | |
2893 | * can minimize the number of extensions we have to scan through to | |
2894 | * find it during load. Write it out regardless of the | |
2895 | * strip_extensions parameter as we need it when loading the shared | |
2896 | * index. | |
2897 | */ | |
3255089a BP |
2898 | if (ieot) { |
2899 | struct strbuf sb = STRBUF_INIT; | |
2900 | ||
2901 | write_ieot_extension(&sb, ieot); | |
2902 | err = write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_INDEXENTRYOFFSETTABLE, sb.len) < 0 | |
2903 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2904 | strbuf_release(&sb); | |
2905 | free(ieot); | |
2906 | if (err) | |
2907 | return -1; | |
2908 | } | |
3255089a | 2909 | |
6e37c8ed NTND |
2910 | if (!strip_extensions && istate->split_index && |
2911 | !is_null_oid(&istate->split_index->base_oid)) { | |
5fc2fc8f NTND |
2912 | struct strbuf sb = STRBUF_INIT; |
2913 | ||
2914 | err = write_link_extension(&sb, istate) < 0 || | |
3b1d9e04 | 2915 | write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_LINK, |
5fc2fc8f NTND |
2916 | sb.len) < 0 || |
2917 | ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2918 | strbuf_release(&sb); | |
2919 | if (err) | |
2920 | return -1; | |
2921 | } | |
a96d3cc3 | 2922 | if (!strip_extensions && !drop_cache_tree && istate->cache_tree) { |
f285a2d7 | 2923 | struct strbuf sb = STRBUF_INIT; |
1dffb8fa | 2924 | |
1dffb8fa | 2925 | cache_tree_write(&sb, istate->cache_tree); |
3b1d9e04 | 2926 | err = write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_TREE, sb.len) < 0 |
1dffb8fa PH |
2927 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; |
2928 | strbuf_release(&sb); | |
2929 | if (err) | |
bad68ec9 | 2930 | return -1; |
bad68ec9 | 2931 | } |
c18b80a0 | 2932 | if (!strip_extensions && istate->resolve_undo) { |
cfc5789a JH |
2933 | struct strbuf sb = STRBUF_INIT; |
2934 | ||
2935 | resolve_undo_write(&sb, istate->resolve_undo); | |
3b1d9e04 | 2936 | err = write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_RESOLVE_UNDO, |
cfc5789a JH |
2937 | sb.len) < 0 |
2938 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2939 | strbuf_release(&sb); | |
2940 | if (err) | |
2941 | return -1; | |
2942 | } | |
83c094ad NTND |
2943 | if (!strip_extensions && istate->untracked) { |
2944 | struct strbuf sb = STRBUF_INIT; | |
2945 | ||
2946 | write_untracked_extension(&sb, istate->untracked); | |
3b1d9e04 | 2947 | err = write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_UNTRACKED, |
83c094ad NTND |
2948 | sb.len) < 0 || |
2949 | ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2950 | strbuf_release(&sb); | |
2951 | if (err) | |
2952 | return -1; | |
2953 | } | |
883e248b BP |
2954 | if (!strip_extensions && istate->fsmonitor_last_update) { |
2955 | struct strbuf sb = STRBUF_INIT; | |
2956 | ||
2957 | write_fsmonitor_extension(&sb, istate); | |
3b1d9e04 BP |
2958 | err = write_index_ext_header(&c, &eoie_c, newfd, CACHE_EXT_FSMONITOR, sb.len) < 0 |
2959 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2960 | strbuf_release(&sb); | |
2961 | if (err) | |
2962 | return -1; | |
2963 | } | |
2964 | ||
2965 | /* | |
2966 | * CACHE_EXT_ENDOFINDEXENTRIES must be written as the last entry before the SHA1 | |
2967 | * so that it can be found and processed before all the index entries are | |
2968 | * read. Write it out regardless of the strip_extensions parameter as we need it | |
2969 | * when loading the shared index. | |
2970 | */ | |
d8465500 | 2971 | if (offset && record_eoie()) { |
3b1d9e04 BP |
2972 | struct strbuf sb = STRBUF_INIT; |
2973 | ||
2974 | write_eoie_extension(&sb, &eoie_c, offset); | |
2975 | err = write_index_ext_header(&c, NULL, newfd, CACHE_EXT_ENDOFINDEXENTRIES, sb.len) < 0 | |
883e248b BP |
2976 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; |
2977 | strbuf_release(&sb); | |
2978 | if (err) | |
2979 | return -1; | |
2980 | } | |
e1afca4f | 2981 | |
75691ea3 | 2982 | if (ce_flush(&c, newfd, istate->oid.hash)) |
9f41c7a6 | 2983 | return -1; |
49bd0fc2 JK |
2984 | if (close_tempfile_gently(tempfile)) { |
2985 | error(_("could not close '%s'"), tempfile->filename.buf); | |
49bd0fc2 JK |
2986 | return -1; |
2987 | } | |
9f41c7a6 | 2988 | if (stat(tempfile->filename.buf, &st)) |
e1afca4f | 2989 | return -1; |
5bcf109c KB |
2990 | istate->timestamp.sec = (unsigned int)st.st_mtime; |
2991 | istate->timestamp.nsec = ST_MTIME_NSEC(st); | |
ca54d9ba | 2992 | trace_performance_since(start, "write index, changed mask = %x", istate->cache_changed); |
42fee7a3 JH |
2993 | |
2994 | /* | |
2995 | * TODO trace2: replace "the_repository" with the actual repo instance | |
2996 | * that is associated with the given "istate". | |
2997 | */ | |
2998 | trace2_data_intmax("index", the_repository, "write/version", | |
2999 | istate->version); | |
3000 | trace2_data_intmax("index", the_repository, "write/cache_nr", | |
3001 | istate->cache_nr); | |
3002 | ||
e1afca4f | 3003 | return 0; |
197ee8c9 | 3004 | } |
e46bbcf6 | 3005 | |
626f35c8 NTND |
3006 | void set_alternate_index_output(const char *name) |
3007 | { | |
3008 | alternate_index_output = name; | |
3009 | } | |
3010 | ||
3011 | static int commit_locked_index(struct lock_file *lk) | |
3012 | { | |
751baced MH |
3013 | if (alternate_index_output) |
3014 | return commit_lock_file_to(lk, alternate_index_output); | |
3015 | else | |
626f35c8 | 3016 | return commit_lock_file(lk); |
626f35c8 NTND |
3017 | } |
3018 | ||
03b86647 NTND |
3019 | static int do_write_locked_index(struct index_state *istate, struct lock_file *lock, |
3020 | unsigned flags) | |
3021 | { | |
42fee7a3 JH |
3022 | int ret; |
3023 | ||
3024 | /* | |
3025 | * TODO trace2: replace "the_repository" with the actual repo instance | |
3026 | * that is associated with the given "istate". | |
3027 | */ | |
3028 | trace2_region_enter_printf("index", "do_write_index", the_repository, | |
3029 | "%s", lock->tempfile->filename.buf); | |
3030 | ret = do_write_index(istate, lock->tempfile, 0); | |
3031 | trace2_region_leave_printf("index", "do_write_index", the_repository, | |
3032 | "%s", lock->tempfile->filename.buf); | |
3033 | ||
03b86647 NTND |
3034 | if (ret) |
3035 | return ret; | |
03b86647 | 3036 | if (flags & COMMIT_LOCK) |
1956ecd0 BP |
3037 | ret = commit_locked_index(lock); |
3038 | else | |
3039 | ret = close_lock_file_gently(lock); | |
3040 | ||
3041 | run_hook_le(NULL, "post-index-change", | |
3042 | istate->updated_workdir ? "1" : "0", | |
3043 | istate->updated_skipworktree ? "1" : "0", NULL); | |
3044 | istate->updated_workdir = 0; | |
3045 | istate->updated_skipworktree = 0; | |
3046 | ||
3047 | return ret; | |
03b86647 NTND |
3048 | } |
3049 | ||
5fc2fc8f NTND |
3050 | static int write_split_index(struct index_state *istate, |
3051 | struct lock_file *lock, | |
3052 | unsigned flags) | |
3053 | { | |
3054 | int ret; | |
3055 | prepare_to_write_split_index(istate); | |
3056 | ret = do_write_locked_index(istate, lock, flags); | |
3057 | finish_writing_split_index(istate); | |
3058 | return ret; | |
3059 | } | |
3060 | ||
b9683722 CC |
3061 | static const char *shared_index_expire = "2.weeks.ago"; |
3062 | ||
3063 | static unsigned long get_shared_index_expire_date(void) | |
3064 | { | |
3065 | static unsigned long shared_index_expire_date; | |
3066 | static int shared_index_expire_date_prepared; | |
3067 | ||
3068 | if (!shared_index_expire_date_prepared) { | |
3069 | git_config_get_expiry("splitindex.sharedindexexpire", | |
3070 | &shared_index_expire); | |
3071 | shared_index_expire_date = approxidate(shared_index_expire); | |
3072 | shared_index_expire_date_prepared = 1; | |
3073 | } | |
3074 | ||
3075 | return shared_index_expire_date; | |
3076 | } | |
3077 | ||
3078 | static int should_delete_shared_index(const char *shared_index_path) | |
3079 | { | |
3080 | struct stat st; | |
3081 | unsigned long expiration; | |
3082 | ||
3083 | /* Check timestamp */ | |
3084 | expiration = get_shared_index_expire_date(); | |
3085 | if (!expiration) | |
3086 | return 0; | |
3087 | if (stat(shared_index_path, &st)) | |
78bde923 | 3088 | return error_errno(_("could not stat '%s'"), shared_index_path); |
b9683722 CC |
3089 | if (st.st_mtime > expiration) |
3090 | return 0; | |
3091 | ||
3092 | return 1; | |
3093 | } | |
3094 | ||
3095 | static int clean_shared_index_files(const char *current_hex) | |
3096 | { | |
3097 | struct dirent *de; | |
3098 | DIR *dir = opendir(get_git_dir()); | |
3099 | ||
3100 | if (!dir) | |
3101 | return error_errno(_("unable to open git dir: %s"), get_git_dir()); | |
3102 | ||
3103 | while ((de = readdir(dir)) != NULL) { | |
3104 | const char *sha1_hex; | |
3105 | const char *shared_index_path; | |
3106 | if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex)) | |
3107 | continue; | |
3108 | if (!strcmp(sha1_hex, current_hex)) | |
3109 | continue; | |
3110 | shared_index_path = git_path("%s", de->d_name); | |
3111 | if (should_delete_shared_index(shared_index_path) > 0 && | |
3112 | unlink(shared_index_path)) | |
3113 | warning_errno(_("unable to unlink: %s"), shared_index_path); | |
3114 | } | |
3115 | closedir(dir); | |
3116 | ||
3117 | return 0; | |
3118 | } | |
3119 | ||
a0a96756 | 3120 | static int write_shared_index(struct index_state *istate, |
59f9d2dd | 3121 | struct tempfile **temp) |
c18b80a0 NTND |
3122 | { |
3123 | struct split_index *si = istate->split_index; | |
076aa2cb | 3124 | int ret; |
c18b80a0 | 3125 | |
c18b80a0 | 3126 | move_cache_to_base_index(istate); |
42fee7a3 JH |
3127 | |
3128 | trace2_region_enter_printf("index", "shared/do_write_index", | |
3129 | the_repository, "%s", (*temp)->filename.buf); | |
7db2d08c | 3130 | ret = do_write_index(si->base, *temp, 1); |
42fee7a3 JH |
3131 | trace2_region_enter_printf("index", "shared/do_write_index", |
3132 | the_repository, "%s", (*temp)->filename.buf); | |
3133 | ||
59f9d2dd | 3134 | if (ret) |
c18b80a0 | 3135 | return ret; |
7db2d08c | 3136 | ret = adjust_shared_perm(get_tempfile_path(*temp)); |
df801f3f | 3137 | if (ret) { |
9d0a9e90 | 3138 | error(_("cannot fix permission bits on '%s'"), get_tempfile_path(*temp)); |
df801f3f CC |
3139 | return ret; |
3140 | } | |
7db2d08c | 3141 | ret = rename_tempfile(temp, |
75691ea3 | 3142 | git_path("sharedindex.%s", oid_to_hex(&si->base->oid))); |
b9683722 | 3143 | if (!ret) { |
75691ea3 | 3144 | oidcpy(&si->base_oid, &si->base->oid); |
3145 | clean_shared_index_files(oid_to_hex(&si->base->oid)); | |
b9683722 CC |
3146 | } |
3147 | ||
c18b80a0 NTND |
3148 | return ret; |
3149 | } | |
3150 | ||
e6a1dd77 CC |
3151 | static const int default_max_percent_split_change = 20; |
3152 | ||
3153 | static int too_many_not_shared_entries(struct index_state *istate) | |
3154 | { | |
3155 | int i, not_shared = 0; | |
3156 | int max_split = git_config_get_max_percent_split_change(); | |
3157 | ||
3158 | switch (max_split) { | |
3159 | case -1: | |
3160 | /* not or badly configured: use the default value */ | |
3161 | max_split = default_max_percent_split_change; | |
3162 | break; | |
3163 | case 0: | |
3164 | return 1; /* 0% means always write a new shared index */ | |
3165 | case 100: | |
3166 | return 0; /* 100% means never write a new shared index */ | |
3167 | default: | |
3168 | break; /* just use the configured value */ | |
3169 | } | |
3170 | ||
3171 | /* Count not shared entries */ | |
3172 | for (i = 0; i < istate->cache_nr; i++) { | |
3173 | struct cache_entry *ce = istate->cache[i]; | |
3174 | if (!ce->index) | |
3175 | not_shared++; | |
3176 | } | |
3177 | ||
3178 | return (int64_t)istate->cache_nr * max_split < (int64_t)not_shared * 100; | |
3179 | } | |
3180 | ||
03b86647 NTND |
3181 | int write_locked_index(struct index_state *istate, struct lock_file *lock, |
3182 | unsigned flags) | |
3183 | { | |
0d59ffb4 | 3184 | int new_shared_index, ret; |
5fc2fc8f NTND |
3185 | struct split_index *si = istate->split_index; |
3186 | ||
4592e608 | 3187 | if (git_env_bool("GIT_TEST_CHECK_CACHE_TREE", 0)) |
c207e9e1 | 3188 | cache_tree_verify(the_repository, istate); |
4592e608 | 3189 | |
61000814 MÅ |
3190 | if ((flags & SKIP_IF_UNCHANGED) && !istate->cache_changed) { |
3191 | if (flags & COMMIT_LOCK) | |
3192 | rollback_lock_file(lock); | |
3193 | return 0; | |
3194 | } | |
3195 | ||
3bd28eb2 AV |
3196 | if (istate->fsmonitor_last_update) |
3197 | fill_fsmonitor_bitmap(istate); | |
3198 | ||
5165dd59 NTND |
3199 | if (!si || alternate_index_output || |
3200 | (istate->cache_changed & ~EXTMASK)) { | |
5fc2fc8f | 3201 | if (si) |
2182abd9 | 3202 | oidclr(&si->base_oid); |
df60cf57 MÅ |
3203 | ret = do_write_locked_index(istate, lock, flags); |
3204 | goto out; | |
5fc2fc8f NTND |
3205 | } |
3206 | ||
4c2db938 | 3207 | if (git_env_bool("GIT_TEST_SPLIT_INDEX", 0)) { |
2182abd9 | 3208 | int v = si->base_oid.hash[0]; |
d6e3c181 NTND |
3209 | if ((v & 15) < 6) |
3210 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
3211 | } | |
e6a1dd77 CC |
3212 | if (too_many_not_shared_entries(istate)) |
3213 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
0d59ffb4 CC |
3214 | |
3215 | new_shared_index = istate->cache_changed & SPLIT_INDEX_ORDERED; | |
3216 | ||
3217 | if (new_shared_index) { | |
59f9d2dd NTND |
3218 | struct tempfile *temp; |
3219 | int saved_errno; | |
3220 | ||
c9d6c788 ÆAB |
3221 | /* Same initial permissions as the main .git/index file */ |
3222 | temp = mks_tempfile_sm(git_path("sharedindex_XXXXXX"), 0, 0666); | |
59f9d2dd | 3223 | if (!temp) { |
2182abd9 | 3224 | oidclr(&si->base_oid); |
59f9d2dd | 3225 | ret = do_write_locked_index(istate, lock, flags); |
ef5b3a6c NTND |
3226 | goto out; |
3227 | } | |
3228 | ret = write_shared_index(istate, &temp); | |
59f9d2dd NTND |
3229 | |
3230 | saved_errno = errno; | |
3231 | if (is_tempfile_active(temp)) | |
3232 | delete_tempfile(&temp); | |
3233 | errno = saved_errno; | |
3234 | ||
c18b80a0 | 3235 | if (ret) |
df60cf57 | 3236 | goto out; |
c18b80a0 NTND |
3237 | } |
3238 | ||
0d59ffb4 CC |
3239 | ret = write_split_index(istate, lock, flags); |
3240 | ||
3241 | /* Freshen the shared index only if the split-index was written */ | |
6e37c8ed | 3242 | if (!ret && !new_shared_index && !is_null_oid(&si->base_oid)) { |
a125a223 | 3243 | const char *shared_index = git_path("sharedindex.%s", |
2182abd9 | 3244 | oid_to_hex(&si->base_oid)); |
a125a223 TG |
3245 | freshen_shared_index(shared_index, 1); |
3246 | } | |
0d59ffb4 | 3247 | |
df60cf57 MÅ |
3248 | out: |
3249 | if (flags & COMMIT_LOCK) | |
3250 | rollback_lock_file(lock); | |
0d59ffb4 | 3251 | return ret; |
03b86647 NTND |
3252 | } |
3253 | ||
e46bbcf6 MV |
3254 | /* |
3255 | * Read the index file that is potentially unmerged into given | |
ad376204 EN |
3256 | * index_state, dropping any unmerged entries to stage #0 (potentially |
3257 | * resulting in a path appearing as both a file and a directory in the | |
3258 | * index; the caller is responsible to clear out the extra entries | |
3259 | * before writing the index to a tree). Returns true if the index is | |
3260 | * unmerged. Callers who want to refuse to work from an unmerged | |
3261 | * state can call this and check its return value, instead of calling | |
3262 | * read_cache(). | |
e46bbcf6 | 3263 | */ |
e1ff0a32 | 3264 | int repo_read_index_unmerged(struct repository *repo) |
e46bbcf6 | 3265 | { |
e1ff0a32 | 3266 | struct index_state *istate; |
e46bbcf6 | 3267 | int i; |
d1a43f2a | 3268 | int unmerged = 0; |
e46bbcf6 | 3269 | |
e1ff0a32 NTND |
3270 | repo_read_index(repo); |
3271 | istate = repo->index; | |
e46bbcf6 MV |
3272 | for (i = 0; i < istate->cache_nr; i++) { |
3273 | struct cache_entry *ce = istate->cache[i]; | |
d1a43f2a | 3274 | struct cache_entry *new_ce; |
a849735b | 3275 | int len; |
d1a43f2a JH |
3276 | |
3277 | if (!ce_stage(ce)) | |
e46bbcf6 | 3278 | continue; |
d1a43f2a | 3279 | unmerged = 1; |
68c4f6a5 | 3280 | len = ce_namelen(ce); |
a849735b | 3281 | new_ce = make_empty_cache_entry(istate, len); |
d1a43f2a | 3282 | memcpy(new_ce->name, ce->name, len); |
b60e188c TG |
3283 | new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED; |
3284 | new_ce->ce_namelen = len; | |
d1a43f2a | 3285 | new_ce->ce_mode = ce->ce_mode; |
ad376204 | 3286 | if (add_index_entry(istate, new_ce, ADD_CACHE_SKIP_DFCHECK)) |
9d0a9e90 | 3287 | return error(_("%s: cannot drop to stage #0"), |
5699d17e | 3288 | new_ce->name); |
e46bbcf6 | 3289 | } |
d1a43f2a | 3290 | return unmerged; |
e46bbcf6 | 3291 | } |
041aee31 | 3292 | |
98fa4738 JK |
3293 | /* |
3294 | * Returns 1 if the path is an "other" path with respect to | |
3295 | * the index; that is, the path is not mentioned in the index at all, | |
3296 | * either as a file, a directory with some files in the index, | |
3297 | * or as an unmerged entry. | |
3298 | * | |
3299 | * We helpfully remove a trailing "/" from directories so that | |
3300 | * the output of read_directory can be used as-is. | |
3301 | */ | |
3302 | int index_name_is_other(const struct index_state *istate, const char *name, | |
3303 | int namelen) | |
3304 | { | |
3305 | int pos; | |
3306 | if (namelen && name[namelen - 1] == '/') | |
3307 | namelen--; | |
3308 | pos = index_name_pos(istate, name, namelen); | |
3309 | if (0 <= pos) | |
3310 | return 0; /* exact match */ | |
3311 | pos = -pos - 1; | |
3312 | if (pos < istate->cache_nr) { | |
3313 | struct cache_entry *ce = istate->cache[pos]; | |
3314 | if (ce_namelen(ce) == namelen && | |
3315 | !memcmp(ce->name, name, namelen)) | |
3316 | return 0; /* Yup, this one exists unmerged */ | |
3317 | } | |
3318 | return 1; | |
3319 | } | |
29fb37b2 | 3320 | |
87542508 BW |
3321 | void *read_blob_data_from_index(const struct index_state *istate, |
3322 | const char *path, unsigned long *size) | |
29fb37b2 LF |
3323 | { |
3324 | int pos, len; | |
3325 | unsigned long sz; | |
3326 | enum object_type type; | |
3327 | void *data; | |
3328 | ||
3329 | len = strlen(path); | |
3330 | pos = index_name_pos(istate, path, len); | |
3331 | if (pos < 0) { | |
3332 | /* | |
3333 | * We might be in the middle of a merge, in which | |
3334 | * case we would read stage #2 (ours). | |
3335 | */ | |
3336 | int i; | |
3337 | for (i = -pos - 1; | |
3338 | (pos < 0 && i < istate->cache_nr && | |
3339 | !strcmp(istate->cache[i]->name, path)); | |
3340 | i++) | |
3341 | if (ce_stage(istate->cache[i]) == 2) | |
3342 | pos = i; | |
3343 | } | |
3344 | if (pos < 0) | |
3345 | return NULL; | |
b4f5aca4 | 3346 | data = read_object_file(&istate->cache[pos]->oid, &type, &sz); |
29fb37b2 LF |
3347 | if (!data || type != OBJ_BLOB) { |
3348 | free(data); | |
3349 | return NULL; | |
3350 | } | |
ff366825 LF |
3351 | if (size) |
3352 | *size = sz; | |
29fb37b2 LF |
3353 | return data; |
3354 | } | |
38612532 MH |
3355 | |
3356 | void stat_validity_clear(struct stat_validity *sv) | |
3357 | { | |
6a83d902 | 3358 | FREE_AND_NULL(sv->sd); |
38612532 MH |
3359 | } |
3360 | ||
3361 | int stat_validity_check(struct stat_validity *sv, const char *path) | |
3362 | { | |
3363 | struct stat st; | |
3364 | ||
3365 | if (stat(path, &st) < 0) | |
3366 | return sv->sd == NULL; | |
3367 | if (!sv->sd) | |
3368 | return 0; | |
3369 | return S_ISREG(st.st_mode) && !match_stat_data(sv->sd, &st); | |
3370 | } | |
3371 | ||
3372 | void stat_validity_update(struct stat_validity *sv, int fd) | |
3373 | { | |
3374 | struct stat st; | |
3375 | ||
3376 | if (fstat(fd, &st) < 0 || !S_ISREG(st.st_mode)) | |
3377 | stat_validity_clear(sv); | |
3378 | else { | |
3379 | if (!sv->sd) | |
3380 | sv->sd = xcalloc(1, sizeof(struct stat_data)); | |
3381 | fill_stat_data(sv->sd, &st); | |
3382 | } | |
3383 | } | |
edf3b905 DT |
3384 | |
3385 | void move_index_extensions(struct index_state *dst, struct index_state *src) | |
3386 | { | |
3387 | dst->untracked = src->untracked; | |
3388 | src->untracked = NULL; | |
836ef2b6 NTND |
3389 | dst->cache_tree = src->cache_tree; |
3390 | src->cache_tree = NULL; | |
edf3b905 | 3391 | } |
a849735b | 3392 | |
8e72d675 JM |
3393 | struct cache_entry *dup_cache_entry(const struct cache_entry *ce, |
3394 | struct index_state *istate) | |
3395 | { | |
3396 | unsigned int size = ce_size(ce); | |
3397 | int mem_pool_allocated; | |
3398 | struct cache_entry *new_entry = make_empty_cache_entry(istate, ce_namelen(ce)); | |
3399 | mem_pool_allocated = new_entry->mem_pool_allocated; | |
3400 | ||
3401 | memcpy(new_entry, ce, size); | |
3402 | new_entry->mem_pool_allocated = mem_pool_allocated; | |
3403 | return new_entry; | |
3404 | } | |
3405 | ||
a849735b JM |
3406 | void discard_cache_entry(struct cache_entry *ce) |
3407 | { | |
8616a2d0 JM |
3408 | if (ce && should_validate_cache_entries()) |
3409 | memset(ce, 0xCD, cache_entry_size(ce->ce_namelen)); | |
3410 | ||
8e72d675 JM |
3411 | if (ce && ce->mem_pool_allocated) |
3412 | return; | |
3413 | ||
a849735b JM |
3414 | free(ce); |
3415 | } | |
8616a2d0 JM |
3416 | |
3417 | int should_validate_cache_entries(void) | |
3418 | { | |
3419 | static int validate_index_cache_entries = -1; | |
3420 | ||
3421 | if (validate_index_cache_entries < 0) { | |
3422 | if (getenv("GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES")) | |
3423 | validate_index_cache_entries = 1; | |
3424 | else | |
3425 | validate_index_cache_entries = 0; | |
3426 | } | |
3427 | ||
3428 | return validate_index_cache_entries; | |
3429 | } | |
3b1d9e04 BP |
3430 | |
3431 | #define EOIE_SIZE (4 + GIT_SHA1_RAWSZ) /* <4-byte offset> + <20-byte hash> */ | |
3432 | #define EOIE_SIZE_WITH_HEADER (4 + 4 + EOIE_SIZE) /* <4-byte signature> + <4-byte length> + EOIE_SIZE */ | |
3433 | ||
3434 | static size_t read_eoie_extension(const char *mmap, size_t mmap_size) | |
3435 | { | |
3436 | /* | |
3437 | * The end of index entries (EOIE) extension is guaranteed to be last | |
3438 | * so that it can be found by scanning backwards from the EOF. | |
3439 | * | |
3440 | * "EOIE" | |
3441 | * <4-byte length> | |
3442 | * <4-byte offset> | |
3443 | * <20-byte hash> | |
3444 | */ | |
3445 | const char *index, *eoie; | |
3446 | uint32_t extsize; | |
3447 | size_t offset, src_offset; | |
3448 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3449 | git_hash_ctx c; | |
3450 | ||
3451 | /* ensure we have an index big enough to contain an EOIE extension */ | |
3452 | if (mmap_size < sizeof(struct cache_header) + EOIE_SIZE_WITH_HEADER + the_hash_algo->rawsz) | |
3453 | return 0; | |
3454 | ||
3455 | /* validate the extension signature */ | |
3456 | index = eoie = mmap + mmap_size - EOIE_SIZE_WITH_HEADER - the_hash_algo->rawsz; | |
3457 | if (CACHE_EXT(index) != CACHE_EXT_ENDOFINDEXENTRIES) | |
3458 | return 0; | |
3459 | index += sizeof(uint32_t); | |
3460 | ||
3461 | /* validate the extension size */ | |
3462 | extsize = get_be32(index); | |
3463 | if (extsize != EOIE_SIZE) | |
3464 | return 0; | |
3465 | index += sizeof(uint32_t); | |
3466 | ||
3467 | /* | |
3468 | * Validate the offset we're going to look for the first extension | |
3469 | * signature is after the index header and before the eoie extension. | |
3470 | */ | |
3471 | offset = get_be32(index); | |
3472 | if (mmap + offset < mmap + sizeof(struct cache_header)) | |
3473 | return 0; | |
3474 | if (mmap + offset >= eoie) | |
3475 | return 0; | |
3476 | index += sizeof(uint32_t); | |
3477 | ||
3478 | /* | |
3479 | * The hash is computed over extension types and their sizes (but not | |
3480 | * their contents). E.g. if we have "TREE" extension that is N-bytes | |
3481 | * long, "REUC" extension that is M-bytes long, followed by "EOIE", | |
3482 | * then the hash would be: | |
3483 | * | |
3484 | * SHA-1("TREE" + <binary representation of N> + | |
3485 | * "REUC" + <binary representation of M>) | |
3486 | */ | |
3487 | src_offset = offset; | |
3488 | the_hash_algo->init_fn(&c); | |
3489 | while (src_offset < mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) { | |
3490 | /* After an array of active_nr index entries, | |
3491 | * there can be arbitrary number of extended | |
3492 | * sections, each of which is prefixed with | |
3493 | * extension name (4-byte) and section length | |
3494 | * in 4-byte network byte order. | |
3495 | */ | |
3496 | uint32_t extsize; | |
3497 | memcpy(&extsize, mmap + src_offset + 4, 4); | |
3498 | extsize = ntohl(extsize); | |
3499 | ||
3500 | /* verify the extension size isn't so large it will wrap around */ | |
3501 | if (src_offset + 8 + extsize < src_offset) | |
3502 | return 0; | |
3503 | ||
3504 | the_hash_algo->update_fn(&c, mmap + src_offset, 8); | |
3505 | ||
3506 | src_offset += 8; | |
3507 | src_offset += extsize; | |
3508 | } | |
3509 | the_hash_algo->final_fn(hash, &c); | |
3510 | if (!hasheq(hash, (const unsigned char *)index)) | |
3511 | return 0; | |
3512 | ||
3513 | /* Validate that the extension offsets returned us back to the eoie extension. */ | |
3514 | if (src_offset != mmap_size - the_hash_algo->rawsz - EOIE_SIZE_WITH_HEADER) | |
3515 | return 0; | |
3516 | ||
3517 | return offset; | |
3518 | } | |
3519 | ||
3520 | static void write_eoie_extension(struct strbuf *sb, git_hash_ctx *eoie_context, size_t offset) | |
3521 | { | |
3522 | uint32_t buffer; | |
3523 | unsigned char hash[GIT_MAX_RAWSZ]; | |
3524 | ||
3525 | /* offset */ | |
3526 | put_be32(&buffer, offset); | |
3527 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3528 | ||
3529 | /* hash */ | |
3530 | the_hash_algo->final_fn(hash, eoie_context); | |
3531 | strbuf_add(sb, hash, the_hash_algo->rawsz); | |
3532 | } | |
3255089a | 3533 | |
3255089a BP |
3534 | #define IEOT_VERSION (1) |
3535 | ||
3536 | static struct index_entry_offset_table *read_ieot_extension(const char *mmap, size_t mmap_size, size_t offset) | |
3537 | { | |
ec36c42a NTND |
3538 | const char *index = NULL; |
3539 | uint32_t extsize, ext_version; | |
3540 | struct index_entry_offset_table *ieot; | |
3541 | int i, nr; | |
3542 | ||
3543 | /* find the IEOT extension */ | |
3544 | if (!offset) | |
3545 | return NULL; | |
3546 | while (offset <= mmap_size - the_hash_algo->rawsz - 8) { | |
3547 | extsize = get_be32(mmap + offset + 4); | |
3548 | if (CACHE_EXT((mmap + offset)) == CACHE_EXT_INDEXENTRYOFFSETTABLE) { | |
3549 | index = mmap + offset + 4 + 4; | |
3550 | break; | |
3551 | } | |
3552 | offset += 8; | |
3553 | offset += extsize; | |
3554 | } | |
3555 | if (!index) | |
3556 | return NULL; | |
3557 | ||
3558 | /* validate the version is IEOT_VERSION */ | |
3559 | ext_version = get_be32(index); | |
3560 | if (ext_version != IEOT_VERSION) { | |
3561 | error("invalid IEOT version %d", ext_version); | |
3562 | return NULL; | |
3563 | } | |
3564 | index += sizeof(uint32_t); | |
3565 | ||
3566 | /* extension size - version bytes / bytes per entry */ | |
3567 | nr = (extsize - sizeof(uint32_t)) / (sizeof(uint32_t) + sizeof(uint32_t)); | |
3568 | if (!nr) { | |
3569 | error("invalid number of IEOT entries %d", nr); | |
3570 | return NULL; | |
3571 | } | |
3572 | ieot = xmalloc(sizeof(struct index_entry_offset_table) | |
3573 | + (nr * sizeof(struct index_entry_offset))); | |
3574 | ieot->nr = nr; | |
3575 | for (i = 0; i < nr; i++) { | |
3576 | ieot->entries[i].offset = get_be32(index); | |
3577 | index += sizeof(uint32_t); | |
3578 | ieot->entries[i].nr = get_be32(index); | |
3579 | index += sizeof(uint32_t); | |
3580 | } | |
3581 | ||
3582 | return ieot; | |
3255089a BP |
3583 | } |
3584 | ||
3585 | static void write_ieot_extension(struct strbuf *sb, struct index_entry_offset_table *ieot) | |
3586 | { | |
ec36c42a NTND |
3587 | uint32_t buffer; |
3588 | int i; | |
3255089a | 3589 | |
ec36c42a NTND |
3590 | /* version */ |
3591 | put_be32(&buffer, IEOT_VERSION); | |
3592 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3593 | |
ec36c42a NTND |
3594 | /* ieot */ |
3595 | for (i = 0; i < ieot->nr; i++) { | |
3255089a | 3596 | |
ec36c42a NTND |
3597 | /* offset */ |
3598 | put_be32(&buffer, ieot->entries[i].offset); | |
3599 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3255089a | 3600 | |
ec36c42a NTND |
3601 | /* count */ |
3602 | put_be32(&buffer, ieot->entries[i].nr); | |
3603 | strbuf_add(sb, &buffer, sizeof(uint32_t)); | |
3604 | } | |
3255089a | 3605 | } |