]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * GIT - The information manager from hell | |
3 | * | |
4 | * Copyright (C) Linus Torvalds, 2005 | |
5 | */ | |
6 | #define NO_THE_INDEX_COMPATIBILITY_MACROS | |
7 | #include "cache.h" | |
8 | #include "config.h" | |
9 | #include "diff.h" | |
10 | #include "diffcore.h" | |
11 | #include "tempfile.h" | |
12 | #include "lockfile.h" | |
13 | #include "cache-tree.h" | |
14 | #include "refs.h" | |
15 | #include "dir.h" | |
16 | #include "object-store.h" | |
17 | #include "tree.h" | |
18 | #include "commit.h" | |
19 | #include "blob.h" | |
20 | #include "resolve-undo.h" | |
21 | #include "strbuf.h" | |
22 | #include "varint.h" | |
23 | #include "split-index.h" | |
24 | #include "utf8.h" | |
25 | #include "fsmonitor.h" | |
26 | ||
27 | /* Mask for the name length in ce_flags in the on-disk index */ | |
28 | ||
29 | #define CE_NAMEMASK (0x0fff) | |
30 | ||
31 | /* Index extensions. | |
32 | * | |
33 | * The first letter should be 'A'..'Z' for extensions that are not | |
34 | * necessary for a correct operation (i.e. optimization data). | |
35 | * When new extensions are added that _needs_ to be understood in | |
36 | * order to correctly interpret the index file, pick character that | |
37 | * is outside the range, to cause the reader to abort. | |
38 | */ | |
39 | ||
40 | #define CACHE_EXT(s) ( (s[0]<<24)|(s[1]<<16)|(s[2]<<8)|(s[3]) ) | |
41 | #define CACHE_EXT_TREE 0x54524545 /* "TREE" */ | |
42 | #define CACHE_EXT_RESOLVE_UNDO 0x52455543 /* "REUC" */ | |
43 | #define CACHE_EXT_LINK 0x6c696e6b /* "link" */ | |
44 | #define CACHE_EXT_UNTRACKED 0x554E5452 /* "UNTR" */ | |
45 | #define CACHE_EXT_FSMONITOR 0x46534D4E /* "FSMN" */ | |
46 | ||
47 | /* changes that can be kept in $GIT_DIR/index (basically all extensions) */ | |
48 | #define EXTMASK (RESOLVE_UNDO_CHANGED | CACHE_TREE_CHANGED | \ | |
49 | CE_ENTRY_ADDED | CE_ENTRY_REMOVED | CE_ENTRY_CHANGED | \ | |
50 | SPLIT_INDEX_ORDERED | UNTRACKED_CHANGED | FSMONITOR_CHANGED) | |
51 | ||
52 | ||
53 | /* | |
54 | * This is an estimate of the pathname length in the index. We use | |
55 | * this for V4 index files to guess the un-deltafied size of the index | |
56 | * in memory because of pathname deltafication. This is not required | |
57 | * for V2/V3 index formats because their pathnames are not compressed. | |
58 | * If the initial amount of memory set aside is not sufficient, the | |
59 | * mem pool will allocate extra memory. | |
60 | */ | |
61 | #define CACHE_ENTRY_PATH_LENGTH 80 | |
62 | ||
63 | static inline struct cache_entry *mem_pool__ce_alloc(struct mem_pool *mem_pool, size_t len) | |
64 | { | |
65 | struct cache_entry *ce; | |
66 | ce = mem_pool_alloc(mem_pool, cache_entry_size(len)); | |
67 | ce->mem_pool_allocated = 1; | |
68 | return ce; | |
69 | } | |
70 | ||
71 | static inline struct cache_entry *mem_pool__ce_calloc(struct mem_pool *mem_pool, size_t len) | |
72 | { | |
73 | struct cache_entry * ce; | |
74 | ce = mem_pool_calloc(mem_pool, 1, cache_entry_size(len)); | |
75 | ce->mem_pool_allocated = 1; | |
76 | return ce; | |
77 | } | |
78 | ||
79 | static struct mem_pool *find_mem_pool(struct index_state *istate) | |
80 | { | |
81 | struct mem_pool **pool_ptr; | |
82 | ||
83 | if (istate->split_index && istate->split_index->base) | |
84 | pool_ptr = &istate->split_index->base->ce_mem_pool; | |
85 | else | |
86 | pool_ptr = &istate->ce_mem_pool; | |
87 | ||
88 | if (!*pool_ptr) | |
89 | mem_pool_init(pool_ptr, 0); | |
90 | ||
91 | return *pool_ptr; | |
92 | } | |
93 | ||
94 | struct index_state the_index; | |
95 | static const char *alternate_index_output; | |
96 | ||
97 | static void set_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) | |
98 | { | |
99 | istate->cache[nr] = ce; | |
100 | add_name_hash(istate, ce); | |
101 | } | |
102 | ||
103 | static void replace_index_entry(struct index_state *istate, int nr, struct cache_entry *ce) | |
104 | { | |
105 | struct cache_entry *old = istate->cache[nr]; | |
106 | ||
107 | replace_index_entry_in_base(istate, old, ce); | |
108 | remove_name_hash(istate, old); | |
109 | discard_cache_entry(old); | |
110 | ce->ce_flags &= ~CE_HASHED; | |
111 | set_index_entry(istate, nr, ce); | |
112 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
113 | mark_fsmonitor_invalid(istate, ce); | |
114 | istate->cache_changed |= CE_ENTRY_CHANGED; | |
115 | } | |
116 | ||
117 | void rename_index_entry_at(struct index_state *istate, int nr, const char *new_name) | |
118 | { | |
119 | struct cache_entry *old_entry = istate->cache[nr], *new_entry; | |
120 | int namelen = strlen(new_name); | |
121 | ||
122 | new_entry = make_empty_cache_entry(istate, namelen); | |
123 | copy_cache_entry(new_entry, old_entry); | |
124 | new_entry->ce_flags &= ~CE_HASHED; | |
125 | new_entry->ce_namelen = namelen; | |
126 | new_entry->index = 0; | |
127 | memcpy(new_entry->name, new_name, namelen + 1); | |
128 | ||
129 | cache_tree_invalidate_path(istate, old_entry->name); | |
130 | untracked_cache_remove_from_index(istate, old_entry->name); | |
131 | remove_index_entry_at(istate, nr); | |
132 | add_index_entry(istate, new_entry, ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE); | |
133 | } | |
134 | ||
135 | void fill_stat_data(struct stat_data *sd, struct stat *st) | |
136 | { | |
137 | sd->sd_ctime.sec = (unsigned int)st->st_ctime; | |
138 | sd->sd_mtime.sec = (unsigned int)st->st_mtime; | |
139 | sd->sd_ctime.nsec = ST_CTIME_NSEC(*st); | |
140 | sd->sd_mtime.nsec = ST_MTIME_NSEC(*st); | |
141 | sd->sd_dev = st->st_dev; | |
142 | sd->sd_ino = st->st_ino; | |
143 | sd->sd_uid = st->st_uid; | |
144 | sd->sd_gid = st->st_gid; | |
145 | sd->sd_size = st->st_size; | |
146 | } | |
147 | ||
148 | int match_stat_data(const struct stat_data *sd, struct stat *st) | |
149 | { | |
150 | int changed = 0; | |
151 | ||
152 | if (sd->sd_mtime.sec != (unsigned int)st->st_mtime) | |
153 | changed |= MTIME_CHANGED; | |
154 | if (trust_ctime && check_stat && | |
155 | sd->sd_ctime.sec != (unsigned int)st->st_ctime) | |
156 | changed |= CTIME_CHANGED; | |
157 | ||
158 | #ifdef USE_NSEC | |
159 | if (check_stat && sd->sd_mtime.nsec != ST_MTIME_NSEC(*st)) | |
160 | changed |= MTIME_CHANGED; | |
161 | if (trust_ctime && check_stat && | |
162 | sd->sd_ctime.nsec != ST_CTIME_NSEC(*st)) | |
163 | changed |= CTIME_CHANGED; | |
164 | #endif | |
165 | ||
166 | if (check_stat) { | |
167 | if (sd->sd_uid != (unsigned int) st->st_uid || | |
168 | sd->sd_gid != (unsigned int) st->st_gid) | |
169 | changed |= OWNER_CHANGED; | |
170 | if (sd->sd_ino != (unsigned int) st->st_ino) | |
171 | changed |= INODE_CHANGED; | |
172 | } | |
173 | ||
174 | #ifdef USE_STDEV | |
175 | /* | |
176 | * st_dev breaks on network filesystems where different | |
177 | * clients will have different views of what "device" | |
178 | * the filesystem is on | |
179 | */ | |
180 | if (check_stat && sd->sd_dev != (unsigned int) st->st_dev) | |
181 | changed |= INODE_CHANGED; | |
182 | #endif | |
183 | ||
184 | if (sd->sd_size != (unsigned int) st->st_size) | |
185 | changed |= DATA_CHANGED; | |
186 | ||
187 | return changed; | |
188 | } | |
189 | ||
190 | /* | |
191 | * This only updates the "non-critical" parts of the directory | |
192 | * cache, ie the parts that aren't tracked by GIT, and only used | |
193 | * to validate the cache. | |
194 | */ | |
195 | void fill_stat_cache_info(struct cache_entry *ce, struct stat *st) | |
196 | { | |
197 | fill_stat_data(&ce->ce_stat_data, st); | |
198 | ||
199 | if (assume_unchanged) | |
200 | ce->ce_flags |= CE_VALID; | |
201 | ||
202 | if (S_ISREG(st->st_mode)) { | |
203 | ce_mark_uptodate(ce); | |
204 | mark_fsmonitor_valid(ce); | |
205 | } | |
206 | } | |
207 | ||
208 | static int ce_compare_data(const struct cache_entry *ce, struct stat *st) | |
209 | { | |
210 | int match = -1; | |
211 | int fd = git_open_cloexec(ce->name, O_RDONLY); | |
212 | ||
213 | if (fd >= 0) { | |
214 | struct object_id oid; | |
215 | if (!index_fd(&oid, fd, st, OBJ_BLOB, ce->name, 0)) | |
216 | match = !oideq(&oid, &ce->oid); | |
217 | /* index_fd() closed the file descriptor already */ | |
218 | } | |
219 | return match; | |
220 | } | |
221 | ||
222 | static int ce_compare_link(const struct cache_entry *ce, size_t expected_size) | |
223 | { | |
224 | int match = -1; | |
225 | void *buffer; | |
226 | unsigned long size; | |
227 | enum object_type type; | |
228 | struct strbuf sb = STRBUF_INIT; | |
229 | ||
230 | if (strbuf_readlink(&sb, ce->name, expected_size)) | |
231 | return -1; | |
232 | ||
233 | buffer = read_object_file(&ce->oid, &type, &size); | |
234 | if (buffer) { | |
235 | if (size == sb.len) | |
236 | match = memcmp(buffer, sb.buf, size); | |
237 | free(buffer); | |
238 | } | |
239 | strbuf_release(&sb); | |
240 | return match; | |
241 | } | |
242 | ||
243 | static int ce_compare_gitlink(const struct cache_entry *ce) | |
244 | { | |
245 | struct object_id oid; | |
246 | ||
247 | /* | |
248 | * We don't actually require that the .git directory | |
249 | * under GITLINK directory be a valid git directory. It | |
250 | * might even be missing (in case nobody populated that | |
251 | * sub-project). | |
252 | * | |
253 | * If so, we consider it always to match. | |
254 | */ | |
255 | if (resolve_gitlink_ref(ce->name, "HEAD", &oid) < 0) | |
256 | return 0; | |
257 | return !oideq(&oid, &ce->oid); | |
258 | } | |
259 | ||
260 | static int ce_modified_check_fs(const struct cache_entry *ce, struct stat *st) | |
261 | { | |
262 | switch (st->st_mode & S_IFMT) { | |
263 | case S_IFREG: | |
264 | if (ce_compare_data(ce, st)) | |
265 | return DATA_CHANGED; | |
266 | break; | |
267 | case S_IFLNK: | |
268 | if (ce_compare_link(ce, xsize_t(st->st_size))) | |
269 | return DATA_CHANGED; | |
270 | break; | |
271 | case S_IFDIR: | |
272 | if (S_ISGITLINK(ce->ce_mode)) | |
273 | return ce_compare_gitlink(ce) ? DATA_CHANGED : 0; | |
274 | /* else fallthrough */ | |
275 | default: | |
276 | return TYPE_CHANGED; | |
277 | } | |
278 | return 0; | |
279 | } | |
280 | ||
281 | static int ce_match_stat_basic(const struct cache_entry *ce, struct stat *st) | |
282 | { | |
283 | unsigned int changed = 0; | |
284 | ||
285 | if (ce->ce_flags & CE_REMOVE) | |
286 | return MODE_CHANGED | DATA_CHANGED | TYPE_CHANGED; | |
287 | ||
288 | switch (ce->ce_mode & S_IFMT) { | |
289 | case S_IFREG: | |
290 | changed |= !S_ISREG(st->st_mode) ? TYPE_CHANGED : 0; | |
291 | /* We consider only the owner x bit to be relevant for | |
292 | * "mode changes" | |
293 | */ | |
294 | if (trust_executable_bit && | |
295 | (0100 & (ce->ce_mode ^ st->st_mode))) | |
296 | changed |= MODE_CHANGED; | |
297 | break; | |
298 | case S_IFLNK: | |
299 | if (!S_ISLNK(st->st_mode) && | |
300 | (has_symlinks || !S_ISREG(st->st_mode))) | |
301 | changed |= TYPE_CHANGED; | |
302 | break; | |
303 | case S_IFGITLINK: | |
304 | /* We ignore most of the st_xxx fields for gitlinks */ | |
305 | if (!S_ISDIR(st->st_mode)) | |
306 | changed |= TYPE_CHANGED; | |
307 | else if (ce_compare_gitlink(ce)) | |
308 | changed |= DATA_CHANGED; | |
309 | return changed; | |
310 | default: | |
311 | die("internal error: ce_mode is %o", ce->ce_mode); | |
312 | } | |
313 | ||
314 | changed |= match_stat_data(&ce->ce_stat_data, st); | |
315 | ||
316 | /* Racily smudged entry? */ | |
317 | if (!ce->ce_stat_data.sd_size) { | |
318 | if (!is_empty_blob_sha1(ce->oid.hash)) | |
319 | changed |= DATA_CHANGED; | |
320 | } | |
321 | ||
322 | return changed; | |
323 | } | |
324 | ||
325 | static int is_racy_stat(const struct index_state *istate, | |
326 | const struct stat_data *sd) | |
327 | { | |
328 | return (istate->timestamp.sec && | |
329 | #ifdef USE_NSEC | |
330 | /* nanosecond timestamped files can also be racy! */ | |
331 | (istate->timestamp.sec < sd->sd_mtime.sec || | |
332 | (istate->timestamp.sec == sd->sd_mtime.sec && | |
333 | istate->timestamp.nsec <= sd->sd_mtime.nsec)) | |
334 | #else | |
335 | istate->timestamp.sec <= sd->sd_mtime.sec | |
336 | #endif | |
337 | ); | |
338 | } | |
339 | ||
340 | static int is_racy_timestamp(const struct index_state *istate, | |
341 | const struct cache_entry *ce) | |
342 | { | |
343 | return (!S_ISGITLINK(ce->ce_mode) && | |
344 | is_racy_stat(istate, &ce->ce_stat_data)); | |
345 | } | |
346 | ||
347 | int match_stat_data_racy(const struct index_state *istate, | |
348 | const struct stat_data *sd, struct stat *st) | |
349 | { | |
350 | if (is_racy_stat(istate, sd)) | |
351 | return MTIME_CHANGED; | |
352 | return match_stat_data(sd, st); | |
353 | } | |
354 | ||
355 | int ie_match_stat(struct index_state *istate, | |
356 | const struct cache_entry *ce, struct stat *st, | |
357 | unsigned int options) | |
358 | { | |
359 | unsigned int changed; | |
360 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; | |
361 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; | |
362 | int assume_racy_is_modified = options & CE_MATCH_RACY_IS_DIRTY; | |
363 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; | |
364 | ||
365 | if (!ignore_fsmonitor) | |
366 | refresh_fsmonitor(istate); | |
367 | /* | |
368 | * If it's marked as always valid in the index, it's | |
369 | * valid whatever the checked-out copy says. | |
370 | * | |
371 | * skip-worktree has the same effect with higher precedence | |
372 | */ | |
373 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) | |
374 | return 0; | |
375 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) | |
376 | return 0; | |
377 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) | |
378 | return 0; | |
379 | ||
380 | /* | |
381 | * Intent-to-add entries have not been added, so the index entry | |
382 | * by definition never matches what is in the work tree until it | |
383 | * actually gets added. | |
384 | */ | |
385 | if (ce_intent_to_add(ce)) | |
386 | return DATA_CHANGED | TYPE_CHANGED | MODE_CHANGED; | |
387 | ||
388 | changed = ce_match_stat_basic(ce, st); | |
389 | ||
390 | /* | |
391 | * Within 1 second of this sequence: | |
392 | * echo xyzzy >file && git-update-index --add file | |
393 | * running this command: | |
394 | * echo frotz >file | |
395 | * would give a falsely clean cache entry. The mtime and | |
396 | * length match the cache, and other stat fields do not change. | |
397 | * | |
398 | * We could detect this at update-index time (the cache entry | |
399 | * being registered/updated records the same time as "now") | |
400 | * and delay the return from git-update-index, but that would | |
401 | * effectively mean we can make at most one commit per second, | |
402 | * which is not acceptable. Instead, we check cache entries | |
403 | * whose mtime are the same as the index file timestamp more | |
404 | * carefully than others. | |
405 | */ | |
406 | if (!changed && is_racy_timestamp(istate, ce)) { | |
407 | if (assume_racy_is_modified) | |
408 | changed |= DATA_CHANGED; | |
409 | else | |
410 | changed |= ce_modified_check_fs(ce, st); | |
411 | } | |
412 | ||
413 | return changed; | |
414 | } | |
415 | ||
416 | int ie_modified(struct index_state *istate, | |
417 | const struct cache_entry *ce, | |
418 | struct stat *st, unsigned int options) | |
419 | { | |
420 | int changed, changed_fs; | |
421 | ||
422 | changed = ie_match_stat(istate, ce, st, options); | |
423 | if (!changed) | |
424 | return 0; | |
425 | /* | |
426 | * If the mode or type has changed, there's no point in trying | |
427 | * to refresh the entry - it's not going to match | |
428 | */ | |
429 | if (changed & (MODE_CHANGED | TYPE_CHANGED)) | |
430 | return changed; | |
431 | ||
432 | /* | |
433 | * Immediately after read-tree or update-index --cacheinfo, | |
434 | * the length field is zero, as we have never even read the | |
435 | * lstat(2) information once, and we cannot trust DATA_CHANGED | |
436 | * returned by ie_match_stat() which in turn was returned by | |
437 | * ce_match_stat_basic() to signal that the filesize of the | |
438 | * blob changed. We have to actually go to the filesystem to | |
439 | * see if the contents match, and if so, should answer "unchanged". | |
440 | * | |
441 | * The logic does not apply to gitlinks, as ce_match_stat_basic() | |
442 | * already has checked the actual HEAD from the filesystem in the | |
443 | * subproject. If ie_match_stat() already said it is different, | |
444 | * then we know it is. | |
445 | */ | |
446 | if ((changed & DATA_CHANGED) && | |
447 | (S_ISGITLINK(ce->ce_mode) || ce->ce_stat_data.sd_size != 0)) | |
448 | return changed; | |
449 | ||
450 | changed_fs = ce_modified_check_fs(ce, st); | |
451 | if (changed_fs) | |
452 | return changed | changed_fs; | |
453 | return 0; | |
454 | } | |
455 | ||
456 | int base_name_compare(const char *name1, int len1, int mode1, | |
457 | const char *name2, int len2, int mode2) | |
458 | { | |
459 | unsigned char c1, c2; | |
460 | int len = len1 < len2 ? len1 : len2; | |
461 | int cmp; | |
462 | ||
463 | cmp = memcmp(name1, name2, len); | |
464 | if (cmp) | |
465 | return cmp; | |
466 | c1 = name1[len]; | |
467 | c2 = name2[len]; | |
468 | if (!c1 && S_ISDIR(mode1)) | |
469 | c1 = '/'; | |
470 | if (!c2 && S_ISDIR(mode2)) | |
471 | c2 = '/'; | |
472 | return (c1 < c2) ? -1 : (c1 > c2) ? 1 : 0; | |
473 | } | |
474 | ||
475 | /* | |
476 | * df_name_compare() is identical to base_name_compare(), except it | |
477 | * compares conflicting directory/file entries as equal. Note that | |
478 | * while a directory name compares as equal to a regular file, they | |
479 | * then individually compare _differently_ to a filename that has | |
480 | * a dot after the basename (because '\0' < '.' < '/'). | |
481 | * | |
482 | * This is used by routines that want to traverse the git namespace | |
483 | * but then handle conflicting entries together when possible. | |
484 | */ | |
485 | int df_name_compare(const char *name1, int len1, int mode1, | |
486 | const char *name2, int len2, int mode2) | |
487 | { | |
488 | int len = len1 < len2 ? len1 : len2, cmp; | |
489 | unsigned char c1, c2; | |
490 | ||
491 | cmp = memcmp(name1, name2, len); | |
492 | if (cmp) | |
493 | return cmp; | |
494 | /* Directories and files compare equal (same length, same name) */ | |
495 | if (len1 == len2) | |
496 | return 0; | |
497 | c1 = name1[len]; | |
498 | if (!c1 && S_ISDIR(mode1)) | |
499 | c1 = '/'; | |
500 | c2 = name2[len]; | |
501 | if (!c2 && S_ISDIR(mode2)) | |
502 | c2 = '/'; | |
503 | if (c1 == '/' && !c2) | |
504 | return 0; | |
505 | if (c2 == '/' && !c1) | |
506 | return 0; | |
507 | return c1 - c2; | |
508 | } | |
509 | ||
510 | int name_compare(const char *name1, size_t len1, const char *name2, size_t len2) | |
511 | { | |
512 | size_t min_len = (len1 < len2) ? len1 : len2; | |
513 | int cmp = memcmp(name1, name2, min_len); | |
514 | if (cmp) | |
515 | return cmp; | |
516 | if (len1 < len2) | |
517 | return -1; | |
518 | if (len1 > len2) | |
519 | return 1; | |
520 | return 0; | |
521 | } | |
522 | ||
523 | int cache_name_stage_compare(const char *name1, int len1, int stage1, const char *name2, int len2, int stage2) | |
524 | { | |
525 | int cmp; | |
526 | ||
527 | cmp = name_compare(name1, len1, name2, len2); | |
528 | if (cmp) | |
529 | return cmp; | |
530 | ||
531 | if (stage1 < stage2) | |
532 | return -1; | |
533 | if (stage1 > stage2) | |
534 | return 1; | |
535 | return 0; | |
536 | } | |
537 | ||
538 | static int index_name_stage_pos(const struct index_state *istate, const char *name, int namelen, int stage) | |
539 | { | |
540 | int first, last; | |
541 | ||
542 | first = 0; | |
543 | last = istate->cache_nr; | |
544 | while (last > first) { | |
545 | int next = (last + first) >> 1; | |
546 | struct cache_entry *ce = istate->cache[next]; | |
547 | int cmp = cache_name_stage_compare(name, namelen, stage, ce->name, ce_namelen(ce), ce_stage(ce)); | |
548 | if (!cmp) | |
549 | return next; | |
550 | if (cmp < 0) { | |
551 | last = next; | |
552 | continue; | |
553 | } | |
554 | first = next+1; | |
555 | } | |
556 | return -first-1; | |
557 | } | |
558 | ||
559 | int index_name_pos(const struct index_state *istate, const char *name, int namelen) | |
560 | { | |
561 | return index_name_stage_pos(istate, name, namelen, 0); | |
562 | } | |
563 | ||
564 | int remove_index_entry_at(struct index_state *istate, int pos) | |
565 | { | |
566 | struct cache_entry *ce = istate->cache[pos]; | |
567 | ||
568 | record_resolve_undo(istate, ce); | |
569 | remove_name_hash(istate, ce); | |
570 | save_or_free_index_entry(istate, ce); | |
571 | istate->cache_changed |= CE_ENTRY_REMOVED; | |
572 | istate->cache_nr--; | |
573 | if (pos >= istate->cache_nr) | |
574 | return 0; | |
575 | MOVE_ARRAY(istate->cache + pos, istate->cache + pos + 1, | |
576 | istate->cache_nr - pos); | |
577 | return 1; | |
578 | } | |
579 | ||
580 | /* | |
581 | * Remove all cache entries marked for removal, that is where | |
582 | * CE_REMOVE is set in ce_flags. This is much more effective than | |
583 | * calling remove_index_entry_at() for each entry to be removed. | |
584 | */ | |
585 | void remove_marked_cache_entries(struct index_state *istate) | |
586 | { | |
587 | struct cache_entry **ce_array = istate->cache; | |
588 | unsigned int i, j; | |
589 | ||
590 | for (i = j = 0; i < istate->cache_nr; i++) { | |
591 | if (ce_array[i]->ce_flags & CE_REMOVE) { | |
592 | remove_name_hash(istate, ce_array[i]); | |
593 | save_or_free_index_entry(istate, ce_array[i]); | |
594 | } | |
595 | else | |
596 | ce_array[j++] = ce_array[i]; | |
597 | } | |
598 | if (j == istate->cache_nr) | |
599 | return; | |
600 | istate->cache_changed |= CE_ENTRY_REMOVED; | |
601 | istate->cache_nr = j; | |
602 | } | |
603 | ||
604 | int remove_file_from_index(struct index_state *istate, const char *path) | |
605 | { | |
606 | int pos = index_name_pos(istate, path, strlen(path)); | |
607 | if (pos < 0) | |
608 | pos = -pos-1; | |
609 | cache_tree_invalidate_path(istate, path); | |
610 | untracked_cache_remove_from_index(istate, path); | |
611 | while (pos < istate->cache_nr && !strcmp(istate->cache[pos]->name, path)) | |
612 | remove_index_entry_at(istate, pos); | |
613 | return 0; | |
614 | } | |
615 | ||
616 | static int compare_name(struct cache_entry *ce, const char *path, int namelen) | |
617 | { | |
618 | return namelen != ce_namelen(ce) || memcmp(path, ce->name, namelen); | |
619 | } | |
620 | ||
621 | static int index_name_pos_also_unmerged(struct index_state *istate, | |
622 | const char *path, int namelen) | |
623 | { | |
624 | int pos = index_name_pos(istate, path, namelen); | |
625 | struct cache_entry *ce; | |
626 | ||
627 | if (pos >= 0) | |
628 | return pos; | |
629 | ||
630 | /* maybe unmerged? */ | |
631 | pos = -1 - pos; | |
632 | if (pos >= istate->cache_nr || | |
633 | compare_name((ce = istate->cache[pos]), path, namelen)) | |
634 | return -1; | |
635 | ||
636 | /* order of preference: stage 2, 1, 3 */ | |
637 | if (ce_stage(ce) == 1 && pos + 1 < istate->cache_nr && | |
638 | ce_stage((ce = istate->cache[pos + 1])) == 2 && | |
639 | !compare_name(ce, path, namelen)) | |
640 | pos++; | |
641 | return pos; | |
642 | } | |
643 | ||
644 | static int different_name(struct cache_entry *ce, struct cache_entry *alias) | |
645 | { | |
646 | int len = ce_namelen(ce); | |
647 | return ce_namelen(alias) != len || memcmp(ce->name, alias->name, len); | |
648 | } | |
649 | ||
650 | /* | |
651 | * If we add a filename that aliases in the cache, we will use the | |
652 | * name that we already have - but we don't want to update the same | |
653 | * alias twice, because that implies that there were actually two | |
654 | * different files with aliasing names! | |
655 | * | |
656 | * So we use the CE_ADDED flag to verify that the alias was an old | |
657 | * one before we accept it as | |
658 | */ | |
659 | static struct cache_entry *create_alias_ce(struct index_state *istate, | |
660 | struct cache_entry *ce, | |
661 | struct cache_entry *alias) | |
662 | { | |
663 | int len; | |
664 | struct cache_entry *new_entry; | |
665 | ||
666 | if (alias->ce_flags & CE_ADDED) | |
667 | die("Will not add file alias '%s' ('%s' already exists in index)", ce->name, alias->name); | |
668 | ||
669 | /* Ok, create the new entry using the name of the existing alias */ | |
670 | len = ce_namelen(alias); | |
671 | new_entry = make_empty_cache_entry(istate, len); | |
672 | memcpy(new_entry->name, alias->name, len); | |
673 | copy_cache_entry(new_entry, ce); | |
674 | save_or_free_index_entry(istate, ce); | |
675 | return new_entry; | |
676 | } | |
677 | ||
678 | void set_object_name_for_intent_to_add_entry(struct cache_entry *ce) | |
679 | { | |
680 | struct object_id oid; | |
681 | if (write_object_file("", 0, blob_type, &oid)) | |
682 | die("cannot create an empty blob in the object database"); | |
683 | oidcpy(&ce->oid, &oid); | |
684 | } | |
685 | ||
686 | int add_to_index(struct index_state *istate, const char *path, struct stat *st, int flags) | |
687 | { | |
688 | int namelen, was_same; | |
689 | mode_t st_mode = st->st_mode; | |
690 | struct cache_entry *ce, *alias = NULL; | |
691 | unsigned ce_option = CE_MATCH_IGNORE_VALID|CE_MATCH_IGNORE_SKIP_WORKTREE|CE_MATCH_RACY_IS_DIRTY; | |
692 | int verbose = flags & (ADD_CACHE_VERBOSE | ADD_CACHE_PRETEND); | |
693 | int pretend = flags & ADD_CACHE_PRETEND; | |
694 | int intent_only = flags & ADD_CACHE_INTENT; | |
695 | int add_option = (ADD_CACHE_OK_TO_ADD|ADD_CACHE_OK_TO_REPLACE| | |
696 | (intent_only ? ADD_CACHE_NEW_ONLY : 0)); | |
697 | int newflags = HASH_WRITE_OBJECT; | |
698 | ||
699 | if (flags & HASH_RENORMALIZE) | |
700 | newflags |= HASH_RENORMALIZE; | |
701 | ||
702 | if (!S_ISREG(st_mode) && !S_ISLNK(st_mode) && !S_ISDIR(st_mode)) | |
703 | return error("%s: can only add regular files, symbolic links or git-directories", path); | |
704 | ||
705 | namelen = strlen(path); | |
706 | if (S_ISDIR(st_mode)) { | |
707 | while (namelen && path[namelen-1] == '/') | |
708 | namelen--; | |
709 | } | |
710 | ce = make_empty_cache_entry(istate, namelen); | |
711 | memcpy(ce->name, path, namelen); | |
712 | ce->ce_namelen = namelen; | |
713 | if (!intent_only) | |
714 | fill_stat_cache_info(ce, st); | |
715 | else | |
716 | ce->ce_flags |= CE_INTENT_TO_ADD; | |
717 | ||
718 | ||
719 | if (trust_executable_bit && has_symlinks) { | |
720 | ce->ce_mode = create_ce_mode(st_mode); | |
721 | } else { | |
722 | /* If there is an existing entry, pick the mode bits and type | |
723 | * from it, otherwise assume unexecutable regular file. | |
724 | */ | |
725 | struct cache_entry *ent; | |
726 | int pos = index_name_pos_also_unmerged(istate, path, namelen); | |
727 | ||
728 | ent = (0 <= pos) ? istate->cache[pos] : NULL; | |
729 | ce->ce_mode = ce_mode_from_stat(ent, st_mode); | |
730 | } | |
731 | ||
732 | /* When core.ignorecase=true, determine if a directory of the same name but differing | |
733 | * case already exists within the Git repository. If it does, ensure the directory | |
734 | * case of the file being added to the repository matches (is folded into) the existing | |
735 | * entry's directory case. | |
736 | */ | |
737 | if (ignore_case) { | |
738 | adjust_dirname_case(istate, ce->name); | |
739 | } | |
740 | if (!(flags & HASH_RENORMALIZE)) { | |
741 | alias = index_file_exists(istate, ce->name, | |
742 | ce_namelen(ce), ignore_case); | |
743 | if (alias && | |
744 | !ce_stage(alias) && | |
745 | !ie_match_stat(istate, alias, st, ce_option)) { | |
746 | /* Nothing changed, really */ | |
747 | if (!S_ISGITLINK(alias->ce_mode)) | |
748 | ce_mark_uptodate(alias); | |
749 | alias->ce_flags |= CE_ADDED; | |
750 | ||
751 | discard_cache_entry(ce); | |
752 | return 0; | |
753 | } | |
754 | } | |
755 | if (!intent_only) { | |
756 | if (index_path(&ce->oid, path, st, newflags)) { | |
757 | discard_cache_entry(ce); | |
758 | return error("unable to index file %s", path); | |
759 | } | |
760 | } else | |
761 | set_object_name_for_intent_to_add_entry(ce); | |
762 | ||
763 | if (ignore_case && alias && different_name(ce, alias)) | |
764 | ce = create_alias_ce(istate, ce, alias); | |
765 | ce->ce_flags |= CE_ADDED; | |
766 | ||
767 | /* It was suspected to be racily clean, but it turns out to be Ok */ | |
768 | was_same = (alias && | |
769 | !ce_stage(alias) && | |
770 | oideq(&alias->oid, &ce->oid) && | |
771 | ce->ce_mode == alias->ce_mode); | |
772 | ||
773 | if (pretend) | |
774 | discard_cache_entry(ce); | |
775 | else if (add_index_entry(istate, ce, add_option)) { | |
776 | discard_cache_entry(ce); | |
777 | return error("unable to add %s to index", path); | |
778 | } | |
779 | if (verbose && !was_same) | |
780 | printf("add '%s'\n", path); | |
781 | return 0; | |
782 | } | |
783 | ||
784 | int add_file_to_index(struct index_state *istate, const char *path, int flags) | |
785 | { | |
786 | struct stat st; | |
787 | if (lstat(path, &st)) | |
788 | die_errno("unable to stat '%s'", path); | |
789 | return add_to_index(istate, path, &st, flags); | |
790 | } | |
791 | ||
792 | struct cache_entry *make_empty_cache_entry(struct index_state *istate, size_t len) | |
793 | { | |
794 | return mem_pool__ce_calloc(find_mem_pool(istate), len); | |
795 | } | |
796 | ||
797 | struct cache_entry *make_empty_transient_cache_entry(size_t len) | |
798 | { | |
799 | return xcalloc(1, cache_entry_size(len)); | |
800 | } | |
801 | ||
802 | struct cache_entry *make_cache_entry(struct index_state *istate, | |
803 | unsigned int mode, | |
804 | const struct object_id *oid, | |
805 | const char *path, | |
806 | int stage, | |
807 | unsigned int refresh_options) | |
808 | { | |
809 | struct cache_entry *ce, *ret; | |
810 | int len; | |
811 | ||
812 | if (!verify_path(path, mode)) { | |
813 | error("Invalid path '%s'", path); | |
814 | return NULL; | |
815 | } | |
816 | ||
817 | len = strlen(path); | |
818 | ce = make_empty_cache_entry(istate, len); | |
819 | ||
820 | oidcpy(&ce->oid, oid); | |
821 | memcpy(ce->name, path, len); | |
822 | ce->ce_flags = create_ce_flags(stage); | |
823 | ce->ce_namelen = len; | |
824 | ce->ce_mode = create_ce_mode(mode); | |
825 | ||
826 | ret = refresh_cache_entry(&the_index, ce, refresh_options); | |
827 | if (ret != ce) | |
828 | discard_cache_entry(ce); | |
829 | return ret; | |
830 | } | |
831 | ||
832 | struct cache_entry *make_transient_cache_entry(unsigned int mode, const struct object_id *oid, | |
833 | const char *path, int stage) | |
834 | { | |
835 | struct cache_entry *ce; | |
836 | int len; | |
837 | ||
838 | if (!verify_path(path, mode)) { | |
839 | error("Invalid path '%s'", path); | |
840 | return NULL; | |
841 | } | |
842 | ||
843 | len = strlen(path); | |
844 | ce = make_empty_transient_cache_entry(len); | |
845 | ||
846 | oidcpy(&ce->oid, oid); | |
847 | memcpy(ce->name, path, len); | |
848 | ce->ce_flags = create_ce_flags(stage); | |
849 | ce->ce_namelen = len; | |
850 | ce->ce_mode = create_ce_mode(mode); | |
851 | ||
852 | return ce; | |
853 | } | |
854 | ||
855 | /* | |
856 | * Chmod an index entry with either +x or -x. | |
857 | * | |
858 | * Returns -1 if the chmod for the particular cache entry failed (if it's | |
859 | * not a regular file), -2 if an invalid flip argument is passed in, 0 | |
860 | * otherwise. | |
861 | */ | |
862 | int chmod_index_entry(struct index_state *istate, struct cache_entry *ce, | |
863 | char flip) | |
864 | { | |
865 | if (!S_ISREG(ce->ce_mode)) | |
866 | return -1; | |
867 | switch (flip) { | |
868 | case '+': | |
869 | ce->ce_mode |= 0111; | |
870 | break; | |
871 | case '-': | |
872 | ce->ce_mode &= ~0111; | |
873 | break; | |
874 | default: | |
875 | return -2; | |
876 | } | |
877 | cache_tree_invalidate_path(istate, ce->name); | |
878 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
879 | mark_fsmonitor_invalid(istate, ce); | |
880 | istate->cache_changed |= CE_ENTRY_CHANGED; | |
881 | ||
882 | return 0; | |
883 | } | |
884 | ||
885 | int ce_same_name(const struct cache_entry *a, const struct cache_entry *b) | |
886 | { | |
887 | int len = ce_namelen(a); | |
888 | return ce_namelen(b) == len && !memcmp(a->name, b->name, len); | |
889 | } | |
890 | ||
891 | /* | |
892 | * We fundamentally don't like some paths: we don't want | |
893 | * dot or dot-dot anywhere, and for obvious reasons don't | |
894 | * want to recurse into ".git" either. | |
895 | * | |
896 | * Also, we don't want double slashes or slashes at the | |
897 | * end that can make pathnames ambiguous. | |
898 | */ | |
899 | static int verify_dotfile(const char *rest, unsigned mode) | |
900 | { | |
901 | /* | |
902 | * The first character was '.', but that | |
903 | * has already been discarded, we now test | |
904 | * the rest. | |
905 | */ | |
906 | ||
907 | /* "." is not allowed */ | |
908 | if (*rest == '\0' || is_dir_sep(*rest)) | |
909 | return 0; | |
910 | ||
911 | switch (*rest) { | |
912 | /* | |
913 | * ".git" followed by NUL or slash is bad. Note that we match | |
914 | * case-insensitively here, even if ignore_case is not set. | |
915 | * This outlaws ".GIT" everywhere out of an abundance of caution, | |
916 | * since there's really no good reason to allow it. | |
917 | * | |
918 | * Once we've seen ".git", we can also find ".gitmodules", etc (also | |
919 | * case-insensitively). | |
920 | */ | |
921 | case 'g': | |
922 | case 'G': | |
923 | if (rest[1] != 'i' && rest[1] != 'I') | |
924 | break; | |
925 | if (rest[2] != 't' && rest[2] != 'T') | |
926 | break; | |
927 | if (rest[3] == '\0' || is_dir_sep(rest[3])) | |
928 | return 0; | |
929 | if (S_ISLNK(mode)) { | |
930 | rest += 3; | |
931 | if (skip_iprefix(rest, "modules", &rest) && | |
932 | (*rest == '\0' || is_dir_sep(*rest))) | |
933 | return 0; | |
934 | } | |
935 | break; | |
936 | case '.': | |
937 | if (rest[1] == '\0' || is_dir_sep(rest[1])) | |
938 | return 0; | |
939 | } | |
940 | return 1; | |
941 | } | |
942 | ||
943 | int verify_path(const char *path, unsigned mode) | |
944 | { | |
945 | char c; | |
946 | ||
947 | if (has_dos_drive_prefix(path)) | |
948 | return 0; | |
949 | ||
950 | goto inside; | |
951 | for (;;) { | |
952 | if (!c) | |
953 | return 1; | |
954 | if (is_dir_sep(c)) { | |
955 | inside: | |
956 | if (protect_hfs) { | |
957 | if (is_hfs_dotgit(path)) | |
958 | return 0; | |
959 | if (S_ISLNK(mode)) { | |
960 | if (is_hfs_dotgitmodules(path)) | |
961 | return 0; | |
962 | } | |
963 | } | |
964 | if (protect_ntfs) { | |
965 | if (is_ntfs_dotgit(path)) | |
966 | return 0; | |
967 | if (S_ISLNK(mode)) { | |
968 | if (is_ntfs_dotgitmodules(path)) | |
969 | return 0; | |
970 | } | |
971 | } | |
972 | ||
973 | c = *path++; | |
974 | if ((c == '.' && !verify_dotfile(path, mode)) || | |
975 | is_dir_sep(c) || c == '\0') | |
976 | return 0; | |
977 | } | |
978 | c = *path++; | |
979 | } | |
980 | } | |
981 | ||
982 | /* | |
983 | * Do we have another file that has the beginning components being a | |
984 | * proper superset of the name we're trying to add? | |
985 | */ | |
986 | static int has_file_name(struct index_state *istate, | |
987 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
988 | { | |
989 | int retval = 0; | |
990 | int len = ce_namelen(ce); | |
991 | int stage = ce_stage(ce); | |
992 | const char *name = ce->name; | |
993 | ||
994 | while (pos < istate->cache_nr) { | |
995 | struct cache_entry *p = istate->cache[pos++]; | |
996 | ||
997 | if (len >= ce_namelen(p)) | |
998 | break; | |
999 | if (memcmp(name, p->name, len)) | |
1000 | break; | |
1001 | if (ce_stage(p) != stage) | |
1002 | continue; | |
1003 | if (p->name[len] != '/') | |
1004 | continue; | |
1005 | if (p->ce_flags & CE_REMOVE) | |
1006 | continue; | |
1007 | retval = -1; | |
1008 | if (!ok_to_replace) | |
1009 | break; | |
1010 | remove_index_entry_at(istate, --pos); | |
1011 | } | |
1012 | return retval; | |
1013 | } | |
1014 | ||
1015 | ||
1016 | /* | |
1017 | * Like strcmp(), but also return the offset of the first change. | |
1018 | * If strings are equal, return the length. | |
1019 | */ | |
1020 | int strcmp_offset(const char *s1, const char *s2, size_t *first_change) | |
1021 | { | |
1022 | size_t k; | |
1023 | ||
1024 | if (!first_change) | |
1025 | return strcmp(s1, s2); | |
1026 | ||
1027 | for (k = 0; s1[k] == s2[k]; k++) | |
1028 | if (s1[k] == '\0') | |
1029 | break; | |
1030 | ||
1031 | *first_change = k; | |
1032 | return (unsigned char)s1[k] - (unsigned char)s2[k]; | |
1033 | } | |
1034 | ||
1035 | /* | |
1036 | * Do we have another file with a pathname that is a proper | |
1037 | * subset of the name we're trying to add? | |
1038 | * | |
1039 | * That is, is there another file in the index with a path | |
1040 | * that matches a sub-directory in the given entry? | |
1041 | */ | |
1042 | static int has_dir_name(struct index_state *istate, | |
1043 | const struct cache_entry *ce, int pos, int ok_to_replace) | |
1044 | { | |
1045 | int retval = 0; | |
1046 | int stage = ce_stage(ce); | |
1047 | const char *name = ce->name; | |
1048 | const char *slash = name + ce_namelen(ce); | |
1049 | size_t len_eq_last; | |
1050 | int cmp_last = 0; | |
1051 | ||
1052 | /* | |
1053 | * We are frequently called during an iteration on a sorted | |
1054 | * list of pathnames and while building a new index. Therefore, | |
1055 | * there is a high probability that this entry will eventually | |
1056 | * be appended to the index, rather than inserted in the middle. | |
1057 | * If we can confirm that, we can avoid binary searches on the | |
1058 | * components of the pathname. | |
1059 | * | |
1060 | * Compare the entry's full path with the last path in the index. | |
1061 | */ | |
1062 | if (istate->cache_nr > 0) { | |
1063 | cmp_last = strcmp_offset(name, | |
1064 | istate->cache[istate->cache_nr - 1]->name, | |
1065 | &len_eq_last); | |
1066 | if (cmp_last > 0) { | |
1067 | if (len_eq_last == 0) { | |
1068 | /* | |
1069 | * The entry sorts AFTER the last one in the | |
1070 | * index and their paths have no common prefix, | |
1071 | * so there cannot be a F/D conflict. | |
1072 | */ | |
1073 | return retval; | |
1074 | } else { | |
1075 | /* | |
1076 | * The entry sorts AFTER the last one in the | |
1077 | * index, but has a common prefix. Fall through | |
1078 | * to the loop below to disect the entry's path | |
1079 | * and see where the difference is. | |
1080 | */ | |
1081 | } | |
1082 | } else if (cmp_last == 0) { | |
1083 | /* | |
1084 | * The entry exactly matches the last one in the | |
1085 | * index, but because of multiple stage and CE_REMOVE | |
1086 | * items, we fall through and let the regular search | |
1087 | * code handle it. | |
1088 | */ | |
1089 | } | |
1090 | } | |
1091 | ||
1092 | for (;;) { | |
1093 | size_t len; | |
1094 | ||
1095 | for (;;) { | |
1096 | if (*--slash == '/') | |
1097 | break; | |
1098 | if (slash <= ce->name) | |
1099 | return retval; | |
1100 | } | |
1101 | len = slash - name; | |
1102 | ||
1103 | if (cmp_last > 0) { | |
1104 | /* | |
1105 | * (len + 1) is a directory boundary (including | |
1106 | * the trailing slash). And since the loop is | |
1107 | * decrementing "slash", the first iteration is | |
1108 | * the longest directory prefix; subsequent | |
1109 | * iterations consider parent directories. | |
1110 | */ | |
1111 | ||
1112 | if (len + 1 <= len_eq_last) { | |
1113 | /* | |
1114 | * The directory prefix (including the trailing | |
1115 | * slash) also appears as a prefix in the last | |
1116 | * entry, so the remainder cannot collide (because | |
1117 | * strcmp said the whole path was greater). | |
1118 | * | |
1119 | * EQ: last: xxx/A | |
1120 | * this: xxx/B | |
1121 | * | |
1122 | * LT: last: xxx/file_A | |
1123 | * this: xxx/file_B | |
1124 | */ | |
1125 | return retval; | |
1126 | } | |
1127 | ||
1128 | if (len > len_eq_last) { | |
1129 | /* | |
1130 | * This part of the directory prefix (excluding | |
1131 | * the trailing slash) is longer than the known | |
1132 | * equal portions, so this sub-directory cannot | |
1133 | * collide with a file. | |
1134 | * | |
1135 | * GT: last: xxxA | |
1136 | * this: xxxB/file | |
1137 | */ | |
1138 | return retval; | |
1139 | } | |
1140 | ||
1141 | if (istate->cache_nr > 0 && | |
1142 | ce_namelen(istate->cache[istate->cache_nr - 1]) > len) { | |
1143 | /* | |
1144 | * The directory prefix lines up with part of | |
1145 | * a longer file or directory name, but sorts | |
1146 | * after it, so this sub-directory cannot | |
1147 | * collide with a file. | |
1148 | * | |
1149 | * last: xxx/yy-file (because '-' sorts before '/') | |
1150 | * this: xxx/yy/abc | |
1151 | */ | |
1152 | return retval; | |
1153 | } | |
1154 | ||
1155 | /* | |
1156 | * This is a possible collision. Fall through and | |
1157 | * let the regular search code handle it. | |
1158 | * | |
1159 | * last: xxx | |
1160 | * this: xxx/file | |
1161 | */ | |
1162 | } | |
1163 | ||
1164 | pos = index_name_stage_pos(istate, name, len, stage); | |
1165 | if (pos >= 0) { | |
1166 | /* | |
1167 | * Found one, but not so fast. This could | |
1168 | * be a marker that says "I was here, but | |
1169 | * I am being removed". Such an entry is | |
1170 | * not a part of the resulting tree, and | |
1171 | * it is Ok to have a directory at the same | |
1172 | * path. | |
1173 | */ | |
1174 | if (!(istate->cache[pos]->ce_flags & CE_REMOVE)) { | |
1175 | retval = -1; | |
1176 | if (!ok_to_replace) | |
1177 | break; | |
1178 | remove_index_entry_at(istate, pos); | |
1179 | continue; | |
1180 | } | |
1181 | } | |
1182 | else | |
1183 | pos = -pos-1; | |
1184 | ||
1185 | /* | |
1186 | * Trivial optimization: if we find an entry that | |
1187 | * already matches the sub-directory, then we know | |
1188 | * we're ok, and we can exit. | |
1189 | */ | |
1190 | while (pos < istate->cache_nr) { | |
1191 | struct cache_entry *p = istate->cache[pos]; | |
1192 | if ((ce_namelen(p) <= len) || | |
1193 | (p->name[len] != '/') || | |
1194 | memcmp(p->name, name, len)) | |
1195 | break; /* not our subdirectory */ | |
1196 | if (ce_stage(p) == stage && !(p->ce_flags & CE_REMOVE)) | |
1197 | /* | |
1198 | * p is at the same stage as our entry, and | |
1199 | * is a subdirectory of what we are looking | |
1200 | * at, so we cannot have conflicts at our | |
1201 | * level or anything shorter. | |
1202 | */ | |
1203 | return retval; | |
1204 | pos++; | |
1205 | } | |
1206 | } | |
1207 | return retval; | |
1208 | } | |
1209 | ||
1210 | /* We may be in a situation where we already have path/file and path | |
1211 | * is being added, or we already have path and path/file is being | |
1212 | * added. Either one would result in a nonsense tree that has path | |
1213 | * twice when git-write-tree tries to write it out. Prevent it. | |
1214 | * | |
1215 | * If ok-to-replace is specified, we remove the conflicting entries | |
1216 | * from the cache so the caller should recompute the insert position. | |
1217 | * When this happens, we return non-zero. | |
1218 | */ | |
1219 | static int check_file_directory_conflict(struct index_state *istate, | |
1220 | const struct cache_entry *ce, | |
1221 | int pos, int ok_to_replace) | |
1222 | { | |
1223 | int retval; | |
1224 | ||
1225 | /* | |
1226 | * When ce is an "I am going away" entry, we allow it to be added | |
1227 | */ | |
1228 | if (ce->ce_flags & CE_REMOVE) | |
1229 | return 0; | |
1230 | ||
1231 | /* | |
1232 | * We check if the path is a sub-path of a subsequent pathname | |
1233 | * first, since removing those will not change the position | |
1234 | * in the array. | |
1235 | */ | |
1236 | retval = has_file_name(istate, ce, pos, ok_to_replace); | |
1237 | ||
1238 | /* | |
1239 | * Then check if the path might have a clashing sub-directory | |
1240 | * before it. | |
1241 | */ | |
1242 | return retval + has_dir_name(istate, ce, pos, ok_to_replace); | |
1243 | } | |
1244 | ||
1245 | static int add_index_entry_with_check(struct index_state *istate, struct cache_entry *ce, int option) | |
1246 | { | |
1247 | int pos; | |
1248 | int ok_to_add = option & ADD_CACHE_OK_TO_ADD; | |
1249 | int ok_to_replace = option & ADD_CACHE_OK_TO_REPLACE; | |
1250 | int skip_df_check = option & ADD_CACHE_SKIP_DFCHECK; | |
1251 | int new_only = option & ADD_CACHE_NEW_ONLY; | |
1252 | ||
1253 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) | |
1254 | cache_tree_invalidate_path(istate, ce->name); | |
1255 | ||
1256 | /* | |
1257 | * If this entry's path sorts after the last entry in the index, | |
1258 | * we can avoid searching for it. | |
1259 | */ | |
1260 | if (istate->cache_nr > 0 && | |
1261 | strcmp(ce->name, istate->cache[istate->cache_nr - 1]->name) > 0) | |
1262 | pos = -istate->cache_nr - 1; | |
1263 | else | |
1264 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce)); | |
1265 | ||
1266 | /* existing match? Just replace it. */ | |
1267 | if (pos >= 0) { | |
1268 | if (!new_only) | |
1269 | replace_index_entry(istate, pos, ce); | |
1270 | return 0; | |
1271 | } | |
1272 | pos = -pos-1; | |
1273 | ||
1274 | if (!(option & ADD_CACHE_KEEP_CACHE_TREE)) | |
1275 | untracked_cache_add_to_index(istate, ce->name); | |
1276 | ||
1277 | /* | |
1278 | * Inserting a merged entry ("stage 0") into the index | |
1279 | * will always replace all non-merged entries.. | |
1280 | */ | |
1281 | if (pos < istate->cache_nr && ce_stage(ce) == 0) { | |
1282 | while (ce_same_name(istate->cache[pos], ce)) { | |
1283 | ok_to_add = 1; | |
1284 | if (!remove_index_entry_at(istate, pos)) | |
1285 | break; | |
1286 | } | |
1287 | } | |
1288 | ||
1289 | if (!ok_to_add) | |
1290 | return -1; | |
1291 | if (!verify_path(ce->name, ce->ce_mode)) | |
1292 | return error("Invalid path '%s'", ce->name); | |
1293 | ||
1294 | if (!skip_df_check && | |
1295 | check_file_directory_conflict(istate, ce, pos, ok_to_replace)) { | |
1296 | if (!ok_to_replace) | |
1297 | return error("'%s' appears as both a file and as a directory", | |
1298 | ce->name); | |
1299 | pos = index_name_stage_pos(istate, ce->name, ce_namelen(ce), ce_stage(ce)); | |
1300 | pos = -pos-1; | |
1301 | } | |
1302 | return pos + 1; | |
1303 | } | |
1304 | ||
1305 | int add_index_entry(struct index_state *istate, struct cache_entry *ce, int option) | |
1306 | { | |
1307 | int pos; | |
1308 | ||
1309 | if (option & ADD_CACHE_JUST_APPEND) | |
1310 | pos = istate->cache_nr; | |
1311 | else { | |
1312 | int ret; | |
1313 | ret = add_index_entry_with_check(istate, ce, option); | |
1314 | if (ret <= 0) | |
1315 | return ret; | |
1316 | pos = ret - 1; | |
1317 | } | |
1318 | ||
1319 | /* Make sure the array is big enough .. */ | |
1320 | ALLOC_GROW(istate->cache, istate->cache_nr + 1, istate->cache_alloc); | |
1321 | ||
1322 | /* Add it in.. */ | |
1323 | istate->cache_nr++; | |
1324 | if (istate->cache_nr > pos + 1) | |
1325 | MOVE_ARRAY(istate->cache + pos + 1, istate->cache + pos, | |
1326 | istate->cache_nr - pos - 1); | |
1327 | set_index_entry(istate, pos, ce); | |
1328 | istate->cache_changed |= CE_ENTRY_ADDED; | |
1329 | return 0; | |
1330 | } | |
1331 | ||
1332 | /* | |
1333 | * "refresh" does not calculate a new sha1 file or bring the | |
1334 | * cache up-to-date for mode/content changes. But what it | |
1335 | * _does_ do is to "re-match" the stat information of a file | |
1336 | * with the cache, so that you can refresh the cache for a | |
1337 | * file that hasn't been changed but where the stat entry is | |
1338 | * out of date. | |
1339 | * | |
1340 | * For example, you'd want to do this after doing a "git-read-tree", | |
1341 | * to link up the stat cache details with the proper files. | |
1342 | */ | |
1343 | static struct cache_entry *refresh_cache_ent(struct index_state *istate, | |
1344 | struct cache_entry *ce, | |
1345 | unsigned int options, int *err, | |
1346 | int *changed_ret) | |
1347 | { | |
1348 | struct stat st; | |
1349 | struct cache_entry *updated; | |
1350 | int changed; | |
1351 | int refresh = options & CE_MATCH_REFRESH; | |
1352 | int ignore_valid = options & CE_MATCH_IGNORE_VALID; | |
1353 | int ignore_skip_worktree = options & CE_MATCH_IGNORE_SKIP_WORKTREE; | |
1354 | int ignore_missing = options & CE_MATCH_IGNORE_MISSING; | |
1355 | int ignore_fsmonitor = options & CE_MATCH_IGNORE_FSMONITOR; | |
1356 | ||
1357 | if (!refresh || ce_uptodate(ce)) | |
1358 | return ce; | |
1359 | ||
1360 | if (!ignore_fsmonitor) | |
1361 | refresh_fsmonitor(istate); | |
1362 | /* | |
1363 | * CE_VALID or CE_SKIP_WORKTREE means the user promised us | |
1364 | * that the change to the work tree does not matter and told | |
1365 | * us not to worry. | |
1366 | */ | |
1367 | if (!ignore_skip_worktree && ce_skip_worktree(ce)) { | |
1368 | ce_mark_uptodate(ce); | |
1369 | return ce; | |
1370 | } | |
1371 | if (!ignore_valid && (ce->ce_flags & CE_VALID)) { | |
1372 | ce_mark_uptodate(ce); | |
1373 | return ce; | |
1374 | } | |
1375 | if (!ignore_fsmonitor && (ce->ce_flags & CE_FSMONITOR_VALID)) { | |
1376 | ce_mark_uptodate(ce); | |
1377 | return ce; | |
1378 | } | |
1379 | ||
1380 | if (has_symlink_leading_path(ce->name, ce_namelen(ce))) { | |
1381 | if (ignore_missing) | |
1382 | return ce; | |
1383 | if (err) | |
1384 | *err = ENOENT; | |
1385 | return NULL; | |
1386 | } | |
1387 | ||
1388 | if (lstat(ce->name, &st) < 0) { | |
1389 | if (ignore_missing && errno == ENOENT) | |
1390 | return ce; | |
1391 | if (err) | |
1392 | *err = errno; | |
1393 | return NULL; | |
1394 | } | |
1395 | ||
1396 | changed = ie_match_stat(istate, ce, &st, options); | |
1397 | if (changed_ret) | |
1398 | *changed_ret = changed; | |
1399 | if (!changed) { | |
1400 | /* | |
1401 | * The path is unchanged. If we were told to ignore | |
1402 | * valid bit, then we did the actual stat check and | |
1403 | * found that the entry is unmodified. If the entry | |
1404 | * is not marked VALID, this is the place to mark it | |
1405 | * valid again, under "assume unchanged" mode. | |
1406 | */ | |
1407 | if (ignore_valid && assume_unchanged && | |
1408 | !(ce->ce_flags & CE_VALID)) | |
1409 | ; /* mark this one VALID again */ | |
1410 | else { | |
1411 | /* | |
1412 | * We do not mark the index itself "modified" | |
1413 | * because CE_UPTODATE flag is in-core only; | |
1414 | * we are not going to write this change out. | |
1415 | */ | |
1416 | if (!S_ISGITLINK(ce->ce_mode)) { | |
1417 | ce_mark_uptodate(ce); | |
1418 | mark_fsmonitor_valid(ce); | |
1419 | } | |
1420 | return ce; | |
1421 | } | |
1422 | } | |
1423 | ||
1424 | if (ie_modified(istate, ce, &st, options)) { | |
1425 | if (err) | |
1426 | *err = EINVAL; | |
1427 | return NULL; | |
1428 | } | |
1429 | ||
1430 | updated = make_empty_cache_entry(istate, ce_namelen(ce)); | |
1431 | copy_cache_entry(updated, ce); | |
1432 | memcpy(updated->name, ce->name, ce->ce_namelen + 1); | |
1433 | fill_stat_cache_info(updated, &st); | |
1434 | /* | |
1435 | * If ignore_valid is not set, we should leave CE_VALID bit | |
1436 | * alone. Otherwise, paths marked with --no-assume-unchanged | |
1437 | * (i.e. things to be edited) will reacquire CE_VALID bit | |
1438 | * automatically, which is not really what we want. | |
1439 | */ | |
1440 | if (!ignore_valid && assume_unchanged && | |
1441 | !(ce->ce_flags & CE_VALID)) | |
1442 | updated->ce_flags &= ~CE_VALID; | |
1443 | ||
1444 | /* istate->cache_changed is updated in the caller */ | |
1445 | return updated; | |
1446 | } | |
1447 | ||
1448 | static void show_file(const char * fmt, const char * name, int in_porcelain, | |
1449 | int * first, const char *header_msg) | |
1450 | { | |
1451 | if (in_porcelain && *first && header_msg) { | |
1452 | printf("%s\n", header_msg); | |
1453 | *first = 0; | |
1454 | } | |
1455 | printf(fmt, name); | |
1456 | } | |
1457 | ||
1458 | int refresh_index(struct index_state *istate, unsigned int flags, | |
1459 | const struct pathspec *pathspec, | |
1460 | char *seen, const char *header_msg) | |
1461 | { | |
1462 | int i; | |
1463 | int has_errors = 0; | |
1464 | int really = (flags & REFRESH_REALLY) != 0; | |
1465 | int allow_unmerged = (flags & REFRESH_UNMERGED) != 0; | |
1466 | int quiet = (flags & REFRESH_QUIET) != 0; | |
1467 | int not_new = (flags & REFRESH_IGNORE_MISSING) != 0; | |
1468 | int ignore_submodules = (flags & REFRESH_IGNORE_SUBMODULES) != 0; | |
1469 | int first = 1; | |
1470 | int in_porcelain = (flags & REFRESH_IN_PORCELAIN); | |
1471 | unsigned int options = (CE_MATCH_REFRESH | | |
1472 | (really ? CE_MATCH_IGNORE_VALID : 0) | | |
1473 | (not_new ? CE_MATCH_IGNORE_MISSING : 0)); | |
1474 | const char *modified_fmt; | |
1475 | const char *deleted_fmt; | |
1476 | const char *typechange_fmt; | |
1477 | const char *added_fmt; | |
1478 | const char *unmerged_fmt; | |
1479 | ||
1480 | trace_performance_enter(); | |
1481 | modified_fmt = (in_porcelain ? "M\t%s\n" : "%s: needs update\n"); | |
1482 | deleted_fmt = (in_porcelain ? "D\t%s\n" : "%s: needs update\n"); | |
1483 | typechange_fmt = (in_porcelain ? "T\t%s\n" : "%s needs update\n"); | |
1484 | added_fmt = (in_porcelain ? "A\t%s\n" : "%s needs update\n"); | |
1485 | unmerged_fmt = (in_porcelain ? "U\t%s\n" : "%s: needs merge\n"); | |
1486 | for (i = 0; i < istate->cache_nr; i++) { | |
1487 | struct cache_entry *ce, *new_entry; | |
1488 | int cache_errno = 0; | |
1489 | int changed = 0; | |
1490 | int filtered = 0; | |
1491 | ||
1492 | ce = istate->cache[i]; | |
1493 | if (ignore_submodules && S_ISGITLINK(ce->ce_mode)) | |
1494 | continue; | |
1495 | ||
1496 | if (pathspec && !ce_path_match(&the_index, ce, pathspec, seen)) | |
1497 | filtered = 1; | |
1498 | ||
1499 | if (ce_stage(ce)) { | |
1500 | while ((i < istate->cache_nr) && | |
1501 | ! strcmp(istate->cache[i]->name, ce->name)) | |
1502 | i++; | |
1503 | i--; | |
1504 | if (allow_unmerged) | |
1505 | continue; | |
1506 | if (!filtered) | |
1507 | show_file(unmerged_fmt, ce->name, in_porcelain, | |
1508 | &first, header_msg); | |
1509 | has_errors = 1; | |
1510 | continue; | |
1511 | } | |
1512 | ||
1513 | if (filtered) | |
1514 | continue; | |
1515 | ||
1516 | new_entry = refresh_cache_ent(istate, ce, options, &cache_errno, &changed); | |
1517 | if (new_entry == ce) | |
1518 | continue; | |
1519 | if (!new_entry) { | |
1520 | const char *fmt; | |
1521 | ||
1522 | if (really && cache_errno == EINVAL) { | |
1523 | /* If we are doing --really-refresh that | |
1524 | * means the index is not valid anymore. | |
1525 | */ | |
1526 | ce->ce_flags &= ~CE_VALID; | |
1527 | ce->ce_flags |= CE_UPDATE_IN_BASE; | |
1528 | mark_fsmonitor_invalid(istate, ce); | |
1529 | istate->cache_changed |= CE_ENTRY_CHANGED; | |
1530 | } | |
1531 | if (quiet) | |
1532 | continue; | |
1533 | ||
1534 | if (cache_errno == ENOENT) | |
1535 | fmt = deleted_fmt; | |
1536 | else if (ce_intent_to_add(ce)) | |
1537 | fmt = added_fmt; /* must be before other checks */ | |
1538 | else if (changed & TYPE_CHANGED) | |
1539 | fmt = typechange_fmt; | |
1540 | else | |
1541 | fmt = modified_fmt; | |
1542 | show_file(fmt, | |
1543 | ce->name, in_porcelain, &first, header_msg); | |
1544 | has_errors = 1; | |
1545 | continue; | |
1546 | } | |
1547 | ||
1548 | replace_index_entry(istate, i, new_entry); | |
1549 | } | |
1550 | trace_performance_leave("refresh index"); | |
1551 | return has_errors; | |
1552 | } | |
1553 | ||
1554 | struct cache_entry *refresh_cache_entry(struct index_state *istate, | |
1555 | struct cache_entry *ce, | |
1556 | unsigned int options) | |
1557 | { | |
1558 | return refresh_cache_ent(istate, ce, options, NULL, NULL); | |
1559 | } | |
1560 | ||
1561 | ||
1562 | /***************************************************************** | |
1563 | * Index File I/O | |
1564 | *****************************************************************/ | |
1565 | ||
1566 | #define INDEX_FORMAT_DEFAULT 3 | |
1567 | ||
1568 | static unsigned int get_index_format_default(void) | |
1569 | { | |
1570 | char *envversion = getenv("GIT_INDEX_VERSION"); | |
1571 | char *endp; | |
1572 | int value; | |
1573 | unsigned int version = INDEX_FORMAT_DEFAULT; | |
1574 | ||
1575 | if (!envversion) { | |
1576 | if (!git_config_get_int("index.version", &value)) | |
1577 | version = value; | |
1578 | if (version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { | |
1579 | warning(_("index.version set, but the value is invalid.\n" | |
1580 | "Using version %i"), INDEX_FORMAT_DEFAULT); | |
1581 | return INDEX_FORMAT_DEFAULT; | |
1582 | } | |
1583 | return version; | |
1584 | } | |
1585 | ||
1586 | version = strtoul(envversion, &endp, 10); | |
1587 | if (*endp || | |
1588 | version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < version) { | |
1589 | warning(_("GIT_INDEX_VERSION set, but the value is invalid.\n" | |
1590 | "Using version %i"), INDEX_FORMAT_DEFAULT); | |
1591 | version = INDEX_FORMAT_DEFAULT; | |
1592 | } | |
1593 | return version; | |
1594 | } | |
1595 | ||
1596 | /* | |
1597 | * dev/ino/uid/gid/size are also just tracked to the low 32 bits | |
1598 | * Again - this is just a (very strong in practice) heuristic that | |
1599 | * the inode hasn't changed. | |
1600 | * | |
1601 | * We save the fields in big-endian order to allow using the | |
1602 | * index file over NFS transparently. | |
1603 | */ | |
1604 | struct ondisk_cache_entry { | |
1605 | struct cache_time ctime; | |
1606 | struct cache_time mtime; | |
1607 | uint32_t dev; | |
1608 | uint32_t ino; | |
1609 | uint32_t mode; | |
1610 | uint32_t uid; | |
1611 | uint32_t gid; | |
1612 | uint32_t size; | |
1613 | unsigned char sha1[20]; | |
1614 | uint16_t flags; | |
1615 | char name[FLEX_ARRAY]; /* more */ | |
1616 | }; | |
1617 | ||
1618 | /* | |
1619 | * This struct is used when CE_EXTENDED bit is 1 | |
1620 | * The struct must match ondisk_cache_entry exactly from | |
1621 | * ctime till flags | |
1622 | */ | |
1623 | struct ondisk_cache_entry_extended { | |
1624 | struct cache_time ctime; | |
1625 | struct cache_time mtime; | |
1626 | uint32_t dev; | |
1627 | uint32_t ino; | |
1628 | uint32_t mode; | |
1629 | uint32_t uid; | |
1630 | uint32_t gid; | |
1631 | uint32_t size; | |
1632 | unsigned char sha1[20]; | |
1633 | uint16_t flags; | |
1634 | uint16_t flags2; | |
1635 | char name[FLEX_ARRAY]; /* more */ | |
1636 | }; | |
1637 | ||
1638 | /* These are only used for v3 or lower */ | |
1639 | #define align_padding_size(size, len) ((size + (len) + 8) & ~7) - (size + len) | |
1640 | #define align_flex_name(STRUCT,len) ((offsetof(struct STRUCT,name) + (len) + 8) & ~7) | |
1641 | #define ondisk_cache_entry_size(len) align_flex_name(ondisk_cache_entry,len) | |
1642 | #define ondisk_cache_entry_extended_size(len) align_flex_name(ondisk_cache_entry_extended,len) | |
1643 | #define ondisk_ce_size(ce) (((ce)->ce_flags & CE_EXTENDED) ? \ | |
1644 | ondisk_cache_entry_extended_size(ce_namelen(ce)) : \ | |
1645 | ondisk_cache_entry_size(ce_namelen(ce))) | |
1646 | ||
1647 | /* Allow fsck to force verification of the index checksum. */ | |
1648 | int verify_index_checksum; | |
1649 | ||
1650 | /* Allow fsck to force verification of the cache entry order. */ | |
1651 | int verify_ce_order; | |
1652 | ||
1653 | static int verify_hdr(const struct cache_header *hdr, unsigned long size) | |
1654 | { | |
1655 | git_hash_ctx c; | |
1656 | unsigned char hash[GIT_MAX_RAWSZ]; | |
1657 | int hdr_version; | |
1658 | ||
1659 | if (hdr->hdr_signature != htonl(CACHE_SIGNATURE)) | |
1660 | return error("bad signature"); | |
1661 | hdr_version = ntohl(hdr->hdr_version); | |
1662 | if (hdr_version < INDEX_FORMAT_LB || INDEX_FORMAT_UB < hdr_version) | |
1663 | return error("bad index version %d", hdr_version); | |
1664 | ||
1665 | if (!verify_index_checksum) | |
1666 | return 0; | |
1667 | ||
1668 | the_hash_algo->init_fn(&c); | |
1669 | the_hash_algo->update_fn(&c, hdr, size - the_hash_algo->rawsz); | |
1670 | the_hash_algo->final_fn(hash, &c); | |
1671 | if (!hasheq(hash, (unsigned char *)hdr + size - the_hash_algo->rawsz)) | |
1672 | return error("bad index file sha1 signature"); | |
1673 | return 0; | |
1674 | } | |
1675 | ||
1676 | static int read_index_extension(struct index_state *istate, | |
1677 | const char *ext, const char *data, unsigned long sz) | |
1678 | { | |
1679 | switch (CACHE_EXT(ext)) { | |
1680 | case CACHE_EXT_TREE: | |
1681 | istate->cache_tree = cache_tree_read(data, sz); | |
1682 | break; | |
1683 | case CACHE_EXT_RESOLVE_UNDO: | |
1684 | istate->resolve_undo = resolve_undo_read(data, sz); | |
1685 | break; | |
1686 | case CACHE_EXT_LINK: | |
1687 | if (read_link_extension(istate, data, sz)) | |
1688 | return -1; | |
1689 | break; | |
1690 | case CACHE_EXT_UNTRACKED: | |
1691 | istate->untracked = read_untracked_extension(data, sz); | |
1692 | break; | |
1693 | case CACHE_EXT_FSMONITOR: | |
1694 | read_fsmonitor_extension(istate, data, sz); | |
1695 | break; | |
1696 | default: | |
1697 | if (*ext < 'A' || 'Z' < *ext) | |
1698 | return error("index uses %.4s extension, which we do not understand", | |
1699 | ext); | |
1700 | fprintf(stderr, "ignoring %.4s extension\n", ext); | |
1701 | break; | |
1702 | } | |
1703 | return 0; | |
1704 | } | |
1705 | ||
1706 | int hold_locked_index(struct lock_file *lk, int lock_flags) | |
1707 | { | |
1708 | return hold_lock_file_for_update(lk, get_index_file(), lock_flags); | |
1709 | } | |
1710 | ||
1711 | int read_index(struct index_state *istate) | |
1712 | { | |
1713 | return read_index_from(istate, get_index_file(), get_git_dir()); | |
1714 | } | |
1715 | ||
1716 | static struct cache_entry *create_from_disk(struct index_state *istate, | |
1717 | struct ondisk_cache_entry *ondisk, | |
1718 | unsigned long *ent_size, | |
1719 | const struct cache_entry *previous_ce) | |
1720 | { | |
1721 | struct cache_entry *ce; | |
1722 | size_t len; | |
1723 | const char *name; | |
1724 | unsigned int flags; | |
1725 | size_t copy_len; | |
1726 | /* | |
1727 | * Adjacent cache entries tend to share the leading paths, so it makes | |
1728 | * sense to only store the differences in later entries. In the v4 | |
1729 | * on-disk format of the index, each on-disk cache entry stores the | |
1730 | * number of bytes to be stripped from the end of the previous name, | |
1731 | * and the bytes to append to the result, to come up with its name. | |
1732 | */ | |
1733 | int expand_name_field = istate->version == 4; | |
1734 | ||
1735 | /* On-disk flags are just 16 bits */ | |
1736 | flags = get_be16(&ondisk->flags); | |
1737 | len = flags & CE_NAMEMASK; | |
1738 | ||
1739 | if (flags & CE_EXTENDED) { | |
1740 | struct ondisk_cache_entry_extended *ondisk2; | |
1741 | int extended_flags; | |
1742 | ondisk2 = (struct ondisk_cache_entry_extended *)ondisk; | |
1743 | extended_flags = get_be16(&ondisk2->flags2) << 16; | |
1744 | /* We do not yet understand any bit out of CE_EXTENDED_FLAGS */ | |
1745 | if (extended_flags & ~CE_EXTENDED_FLAGS) | |
1746 | die("Unknown index entry format %08x", extended_flags); | |
1747 | flags |= extended_flags; | |
1748 | name = ondisk2->name; | |
1749 | } | |
1750 | else | |
1751 | name = ondisk->name; | |
1752 | ||
1753 | if (expand_name_field) { | |
1754 | const unsigned char *cp = (const unsigned char *)name; | |
1755 | size_t strip_len, previous_len; | |
1756 | ||
1757 | previous_len = previous_ce ? previous_ce->ce_namelen : 0; | |
1758 | strip_len = decode_varint(&cp); | |
1759 | if (previous_len < strip_len) { | |
1760 | if (previous_ce) | |
1761 | die(_("malformed name field in the index, near path '%s'"), | |
1762 | previous_ce->name); | |
1763 | else | |
1764 | die(_("malformed name field in the index in the first path")); | |
1765 | } | |
1766 | copy_len = previous_len - strip_len; | |
1767 | name = (const char *)cp; | |
1768 | } | |
1769 | ||
1770 | if (len == CE_NAMEMASK) { | |
1771 | len = strlen(name); | |
1772 | if (expand_name_field) | |
1773 | len += copy_len; | |
1774 | } | |
1775 | ||
1776 | ce = mem_pool__ce_alloc(istate->ce_mem_pool, len); | |
1777 | ||
1778 | ce->ce_stat_data.sd_ctime.sec = get_be32(&ondisk->ctime.sec); | |
1779 | ce->ce_stat_data.sd_mtime.sec = get_be32(&ondisk->mtime.sec); | |
1780 | ce->ce_stat_data.sd_ctime.nsec = get_be32(&ondisk->ctime.nsec); | |
1781 | ce->ce_stat_data.sd_mtime.nsec = get_be32(&ondisk->mtime.nsec); | |
1782 | ce->ce_stat_data.sd_dev = get_be32(&ondisk->dev); | |
1783 | ce->ce_stat_data.sd_ino = get_be32(&ondisk->ino); | |
1784 | ce->ce_mode = get_be32(&ondisk->mode); | |
1785 | ce->ce_stat_data.sd_uid = get_be32(&ondisk->uid); | |
1786 | ce->ce_stat_data.sd_gid = get_be32(&ondisk->gid); | |
1787 | ce->ce_stat_data.sd_size = get_be32(&ondisk->size); | |
1788 | ce->ce_flags = flags & ~CE_NAMEMASK; | |
1789 | ce->ce_namelen = len; | |
1790 | ce->index = 0; | |
1791 | hashcpy(ce->oid.hash, ondisk->sha1); | |
1792 | ||
1793 | if (expand_name_field) { | |
1794 | if (copy_len) | |
1795 | memcpy(ce->name, previous_ce->name, copy_len); | |
1796 | memcpy(ce->name + copy_len, name, len + 1 - copy_len); | |
1797 | *ent_size = (name - ((char *)ondisk)) + len + 1 - copy_len; | |
1798 | } else { | |
1799 | memcpy(ce->name, name, len + 1); | |
1800 | *ent_size = ondisk_ce_size(ce); | |
1801 | } | |
1802 | return ce; | |
1803 | } | |
1804 | ||
1805 | static void check_ce_order(struct index_state *istate) | |
1806 | { | |
1807 | unsigned int i; | |
1808 | ||
1809 | if (!verify_ce_order) | |
1810 | return; | |
1811 | ||
1812 | for (i = 1; i < istate->cache_nr; i++) { | |
1813 | struct cache_entry *ce = istate->cache[i - 1]; | |
1814 | struct cache_entry *next_ce = istate->cache[i]; | |
1815 | int name_compare = strcmp(ce->name, next_ce->name); | |
1816 | ||
1817 | if (0 < name_compare) | |
1818 | die("unordered stage entries in index"); | |
1819 | if (!name_compare) { | |
1820 | if (!ce_stage(ce)) | |
1821 | die("multiple stage entries for merged file '%s'", | |
1822 | ce->name); | |
1823 | if (ce_stage(ce) > ce_stage(next_ce)) | |
1824 | die("unordered stage entries for '%s'", | |
1825 | ce->name); | |
1826 | } | |
1827 | } | |
1828 | } | |
1829 | ||
1830 | static void tweak_untracked_cache(struct index_state *istate) | |
1831 | { | |
1832 | switch (git_config_get_untracked_cache()) { | |
1833 | case -1: /* keep: do nothing */ | |
1834 | break; | |
1835 | case 0: /* false */ | |
1836 | remove_untracked_cache(istate); | |
1837 | break; | |
1838 | case 1: /* true */ | |
1839 | add_untracked_cache(istate); | |
1840 | break; | |
1841 | default: /* unknown value: do nothing */ | |
1842 | break; | |
1843 | } | |
1844 | } | |
1845 | ||
1846 | static void tweak_split_index(struct index_state *istate) | |
1847 | { | |
1848 | switch (git_config_get_split_index()) { | |
1849 | case -1: /* unset: do nothing */ | |
1850 | break; | |
1851 | case 0: /* false */ | |
1852 | remove_split_index(istate); | |
1853 | break; | |
1854 | case 1: /* true */ | |
1855 | add_split_index(istate); | |
1856 | break; | |
1857 | default: /* unknown value: do nothing */ | |
1858 | break; | |
1859 | } | |
1860 | } | |
1861 | ||
1862 | static void post_read_index_from(struct index_state *istate) | |
1863 | { | |
1864 | check_ce_order(istate); | |
1865 | tweak_untracked_cache(istate); | |
1866 | tweak_split_index(istate); | |
1867 | tweak_fsmonitor(istate); | |
1868 | } | |
1869 | ||
1870 | static size_t estimate_cache_size_from_compressed(unsigned int entries) | |
1871 | { | |
1872 | return entries * (sizeof(struct cache_entry) + CACHE_ENTRY_PATH_LENGTH); | |
1873 | } | |
1874 | ||
1875 | static size_t estimate_cache_size(size_t ondisk_size, unsigned int entries) | |
1876 | { | |
1877 | long per_entry = sizeof(struct cache_entry) - sizeof(struct ondisk_cache_entry); | |
1878 | ||
1879 | /* | |
1880 | * Account for potential alignment differences. | |
1881 | */ | |
1882 | per_entry += align_padding_size(sizeof(struct cache_entry), -sizeof(struct ondisk_cache_entry)); | |
1883 | return ondisk_size + entries * per_entry; | |
1884 | } | |
1885 | ||
1886 | /* remember to discard_cache() before reading a different cache! */ | |
1887 | int do_read_index(struct index_state *istate, const char *path, int must_exist) | |
1888 | { | |
1889 | int fd, i; | |
1890 | struct stat st; | |
1891 | unsigned long src_offset; | |
1892 | const struct cache_header *hdr; | |
1893 | const char *mmap; | |
1894 | size_t mmap_size; | |
1895 | const struct cache_entry *previous_ce = NULL; | |
1896 | ||
1897 | if (istate->initialized) | |
1898 | return istate->cache_nr; | |
1899 | ||
1900 | istate->timestamp.sec = 0; | |
1901 | istate->timestamp.nsec = 0; | |
1902 | fd = open(path, O_RDONLY); | |
1903 | if (fd < 0) { | |
1904 | if (!must_exist && errno == ENOENT) | |
1905 | return 0; | |
1906 | die_errno("%s: index file open failed", path); | |
1907 | } | |
1908 | ||
1909 | if (fstat(fd, &st)) | |
1910 | die_errno("cannot stat the open index"); | |
1911 | ||
1912 | mmap_size = xsize_t(st.st_size); | |
1913 | if (mmap_size < sizeof(struct cache_header) + the_hash_algo->rawsz) | |
1914 | die("index file smaller than expected"); | |
1915 | ||
1916 | mmap = xmmap(NULL, mmap_size, PROT_READ, MAP_PRIVATE, fd, 0); | |
1917 | if (mmap == MAP_FAILED) | |
1918 | die_errno("unable to map index file"); | |
1919 | close(fd); | |
1920 | ||
1921 | hdr = (const struct cache_header *)mmap; | |
1922 | if (verify_hdr(hdr, mmap_size) < 0) | |
1923 | goto unmap; | |
1924 | ||
1925 | hashcpy(istate->oid.hash, (const unsigned char *)hdr + mmap_size - the_hash_algo->rawsz); | |
1926 | istate->version = ntohl(hdr->hdr_version); | |
1927 | istate->cache_nr = ntohl(hdr->hdr_entries); | |
1928 | istate->cache_alloc = alloc_nr(istate->cache_nr); | |
1929 | istate->cache = xcalloc(istate->cache_alloc, sizeof(*istate->cache)); | |
1930 | istate->initialized = 1; | |
1931 | ||
1932 | if (istate->version == 4) { | |
1933 | mem_pool_init(&istate->ce_mem_pool, | |
1934 | estimate_cache_size_from_compressed(istate->cache_nr)); | |
1935 | } else { | |
1936 | mem_pool_init(&istate->ce_mem_pool, | |
1937 | estimate_cache_size(mmap_size, istate->cache_nr)); | |
1938 | } | |
1939 | ||
1940 | src_offset = sizeof(*hdr); | |
1941 | for (i = 0; i < istate->cache_nr; i++) { | |
1942 | struct ondisk_cache_entry *disk_ce; | |
1943 | struct cache_entry *ce; | |
1944 | unsigned long consumed; | |
1945 | ||
1946 | disk_ce = (struct ondisk_cache_entry *)(mmap + src_offset); | |
1947 | ce = create_from_disk(istate, disk_ce, &consumed, previous_ce); | |
1948 | set_index_entry(istate, i, ce); | |
1949 | ||
1950 | src_offset += consumed; | |
1951 | previous_ce = ce; | |
1952 | } | |
1953 | istate->timestamp.sec = st.st_mtime; | |
1954 | istate->timestamp.nsec = ST_MTIME_NSEC(st); | |
1955 | ||
1956 | while (src_offset <= mmap_size - the_hash_algo->rawsz - 8) { | |
1957 | /* After an array of active_nr index entries, | |
1958 | * there can be arbitrary number of extended | |
1959 | * sections, each of which is prefixed with | |
1960 | * extension name (4-byte) and section length | |
1961 | * in 4-byte network byte order. | |
1962 | */ | |
1963 | uint32_t extsize; | |
1964 | extsize = get_be32(mmap + src_offset + 4); | |
1965 | if (read_index_extension(istate, | |
1966 | mmap + src_offset, | |
1967 | mmap + src_offset + 8, | |
1968 | extsize) < 0) | |
1969 | goto unmap; | |
1970 | src_offset += 8; | |
1971 | src_offset += extsize; | |
1972 | } | |
1973 | munmap((void *)mmap, mmap_size); | |
1974 | return istate->cache_nr; | |
1975 | ||
1976 | unmap: | |
1977 | munmap((void *)mmap, mmap_size); | |
1978 | die("index file corrupt"); | |
1979 | } | |
1980 | ||
1981 | /* | |
1982 | * Signal that the shared index is used by updating its mtime. | |
1983 | * | |
1984 | * This way, shared index can be removed if they have not been used | |
1985 | * for some time. | |
1986 | */ | |
1987 | static void freshen_shared_index(const char *shared_index, int warn) | |
1988 | { | |
1989 | if (!check_and_freshen_file(shared_index, 1) && warn) | |
1990 | warning("could not freshen shared index '%s'", shared_index); | |
1991 | } | |
1992 | ||
1993 | int read_index_from(struct index_state *istate, const char *path, | |
1994 | const char *gitdir) | |
1995 | { | |
1996 | struct split_index *split_index; | |
1997 | int ret; | |
1998 | char *base_oid_hex; | |
1999 | char *base_path; | |
2000 | ||
2001 | /* istate->initialized covers both .git/index and .git/sharedindex.xxx */ | |
2002 | if (istate->initialized) | |
2003 | return istate->cache_nr; | |
2004 | ||
2005 | trace_performance_enter(); | |
2006 | ret = do_read_index(istate, path, 0); | |
2007 | trace_performance_leave("read cache %s", path); | |
2008 | ||
2009 | split_index = istate->split_index; | |
2010 | if (!split_index || is_null_oid(&split_index->base_oid)) { | |
2011 | post_read_index_from(istate); | |
2012 | return ret; | |
2013 | } | |
2014 | ||
2015 | trace_performance_enter(); | |
2016 | if (split_index->base) | |
2017 | discard_index(split_index->base); | |
2018 | else | |
2019 | split_index->base = xcalloc(1, sizeof(*split_index->base)); | |
2020 | ||
2021 | base_oid_hex = oid_to_hex(&split_index->base_oid); | |
2022 | base_path = xstrfmt("%s/sharedindex.%s", gitdir, base_oid_hex); | |
2023 | ret = do_read_index(split_index->base, base_path, 1); | |
2024 | if (!oideq(&split_index->base_oid, &split_index->base->oid)) | |
2025 | die("broken index, expect %s in %s, got %s", | |
2026 | base_oid_hex, base_path, | |
2027 | oid_to_hex(&split_index->base->oid)); | |
2028 | ||
2029 | freshen_shared_index(base_path, 0); | |
2030 | merge_base_index(istate); | |
2031 | post_read_index_from(istate); | |
2032 | free(base_path); | |
2033 | trace_performance_leave("read cache %s", base_path); | |
2034 | return ret; | |
2035 | } | |
2036 | ||
2037 | int is_index_unborn(struct index_state *istate) | |
2038 | { | |
2039 | return (!istate->cache_nr && !istate->timestamp.sec); | |
2040 | } | |
2041 | ||
2042 | int discard_index(struct index_state *istate) | |
2043 | { | |
2044 | /* | |
2045 | * Cache entries in istate->cache[] should have been allocated | |
2046 | * from the memory pool associated with this index, or from an | |
2047 | * associated split_index. There is no need to free individual | |
2048 | * cache entries. validate_cache_entries can detect when this | |
2049 | * assertion does not hold. | |
2050 | */ | |
2051 | validate_cache_entries(istate); | |
2052 | ||
2053 | resolve_undo_clear_index(istate); | |
2054 | istate->cache_nr = 0; | |
2055 | istate->cache_changed = 0; | |
2056 | istate->timestamp.sec = 0; | |
2057 | istate->timestamp.nsec = 0; | |
2058 | free_name_hash(istate); | |
2059 | cache_tree_free(&(istate->cache_tree)); | |
2060 | istate->initialized = 0; | |
2061 | FREE_AND_NULL(istate->cache); | |
2062 | istate->cache_alloc = 0; | |
2063 | discard_split_index(istate); | |
2064 | free_untracked_cache(istate->untracked); | |
2065 | istate->untracked = NULL; | |
2066 | ||
2067 | if (istate->ce_mem_pool) { | |
2068 | mem_pool_discard(istate->ce_mem_pool, should_validate_cache_entries()); | |
2069 | istate->ce_mem_pool = NULL; | |
2070 | } | |
2071 | ||
2072 | return 0; | |
2073 | } | |
2074 | ||
2075 | /* | |
2076 | * Validate the cache entries of this index. | |
2077 | * All cache entries associated with this index | |
2078 | * should have been allocated by the memory pool | |
2079 | * associated with this index, or by a referenced | |
2080 | * split index. | |
2081 | */ | |
2082 | void validate_cache_entries(const struct index_state *istate) | |
2083 | { | |
2084 | int i; | |
2085 | ||
2086 | if (!should_validate_cache_entries() ||!istate || !istate->initialized) | |
2087 | return; | |
2088 | ||
2089 | for (i = 0; i < istate->cache_nr; i++) { | |
2090 | if (!istate) { | |
2091 | die("internal error: cache entry is not allocated from expected memory pool"); | |
2092 | } else if (!istate->ce_mem_pool || | |
2093 | !mem_pool_contains(istate->ce_mem_pool, istate->cache[i])) { | |
2094 | if (!istate->split_index || | |
2095 | !istate->split_index->base || | |
2096 | !istate->split_index->base->ce_mem_pool || | |
2097 | !mem_pool_contains(istate->split_index->base->ce_mem_pool, istate->cache[i])) { | |
2098 | die("internal error: cache entry is not allocated from expected memory pool"); | |
2099 | } | |
2100 | } | |
2101 | } | |
2102 | ||
2103 | if (istate->split_index) | |
2104 | validate_cache_entries(istate->split_index->base); | |
2105 | } | |
2106 | ||
2107 | int unmerged_index(const struct index_state *istate) | |
2108 | { | |
2109 | int i; | |
2110 | for (i = 0; i < istate->cache_nr; i++) { | |
2111 | if (ce_stage(istate->cache[i])) | |
2112 | return 1; | |
2113 | } | |
2114 | return 0; | |
2115 | } | |
2116 | ||
2117 | int index_has_changes(const struct index_state *istate, | |
2118 | struct tree *tree, | |
2119 | struct strbuf *sb) | |
2120 | { | |
2121 | struct object_id cmp; | |
2122 | int i; | |
2123 | ||
2124 | if (istate != &the_index) { | |
2125 | BUG("index_has_changes cannot yet accept istate != &the_index; do_diff_cache needs updating first."); | |
2126 | } | |
2127 | if (tree) | |
2128 | cmp = tree->object.oid; | |
2129 | if (tree || !get_oid_tree("HEAD", &cmp)) { | |
2130 | struct diff_options opt; | |
2131 | ||
2132 | diff_setup(&opt); | |
2133 | opt.flags.exit_with_status = 1; | |
2134 | if (!sb) | |
2135 | opt.flags.quick = 1; | |
2136 | do_diff_cache(&cmp, &opt); | |
2137 | diffcore_std(&opt); | |
2138 | for (i = 0; sb && i < diff_queued_diff.nr; i++) { | |
2139 | if (i) | |
2140 | strbuf_addch(sb, ' '); | |
2141 | strbuf_addstr(sb, diff_queued_diff.queue[i]->two->path); | |
2142 | } | |
2143 | diff_flush(&opt); | |
2144 | return opt.flags.has_changes != 0; | |
2145 | } else { | |
2146 | for (i = 0; sb && i < istate->cache_nr; i++) { | |
2147 | if (i) | |
2148 | strbuf_addch(sb, ' '); | |
2149 | strbuf_addstr(sb, istate->cache[i]->name); | |
2150 | } | |
2151 | return !!istate->cache_nr; | |
2152 | } | |
2153 | } | |
2154 | ||
2155 | #define WRITE_BUFFER_SIZE 8192 | |
2156 | static unsigned char write_buffer[WRITE_BUFFER_SIZE]; | |
2157 | static unsigned long write_buffer_len; | |
2158 | ||
2159 | static int ce_write_flush(git_hash_ctx *context, int fd) | |
2160 | { | |
2161 | unsigned int buffered = write_buffer_len; | |
2162 | if (buffered) { | |
2163 | the_hash_algo->update_fn(context, write_buffer, buffered); | |
2164 | if (write_in_full(fd, write_buffer, buffered) < 0) | |
2165 | return -1; | |
2166 | write_buffer_len = 0; | |
2167 | } | |
2168 | return 0; | |
2169 | } | |
2170 | ||
2171 | static int ce_write(git_hash_ctx *context, int fd, void *data, unsigned int len) | |
2172 | { | |
2173 | while (len) { | |
2174 | unsigned int buffered = write_buffer_len; | |
2175 | unsigned int partial = WRITE_BUFFER_SIZE - buffered; | |
2176 | if (partial > len) | |
2177 | partial = len; | |
2178 | memcpy(write_buffer + buffered, data, partial); | |
2179 | buffered += partial; | |
2180 | if (buffered == WRITE_BUFFER_SIZE) { | |
2181 | write_buffer_len = buffered; | |
2182 | if (ce_write_flush(context, fd)) | |
2183 | return -1; | |
2184 | buffered = 0; | |
2185 | } | |
2186 | write_buffer_len = buffered; | |
2187 | len -= partial; | |
2188 | data = (char *) data + partial; | |
2189 | } | |
2190 | return 0; | |
2191 | } | |
2192 | ||
2193 | static int write_index_ext_header(git_hash_ctx *context, int fd, | |
2194 | unsigned int ext, unsigned int sz) | |
2195 | { | |
2196 | ext = htonl(ext); | |
2197 | sz = htonl(sz); | |
2198 | return ((ce_write(context, fd, &ext, 4) < 0) || | |
2199 | (ce_write(context, fd, &sz, 4) < 0)) ? -1 : 0; | |
2200 | } | |
2201 | ||
2202 | static int ce_flush(git_hash_ctx *context, int fd, unsigned char *hash) | |
2203 | { | |
2204 | unsigned int left = write_buffer_len; | |
2205 | ||
2206 | if (left) { | |
2207 | write_buffer_len = 0; | |
2208 | the_hash_algo->update_fn(context, write_buffer, left); | |
2209 | } | |
2210 | ||
2211 | /* Flush first if not enough space for hash signature */ | |
2212 | if (left + the_hash_algo->rawsz > WRITE_BUFFER_SIZE) { | |
2213 | if (write_in_full(fd, write_buffer, left) < 0) | |
2214 | return -1; | |
2215 | left = 0; | |
2216 | } | |
2217 | ||
2218 | /* Append the hash signature at the end */ | |
2219 | the_hash_algo->final_fn(write_buffer + left, context); | |
2220 | hashcpy(hash, write_buffer + left); | |
2221 | left += the_hash_algo->rawsz; | |
2222 | return (write_in_full(fd, write_buffer, left) < 0) ? -1 : 0; | |
2223 | } | |
2224 | ||
2225 | static void ce_smudge_racily_clean_entry(struct cache_entry *ce) | |
2226 | { | |
2227 | /* | |
2228 | * The only thing we care about in this function is to smudge the | |
2229 | * falsely clean entry due to touch-update-touch race, so we leave | |
2230 | * everything else as they are. We are called for entries whose | |
2231 | * ce_stat_data.sd_mtime match the index file mtime. | |
2232 | * | |
2233 | * Note that this actually does not do much for gitlinks, for | |
2234 | * which ce_match_stat_basic() always goes to the actual | |
2235 | * contents. The caller checks with is_racy_timestamp() which | |
2236 | * always says "no" for gitlinks, so we are not called for them ;-) | |
2237 | */ | |
2238 | struct stat st; | |
2239 | ||
2240 | if (lstat(ce->name, &st) < 0) | |
2241 | return; | |
2242 | if (ce_match_stat_basic(ce, &st)) | |
2243 | return; | |
2244 | if (ce_modified_check_fs(ce, &st)) { | |
2245 | /* This is "racily clean"; smudge it. Note that this | |
2246 | * is a tricky code. At first glance, it may appear | |
2247 | * that it can break with this sequence: | |
2248 | * | |
2249 | * $ echo xyzzy >frotz | |
2250 | * $ git-update-index --add frotz | |
2251 | * $ : >frotz | |
2252 | * $ sleep 3 | |
2253 | * $ echo filfre >nitfol | |
2254 | * $ git-update-index --add nitfol | |
2255 | * | |
2256 | * but it does not. When the second update-index runs, | |
2257 | * it notices that the entry "frotz" has the same timestamp | |
2258 | * as index, and if we were to smudge it by resetting its | |
2259 | * size to zero here, then the object name recorded | |
2260 | * in index is the 6-byte file but the cached stat information | |
2261 | * becomes zero --- which would then match what we would | |
2262 | * obtain from the filesystem next time we stat("frotz"). | |
2263 | * | |
2264 | * However, the second update-index, before calling | |
2265 | * this function, notices that the cached size is 6 | |
2266 | * bytes and what is on the filesystem is an empty | |
2267 | * file, and never calls us, so the cached size information | |
2268 | * for "frotz" stays 6 which does not match the filesystem. | |
2269 | */ | |
2270 | ce->ce_stat_data.sd_size = 0; | |
2271 | } | |
2272 | } | |
2273 | ||
2274 | /* Copy miscellaneous fields but not the name */ | |
2275 | static void copy_cache_entry_to_ondisk(struct ondisk_cache_entry *ondisk, | |
2276 | struct cache_entry *ce) | |
2277 | { | |
2278 | short flags; | |
2279 | ||
2280 | ondisk->ctime.sec = htonl(ce->ce_stat_data.sd_ctime.sec); | |
2281 | ondisk->mtime.sec = htonl(ce->ce_stat_data.sd_mtime.sec); | |
2282 | ondisk->ctime.nsec = htonl(ce->ce_stat_data.sd_ctime.nsec); | |
2283 | ondisk->mtime.nsec = htonl(ce->ce_stat_data.sd_mtime.nsec); | |
2284 | ondisk->dev = htonl(ce->ce_stat_data.sd_dev); | |
2285 | ondisk->ino = htonl(ce->ce_stat_data.sd_ino); | |
2286 | ondisk->mode = htonl(ce->ce_mode); | |
2287 | ondisk->uid = htonl(ce->ce_stat_data.sd_uid); | |
2288 | ondisk->gid = htonl(ce->ce_stat_data.sd_gid); | |
2289 | ondisk->size = htonl(ce->ce_stat_data.sd_size); | |
2290 | hashcpy(ondisk->sha1, ce->oid.hash); | |
2291 | ||
2292 | flags = ce->ce_flags & ~CE_NAMEMASK; | |
2293 | flags |= (ce_namelen(ce) >= CE_NAMEMASK ? CE_NAMEMASK : ce_namelen(ce)); | |
2294 | ondisk->flags = htons(flags); | |
2295 | if (ce->ce_flags & CE_EXTENDED) { | |
2296 | struct ondisk_cache_entry_extended *ondisk2; | |
2297 | ondisk2 = (struct ondisk_cache_entry_extended *)ondisk; | |
2298 | ondisk2->flags2 = htons((ce->ce_flags & CE_EXTENDED_FLAGS) >> 16); | |
2299 | } | |
2300 | } | |
2301 | ||
2302 | static int ce_write_entry(git_hash_ctx *c, int fd, struct cache_entry *ce, | |
2303 | struct strbuf *previous_name, struct ondisk_cache_entry *ondisk) | |
2304 | { | |
2305 | int size; | |
2306 | int result; | |
2307 | unsigned int saved_namelen; | |
2308 | int stripped_name = 0; | |
2309 | static unsigned char padding[8] = { 0x00 }; | |
2310 | ||
2311 | if (ce->ce_flags & CE_STRIP_NAME) { | |
2312 | saved_namelen = ce_namelen(ce); | |
2313 | ce->ce_namelen = 0; | |
2314 | stripped_name = 1; | |
2315 | } | |
2316 | ||
2317 | if (ce->ce_flags & CE_EXTENDED) | |
2318 | size = offsetof(struct ondisk_cache_entry_extended, name); | |
2319 | else | |
2320 | size = offsetof(struct ondisk_cache_entry, name); | |
2321 | ||
2322 | if (!previous_name) { | |
2323 | int len = ce_namelen(ce); | |
2324 | copy_cache_entry_to_ondisk(ondisk, ce); | |
2325 | result = ce_write(c, fd, ondisk, size); | |
2326 | if (!result) | |
2327 | result = ce_write(c, fd, ce->name, len); | |
2328 | if (!result) | |
2329 | result = ce_write(c, fd, padding, align_padding_size(size, len)); | |
2330 | } else { | |
2331 | int common, to_remove, prefix_size; | |
2332 | unsigned char to_remove_vi[16]; | |
2333 | for (common = 0; | |
2334 | (ce->name[common] && | |
2335 | common < previous_name->len && | |
2336 | ce->name[common] == previous_name->buf[common]); | |
2337 | common++) | |
2338 | ; /* still matching */ | |
2339 | to_remove = previous_name->len - common; | |
2340 | prefix_size = encode_varint(to_remove, to_remove_vi); | |
2341 | ||
2342 | copy_cache_entry_to_ondisk(ondisk, ce); | |
2343 | result = ce_write(c, fd, ondisk, size); | |
2344 | if (!result) | |
2345 | result = ce_write(c, fd, to_remove_vi, prefix_size); | |
2346 | if (!result) | |
2347 | result = ce_write(c, fd, ce->name + common, ce_namelen(ce) - common); | |
2348 | if (!result) | |
2349 | result = ce_write(c, fd, padding, 1); | |
2350 | ||
2351 | strbuf_splice(previous_name, common, to_remove, | |
2352 | ce->name + common, ce_namelen(ce) - common); | |
2353 | } | |
2354 | if (stripped_name) { | |
2355 | ce->ce_namelen = saved_namelen; | |
2356 | ce->ce_flags &= ~CE_STRIP_NAME; | |
2357 | } | |
2358 | ||
2359 | return result; | |
2360 | } | |
2361 | ||
2362 | /* | |
2363 | * This function verifies if index_state has the correct sha1 of the | |
2364 | * index file. Don't die if we have any other failure, just return 0. | |
2365 | */ | |
2366 | static int verify_index_from(const struct index_state *istate, const char *path) | |
2367 | { | |
2368 | int fd; | |
2369 | ssize_t n; | |
2370 | struct stat st; | |
2371 | unsigned char hash[GIT_MAX_RAWSZ]; | |
2372 | ||
2373 | if (!istate->initialized) | |
2374 | return 0; | |
2375 | ||
2376 | fd = open(path, O_RDONLY); | |
2377 | if (fd < 0) | |
2378 | return 0; | |
2379 | ||
2380 | if (fstat(fd, &st)) | |
2381 | goto out; | |
2382 | ||
2383 | if (st.st_size < sizeof(struct cache_header) + the_hash_algo->rawsz) | |
2384 | goto out; | |
2385 | ||
2386 | n = pread_in_full(fd, hash, the_hash_algo->rawsz, st.st_size - the_hash_algo->rawsz); | |
2387 | if (n != the_hash_algo->rawsz) | |
2388 | goto out; | |
2389 | ||
2390 | if (!hasheq(istate->oid.hash, hash)) | |
2391 | goto out; | |
2392 | ||
2393 | close(fd); | |
2394 | return 1; | |
2395 | ||
2396 | out: | |
2397 | close(fd); | |
2398 | return 0; | |
2399 | } | |
2400 | ||
2401 | static int verify_index(const struct index_state *istate) | |
2402 | { | |
2403 | return verify_index_from(istate, get_index_file()); | |
2404 | } | |
2405 | ||
2406 | static int has_racy_timestamp(struct index_state *istate) | |
2407 | { | |
2408 | int entries = istate->cache_nr; | |
2409 | int i; | |
2410 | ||
2411 | for (i = 0; i < entries; i++) { | |
2412 | struct cache_entry *ce = istate->cache[i]; | |
2413 | if (is_racy_timestamp(istate, ce)) | |
2414 | return 1; | |
2415 | } | |
2416 | return 0; | |
2417 | } | |
2418 | ||
2419 | void update_index_if_able(struct index_state *istate, struct lock_file *lockfile) | |
2420 | { | |
2421 | if ((istate->cache_changed || has_racy_timestamp(istate)) && | |
2422 | verify_index(istate)) | |
2423 | write_locked_index(istate, lockfile, COMMIT_LOCK); | |
2424 | else | |
2425 | rollback_lock_file(lockfile); | |
2426 | } | |
2427 | ||
2428 | /* | |
2429 | * On success, `tempfile` is closed. If it is the temporary file | |
2430 | * of a `struct lock_file`, we will therefore effectively perform | |
2431 | * a 'close_lock_file_gently()`. Since that is an implementation | |
2432 | * detail of lockfiles, callers of `do_write_index()` should not | |
2433 | * rely on it. | |
2434 | */ | |
2435 | static int do_write_index(struct index_state *istate, struct tempfile *tempfile, | |
2436 | int strip_extensions) | |
2437 | { | |
2438 | uint64_t start = getnanotime(); | |
2439 | int newfd = tempfile->fd; | |
2440 | git_hash_ctx c; | |
2441 | struct cache_header hdr; | |
2442 | int i, err = 0, removed, extended, hdr_version; | |
2443 | struct cache_entry **cache = istate->cache; | |
2444 | int entries = istate->cache_nr; | |
2445 | struct stat st; | |
2446 | struct ondisk_cache_entry_extended ondisk; | |
2447 | struct strbuf previous_name_buf = STRBUF_INIT, *previous_name; | |
2448 | int drop_cache_tree = istate->drop_cache_tree; | |
2449 | ||
2450 | for (i = removed = extended = 0; i < entries; i++) { | |
2451 | if (cache[i]->ce_flags & CE_REMOVE) | |
2452 | removed++; | |
2453 | ||
2454 | /* reduce extended entries if possible */ | |
2455 | cache[i]->ce_flags &= ~CE_EXTENDED; | |
2456 | if (cache[i]->ce_flags & CE_EXTENDED_FLAGS) { | |
2457 | extended++; | |
2458 | cache[i]->ce_flags |= CE_EXTENDED; | |
2459 | } | |
2460 | } | |
2461 | ||
2462 | if (!istate->version) { | |
2463 | istate->version = get_index_format_default(); | |
2464 | if (git_env_bool("GIT_TEST_SPLIT_INDEX", 0)) | |
2465 | init_split_index(istate); | |
2466 | } | |
2467 | ||
2468 | /* demote version 3 to version 2 when the latter suffices */ | |
2469 | if (istate->version == 3 || istate->version == 2) | |
2470 | istate->version = extended ? 3 : 2; | |
2471 | ||
2472 | hdr_version = istate->version; | |
2473 | ||
2474 | hdr.hdr_signature = htonl(CACHE_SIGNATURE); | |
2475 | hdr.hdr_version = htonl(hdr_version); | |
2476 | hdr.hdr_entries = htonl(entries - removed); | |
2477 | ||
2478 | the_hash_algo->init_fn(&c); | |
2479 | if (ce_write(&c, newfd, &hdr, sizeof(hdr)) < 0) | |
2480 | return -1; | |
2481 | ||
2482 | previous_name = (hdr_version == 4) ? &previous_name_buf : NULL; | |
2483 | ||
2484 | for (i = 0; i < entries; i++) { | |
2485 | struct cache_entry *ce = cache[i]; | |
2486 | if (ce->ce_flags & CE_REMOVE) | |
2487 | continue; | |
2488 | if (!ce_uptodate(ce) && is_racy_timestamp(istate, ce)) | |
2489 | ce_smudge_racily_clean_entry(ce); | |
2490 | if (is_null_oid(&ce->oid)) { | |
2491 | static const char msg[] = "cache entry has null sha1: %s"; | |
2492 | static int allow = -1; | |
2493 | ||
2494 | if (allow < 0) | |
2495 | allow = git_env_bool("GIT_ALLOW_NULL_SHA1", 0); | |
2496 | if (allow) | |
2497 | warning(msg, ce->name); | |
2498 | else | |
2499 | err = error(msg, ce->name); | |
2500 | ||
2501 | drop_cache_tree = 1; | |
2502 | } | |
2503 | if (ce_write_entry(&c, newfd, ce, previous_name, (struct ondisk_cache_entry *)&ondisk) < 0) | |
2504 | err = -1; | |
2505 | ||
2506 | if (err) | |
2507 | break; | |
2508 | } | |
2509 | strbuf_release(&previous_name_buf); | |
2510 | ||
2511 | if (err) | |
2512 | return err; | |
2513 | ||
2514 | /* Write extension data here */ | |
2515 | if (!strip_extensions && istate->split_index) { | |
2516 | struct strbuf sb = STRBUF_INIT; | |
2517 | ||
2518 | err = write_link_extension(&sb, istate) < 0 || | |
2519 | write_index_ext_header(&c, newfd, CACHE_EXT_LINK, | |
2520 | sb.len) < 0 || | |
2521 | ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2522 | strbuf_release(&sb); | |
2523 | if (err) | |
2524 | return -1; | |
2525 | } | |
2526 | if (!strip_extensions && !drop_cache_tree && istate->cache_tree) { | |
2527 | struct strbuf sb = STRBUF_INIT; | |
2528 | ||
2529 | cache_tree_write(&sb, istate->cache_tree); | |
2530 | err = write_index_ext_header(&c, newfd, CACHE_EXT_TREE, sb.len) < 0 | |
2531 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2532 | strbuf_release(&sb); | |
2533 | if (err) | |
2534 | return -1; | |
2535 | } | |
2536 | if (!strip_extensions && istate->resolve_undo) { | |
2537 | struct strbuf sb = STRBUF_INIT; | |
2538 | ||
2539 | resolve_undo_write(&sb, istate->resolve_undo); | |
2540 | err = write_index_ext_header(&c, newfd, CACHE_EXT_RESOLVE_UNDO, | |
2541 | sb.len) < 0 | |
2542 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2543 | strbuf_release(&sb); | |
2544 | if (err) | |
2545 | return -1; | |
2546 | } | |
2547 | if (!strip_extensions && istate->untracked) { | |
2548 | struct strbuf sb = STRBUF_INIT; | |
2549 | ||
2550 | write_untracked_extension(&sb, istate->untracked); | |
2551 | err = write_index_ext_header(&c, newfd, CACHE_EXT_UNTRACKED, | |
2552 | sb.len) < 0 || | |
2553 | ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2554 | strbuf_release(&sb); | |
2555 | if (err) | |
2556 | return -1; | |
2557 | } | |
2558 | if (!strip_extensions && istate->fsmonitor_last_update) { | |
2559 | struct strbuf sb = STRBUF_INIT; | |
2560 | ||
2561 | write_fsmonitor_extension(&sb, istate); | |
2562 | err = write_index_ext_header(&c, newfd, CACHE_EXT_FSMONITOR, sb.len) < 0 | |
2563 | || ce_write(&c, newfd, sb.buf, sb.len) < 0; | |
2564 | strbuf_release(&sb); | |
2565 | if (err) | |
2566 | return -1; | |
2567 | } | |
2568 | ||
2569 | if (ce_flush(&c, newfd, istate->oid.hash)) | |
2570 | return -1; | |
2571 | if (close_tempfile_gently(tempfile)) { | |
2572 | error(_("could not close '%s'"), tempfile->filename.buf); | |
2573 | return -1; | |
2574 | } | |
2575 | if (stat(tempfile->filename.buf, &st)) | |
2576 | return -1; | |
2577 | istate->timestamp.sec = (unsigned int)st.st_mtime; | |
2578 | istate->timestamp.nsec = ST_MTIME_NSEC(st); | |
2579 | trace_performance_since(start, "write index, changed mask = %x", istate->cache_changed); | |
2580 | return 0; | |
2581 | } | |
2582 | ||
2583 | void set_alternate_index_output(const char *name) | |
2584 | { | |
2585 | alternate_index_output = name; | |
2586 | } | |
2587 | ||
2588 | static int commit_locked_index(struct lock_file *lk) | |
2589 | { | |
2590 | if (alternate_index_output) | |
2591 | return commit_lock_file_to(lk, alternate_index_output); | |
2592 | else | |
2593 | return commit_lock_file(lk); | |
2594 | } | |
2595 | ||
2596 | static int do_write_locked_index(struct index_state *istate, struct lock_file *lock, | |
2597 | unsigned flags) | |
2598 | { | |
2599 | int ret = do_write_index(istate, lock->tempfile, 0); | |
2600 | if (ret) | |
2601 | return ret; | |
2602 | if (flags & COMMIT_LOCK) | |
2603 | return commit_locked_index(lock); | |
2604 | return close_lock_file_gently(lock); | |
2605 | } | |
2606 | ||
2607 | static int write_split_index(struct index_state *istate, | |
2608 | struct lock_file *lock, | |
2609 | unsigned flags) | |
2610 | { | |
2611 | int ret; | |
2612 | prepare_to_write_split_index(istate); | |
2613 | ret = do_write_locked_index(istate, lock, flags); | |
2614 | finish_writing_split_index(istate); | |
2615 | return ret; | |
2616 | } | |
2617 | ||
2618 | static const char *shared_index_expire = "2.weeks.ago"; | |
2619 | ||
2620 | static unsigned long get_shared_index_expire_date(void) | |
2621 | { | |
2622 | static unsigned long shared_index_expire_date; | |
2623 | static int shared_index_expire_date_prepared; | |
2624 | ||
2625 | if (!shared_index_expire_date_prepared) { | |
2626 | git_config_get_expiry("splitindex.sharedindexexpire", | |
2627 | &shared_index_expire); | |
2628 | shared_index_expire_date = approxidate(shared_index_expire); | |
2629 | shared_index_expire_date_prepared = 1; | |
2630 | } | |
2631 | ||
2632 | return shared_index_expire_date; | |
2633 | } | |
2634 | ||
2635 | static int should_delete_shared_index(const char *shared_index_path) | |
2636 | { | |
2637 | struct stat st; | |
2638 | unsigned long expiration; | |
2639 | ||
2640 | /* Check timestamp */ | |
2641 | expiration = get_shared_index_expire_date(); | |
2642 | if (!expiration) | |
2643 | return 0; | |
2644 | if (stat(shared_index_path, &st)) | |
2645 | return error_errno(_("could not stat '%s'"), shared_index_path); | |
2646 | if (st.st_mtime > expiration) | |
2647 | return 0; | |
2648 | ||
2649 | return 1; | |
2650 | } | |
2651 | ||
2652 | static int clean_shared_index_files(const char *current_hex) | |
2653 | { | |
2654 | struct dirent *de; | |
2655 | DIR *dir = opendir(get_git_dir()); | |
2656 | ||
2657 | if (!dir) | |
2658 | return error_errno(_("unable to open git dir: %s"), get_git_dir()); | |
2659 | ||
2660 | while ((de = readdir(dir)) != NULL) { | |
2661 | const char *sha1_hex; | |
2662 | const char *shared_index_path; | |
2663 | if (!skip_prefix(de->d_name, "sharedindex.", &sha1_hex)) | |
2664 | continue; | |
2665 | if (!strcmp(sha1_hex, current_hex)) | |
2666 | continue; | |
2667 | shared_index_path = git_path("%s", de->d_name); | |
2668 | if (should_delete_shared_index(shared_index_path) > 0 && | |
2669 | unlink(shared_index_path)) | |
2670 | warning_errno(_("unable to unlink: %s"), shared_index_path); | |
2671 | } | |
2672 | closedir(dir); | |
2673 | ||
2674 | return 0; | |
2675 | } | |
2676 | ||
2677 | static int write_shared_index(struct index_state *istate, | |
2678 | struct tempfile **temp) | |
2679 | { | |
2680 | struct split_index *si = istate->split_index; | |
2681 | int ret; | |
2682 | ||
2683 | move_cache_to_base_index(istate); | |
2684 | ret = do_write_index(si->base, *temp, 1); | |
2685 | if (ret) | |
2686 | return ret; | |
2687 | ret = adjust_shared_perm(get_tempfile_path(*temp)); | |
2688 | if (ret) { | |
2689 | error("cannot fix permission bits on %s", get_tempfile_path(*temp)); | |
2690 | return ret; | |
2691 | } | |
2692 | ret = rename_tempfile(temp, | |
2693 | git_path("sharedindex.%s", oid_to_hex(&si->base->oid))); | |
2694 | if (!ret) { | |
2695 | oidcpy(&si->base_oid, &si->base->oid); | |
2696 | clean_shared_index_files(oid_to_hex(&si->base->oid)); | |
2697 | } | |
2698 | ||
2699 | return ret; | |
2700 | } | |
2701 | ||
2702 | static const int default_max_percent_split_change = 20; | |
2703 | ||
2704 | static int too_many_not_shared_entries(struct index_state *istate) | |
2705 | { | |
2706 | int i, not_shared = 0; | |
2707 | int max_split = git_config_get_max_percent_split_change(); | |
2708 | ||
2709 | switch (max_split) { | |
2710 | case -1: | |
2711 | /* not or badly configured: use the default value */ | |
2712 | max_split = default_max_percent_split_change; | |
2713 | break; | |
2714 | case 0: | |
2715 | return 1; /* 0% means always write a new shared index */ | |
2716 | case 100: | |
2717 | return 0; /* 100% means never write a new shared index */ | |
2718 | default: | |
2719 | break; /* just use the configured value */ | |
2720 | } | |
2721 | ||
2722 | /* Count not shared entries */ | |
2723 | for (i = 0; i < istate->cache_nr; i++) { | |
2724 | struct cache_entry *ce = istate->cache[i]; | |
2725 | if (!ce->index) | |
2726 | not_shared++; | |
2727 | } | |
2728 | ||
2729 | return (int64_t)istate->cache_nr * max_split < (int64_t)not_shared * 100; | |
2730 | } | |
2731 | ||
2732 | int write_locked_index(struct index_state *istate, struct lock_file *lock, | |
2733 | unsigned flags) | |
2734 | { | |
2735 | int new_shared_index, ret; | |
2736 | struct split_index *si = istate->split_index; | |
2737 | ||
2738 | if (git_env_bool("GIT_TEST_CHECK_CACHE_TREE", 0)) | |
2739 | cache_tree_verify(istate); | |
2740 | ||
2741 | if ((flags & SKIP_IF_UNCHANGED) && !istate->cache_changed) { | |
2742 | if (flags & COMMIT_LOCK) | |
2743 | rollback_lock_file(lock); | |
2744 | return 0; | |
2745 | } | |
2746 | ||
2747 | if (istate->fsmonitor_last_update) | |
2748 | fill_fsmonitor_bitmap(istate); | |
2749 | ||
2750 | if (!si || alternate_index_output || | |
2751 | (istate->cache_changed & ~EXTMASK)) { | |
2752 | if (si) | |
2753 | oidclr(&si->base_oid); | |
2754 | ret = do_write_locked_index(istate, lock, flags); | |
2755 | goto out; | |
2756 | } | |
2757 | ||
2758 | if (git_env_bool("GIT_TEST_SPLIT_INDEX", 0)) { | |
2759 | int v = si->base_oid.hash[0]; | |
2760 | if ((v & 15) < 6) | |
2761 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
2762 | } | |
2763 | if (too_many_not_shared_entries(istate)) | |
2764 | istate->cache_changed |= SPLIT_INDEX_ORDERED; | |
2765 | ||
2766 | new_shared_index = istate->cache_changed & SPLIT_INDEX_ORDERED; | |
2767 | ||
2768 | if (new_shared_index) { | |
2769 | struct tempfile *temp; | |
2770 | int saved_errno; | |
2771 | ||
2772 | temp = mks_tempfile(git_path("sharedindex_XXXXXX")); | |
2773 | if (!temp) { | |
2774 | oidclr(&si->base_oid); | |
2775 | ret = do_write_locked_index(istate, lock, flags); | |
2776 | goto out; | |
2777 | } | |
2778 | ret = write_shared_index(istate, &temp); | |
2779 | ||
2780 | saved_errno = errno; | |
2781 | if (is_tempfile_active(temp)) | |
2782 | delete_tempfile(&temp); | |
2783 | errno = saved_errno; | |
2784 | ||
2785 | if (ret) | |
2786 | goto out; | |
2787 | } | |
2788 | ||
2789 | ret = write_split_index(istate, lock, flags); | |
2790 | ||
2791 | /* Freshen the shared index only if the split-index was written */ | |
2792 | if (!ret && !new_shared_index) { | |
2793 | const char *shared_index = git_path("sharedindex.%s", | |
2794 | oid_to_hex(&si->base_oid)); | |
2795 | freshen_shared_index(shared_index, 1); | |
2796 | } | |
2797 | ||
2798 | out: | |
2799 | if (flags & COMMIT_LOCK) | |
2800 | rollback_lock_file(lock); | |
2801 | return ret; | |
2802 | } | |
2803 | ||
2804 | /* | |
2805 | * Read the index file that is potentially unmerged into given | |
2806 | * index_state, dropping any unmerged entries to stage #0 (potentially | |
2807 | * resulting in a path appearing as both a file and a directory in the | |
2808 | * index; the caller is responsible to clear out the extra entries | |
2809 | * before writing the index to a tree). Returns true if the index is | |
2810 | * unmerged. Callers who want to refuse to work from an unmerged | |
2811 | * state can call this and check its return value, instead of calling | |
2812 | * read_cache(). | |
2813 | */ | |
2814 | int read_index_unmerged(struct index_state *istate) | |
2815 | { | |
2816 | int i; | |
2817 | int unmerged = 0; | |
2818 | ||
2819 | read_index(istate); | |
2820 | for (i = 0; i < istate->cache_nr; i++) { | |
2821 | struct cache_entry *ce = istate->cache[i]; | |
2822 | struct cache_entry *new_ce; | |
2823 | int len; | |
2824 | ||
2825 | if (!ce_stage(ce)) | |
2826 | continue; | |
2827 | unmerged = 1; | |
2828 | len = ce_namelen(ce); | |
2829 | new_ce = make_empty_cache_entry(istate, len); | |
2830 | memcpy(new_ce->name, ce->name, len); | |
2831 | new_ce->ce_flags = create_ce_flags(0) | CE_CONFLICTED; | |
2832 | new_ce->ce_namelen = len; | |
2833 | new_ce->ce_mode = ce->ce_mode; | |
2834 | if (add_index_entry(istate, new_ce, ADD_CACHE_SKIP_DFCHECK)) | |
2835 | return error("%s: cannot drop to stage #0", | |
2836 | new_ce->name); | |
2837 | } | |
2838 | return unmerged; | |
2839 | } | |
2840 | ||
2841 | /* | |
2842 | * Returns 1 if the path is an "other" path with respect to | |
2843 | * the index; that is, the path is not mentioned in the index at all, | |
2844 | * either as a file, a directory with some files in the index, | |
2845 | * or as an unmerged entry. | |
2846 | * | |
2847 | * We helpfully remove a trailing "/" from directories so that | |
2848 | * the output of read_directory can be used as-is. | |
2849 | */ | |
2850 | int index_name_is_other(const struct index_state *istate, const char *name, | |
2851 | int namelen) | |
2852 | { | |
2853 | int pos; | |
2854 | if (namelen && name[namelen - 1] == '/') | |
2855 | namelen--; | |
2856 | pos = index_name_pos(istate, name, namelen); | |
2857 | if (0 <= pos) | |
2858 | return 0; /* exact match */ | |
2859 | pos = -pos - 1; | |
2860 | if (pos < istate->cache_nr) { | |
2861 | struct cache_entry *ce = istate->cache[pos]; | |
2862 | if (ce_namelen(ce) == namelen && | |
2863 | !memcmp(ce->name, name, namelen)) | |
2864 | return 0; /* Yup, this one exists unmerged */ | |
2865 | } | |
2866 | return 1; | |
2867 | } | |
2868 | ||
2869 | void *read_blob_data_from_index(const struct index_state *istate, | |
2870 | const char *path, unsigned long *size) | |
2871 | { | |
2872 | int pos, len; | |
2873 | unsigned long sz; | |
2874 | enum object_type type; | |
2875 | void *data; | |
2876 | ||
2877 | len = strlen(path); | |
2878 | pos = index_name_pos(istate, path, len); | |
2879 | if (pos < 0) { | |
2880 | /* | |
2881 | * We might be in the middle of a merge, in which | |
2882 | * case we would read stage #2 (ours). | |
2883 | */ | |
2884 | int i; | |
2885 | for (i = -pos - 1; | |
2886 | (pos < 0 && i < istate->cache_nr && | |
2887 | !strcmp(istate->cache[i]->name, path)); | |
2888 | i++) | |
2889 | if (ce_stage(istate->cache[i]) == 2) | |
2890 | pos = i; | |
2891 | } | |
2892 | if (pos < 0) | |
2893 | return NULL; | |
2894 | data = read_object_file(&istate->cache[pos]->oid, &type, &sz); | |
2895 | if (!data || type != OBJ_BLOB) { | |
2896 | free(data); | |
2897 | return NULL; | |
2898 | } | |
2899 | if (size) | |
2900 | *size = sz; | |
2901 | return data; | |
2902 | } | |
2903 | ||
2904 | void stat_validity_clear(struct stat_validity *sv) | |
2905 | { | |
2906 | FREE_AND_NULL(sv->sd); | |
2907 | } | |
2908 | ||
2909 | int stat_validity_check(struct stat_validity *sv, const char *path) | |
2910 | { | |
2911 | struct stat st; | |
2912 | ||
2913 | if (stat(path, &st) < 0) | |
2914 | return sv->sd == NULL; | |
2915 | if (!sv->sd) | |
2916 | return 0; | |
2917 | return S_ISREG(st.st_mode) && !match_stat_data(sv->sd, &st); | |
2918 | } | |
2919 | ||
2920 | void stat_validity_update(struct stat_validity *sv, int fd) | |
2921 | { | |
2922 | struct stat st; | |
2923 | ||
2924 | if (fstat(fd, &st) < 0 || !S_ISREG(st.st_mode)) | |
2925 | stat_validity_clear(sv); | |
2926 | else { | |
2927 | if (!sv->sd) | |
2928 | sv->sd = xcalloc(1, sizeof(struct stat_data)); | |
2929 | fill_stat_data(sv->sd, &st); | |
2930 | } | |
2931 | } | |
2932 | ||
2933 | void move_index_extensions(struct index_state *dst, struct index_state *src) | |
2934 | { | |
2935 | dst->untracked = src->untracked; | |
2936 | src->untracked = NULL; | |
2937 | dst->cache_tree = src->cache_tree; | |
2938 | src->cache_tree = NULL; | |
2939 | } | |
2940 | ||
2941 | struct cache_entry *dup_cache_entry(const struct cache_entry *ce, | |
2942 | struct index_state *istate) | |
2943 | { | |
2944 | unsigned int size = ce_size(ce); | |
2945 | int mem_pool_allocated; | |
2946 | struct cache_entry *new_entry = make_empty_cache_entry(istate, ce_namelen(ce)); | |
2947 | mem_pool_allocated = new_entry->mem_pool_allocated; | |
2948 | ||
2949 | memcpy(new_entry, ce, size); | |
2950 | new_entry->mem_pool_allocated = mem_pool_allocated; | |
2951 | return new_entry; | |
2952 | } | |
2953 | ||
2954 | void discard_cache_entry(struct cache_entry *ce) | |
2955 | { | |
2956 | if (ce && should_validate_cache_entries()) | |
2957 | memset(ce, 0xCD, cache_entry_size(ce->ce_namelen)); | |
2958 | ||
2959 | if (ce && ce->mem_pool_allocated) | |
2960 | return; | |
2961 | ||
2962 | free(ce); | |
2963 | } | |
2964 | ||
2965 | int should_validate_cache_entries(void) | |
2966 | { | |
2967 | static int validate_index_cache_entries = -1; | |
2968 | ||
2969 | if (validate_index_cache_entries < 0) { | |
2970 | if (getenv("GIT_TEST_VALIDATE_INDEX_CACHE_ENTRIES")) | |
2971 | validate_index_cache_entries = 1; | |
2972 | else | |
2973 | validate_index_cache_entries = 0; | |
2974 | } | |
2975 | ||
2976 | return validate_index_cache_entries; | |
2977 | } |