]>
Commit | Line | Data |
---|---|---|
1 | #include "cache.h" | |
2 | #include "lockfile.h" | |
3 | #include "tree.h" | |
4 | #include "tree-walk.h" | |
5 | #include "cache-tree.h" | |
6 | #include "object-store.h" | |
7 | #include "replace-object.h" | |
8 | #include "promisor-remote.h" | |
9 | #include "sparse-index.h" | |
10 | ||
11 | #ifndef DEBUG_CACHE_TREE | |
12 | #define DEBUG_CACHE_TREE 0 | |
13 | #endif | |
14 | ||
15 | struct cache_tree *cache_tree(void) | |
16 | { | |
17 | struct cache_tree *it = xcalloc(1, sizeof(struct cache_tree)); | |
18 | it->entry_count = -1; | |
19 | return it; | |
20 | } | |
21 | ||
22 | void cache_tree_free(struct cache_tree **it_p) | |
23 | { | |
24 | int i; | |
25 | struct cache_tree *it = *it_p; | |
26 | ||
27 | if (!it) | |
28 | return; | |
29 | for (i = 0; i < it->subtree_nr; i++) | |
30 | if (it->down[i]) { | |
31 | cache_tree_free(&it->down[i]->cache_tree); | |
32 | free(it->down[i]); | |
33 | } | |
34 | free(it->down); | |
35 | free(it); | |
36 | *it_p = NULL; | |
37 | } | |
38 | ||
39 | static int subtree_name_cmp(const char *one, int onelen, | |
40 | const char *two, int twolen) | |
41 | { | |
42 | if (onelen < twolen) | |
43 | return -1; | |
44 | if (twolen < onelen) | |
45 | return 1; | |
46 | return memcmp(one, two, onelen); | |
47 | } | |
48 | ||
49 | int cache_tree_subtree_pos(struct cache_tree *it, const char *path, int pathlen) | |
50 | { | |
51 | struct cache_tree_sub **down = it->down; | |
52 | int lo, hi; | |
53 | lo = 0; | |
54 | hi = it->subtree_nr; | |
55 | while (lo < hi) { | |
56 | int mi = lo + (hi - lo) / 2; | |
57 | struct cache_tree_sub *mdl = down[mi]; | |
58 | int cmp = subtree_name_cmp(path, pathlen, | |
59 | mdl->name, mdl->namelen); | |
60 | if (!cmp) | |
61 | return mi; | |
62 | if (cmp < 0) | |
63 | hi = mi; | |
64 | else | |
65 | lo = mi + 1; | |
66 | } | |
67 | return -lo-1; | |
68 | } | |
69 | ||
70 | static struct cache_tree_sub *find_subtree(struct cache_tree *it, | |
71 | const char *path, | |
72 | int pathlen, | |
73 | int create) | |
74 | { | |
75 | struct cache_tree_sub *down; | |
76 | int pos = cache_tree_subtree_pos(it, path, pathlen); | |
77 | if (0 <= pos) | |
78 | return it->down[pos]; | |
79 | if (!create) | |
80 | return NULL; | |
81 | ||
82 | pos = -pos-1; | |
83 | ALLOC_GROW(it->down, it->subtree_nr + 1, it->subtree_alloc); | |
84 | it->subtree_nr++; | |
85 | ||
86 | FLEX_ALLOC_MEM(down, name, path, pathlen); | |
87 | down->cache_tree = NULL; | |
88 | down->namelen = pathlen; | |
89 | ||
90 | if (pos < it->subtree_nr) | |
91 | MOVE_ARRAY(it->down + pos + 1, it->down + pos, | |
92 | it->subtree_nr - pos - 1); | |
93 | it->down[pos] = down; | |
94 | return down; | |
95 | } | |
96 | ||
97 | struct cache_tree_sub *cache_tree_sub(struct cache_tree *it, const char *path) | |
98 | { | |
99 | int pathlen = strlen(path); | |
100 | return find_subtree(it, path, pathlen, 1); | |
101 | } | |
102 | ||
103 | static int do_invalidate_path(struct cache_tree *it, const char *path) | |
104 | { | |
105 | /* a/b/c | |
106 | * ==> invalidate self | |
107 | * ==> find "a", have it invalidate "b/c" | |
108 | * a | |
109 | * ==> invalidate self | |
110 | * ==> if "a" exists as a subtree, remove it. | |
111 | */ | |
112 | const char *slash; | |
113 | int namelen; | |
114 | struct cache_tree_sub *down; | |
115 | ||
116 | #if DEBUG_CACHE_TREE | |
117 | fprintf(stderr, "cache-tree invalidate <%s>\n", path); | |
118 | #endif | |
119 | ||
120 | if (!it) | |
121 | return 0; | |
122 | slash = strchrnul(path, '/'); | |
123 | namelen = slash - path; | |
124 | it->entry_count = -1; | |
125 | if (!*slash) { | |
126 | int pos; | |
127 | pos = cache_tree_subtree_pos(it, path, namelen); | |
128 | if (0 <= pos) { | |
129 | cache_tree_free(&it->down[pos]->cache_tree); | |
130 | free(it->down[pos]); | |
131 | /* 0 1 2 3 4 5 | |
132 | * ^ ^subtree_nr = 6 | |
133 | * pos | |
134 | * move 4 and 5 up one place (2 entries) | |
135 | * 2 = 6 - 3 - 1 = subtree_nr - pos - 1 | |
136 | */ | |
137 | MOVE_ARRAY(it->down + pos, it->down + pos + 1, | |
138 | it->subtree_nr - pos - 1); | |
139 | it->subtree_nr--; | |
140 | } | |
141 | return 1; | |
142 | } | |
143 | down = find_subtree(it, path, namelen, 0); | |
144 | if (down) | |
145 | do_invalidate_path(down->cache_tree, slash + 1); | |
146 | return 1; | |
147 | } | |
148 | ||
149 | void cache_tree_invalidate_path(struct index_state *istate, const char *path) | |
150 | { | |
151 | if (do_invalidate_path(istate->cache_tree, path)) | |
152 | istate->cache_changed |= CACHE_TREE_CHANGED; | |
153 | } | |
154 | ||
155 | static int verify_cache(struct index_state *istate, int flags) | |
156 | { | |
157 | unsigned i, funny; | |
158 | int silent = flags & WRITE_TREE_SILENT; | |
159 | ||
160 | /* Verify that the tree is merged */ | |
161 | funny = 0; | |
162 | for (i = 0; i < istate->cache_nr; i++) { | |
163 | const struct cache_entry *ce = istate->cache[i]; | |
164 | if (ce_stage(ce)) { | |
165 | if (silent) | |
166 | return -1; | |
167 | if (10 < ++funny) { | |
168 | fprintf(stderr, "...\n"); | |
169 | break; | |
170 | } | |
171 | fprintf(stderr, "%s: unmerged (%s)\n", | |
172 | ce->name, oid_to_hex(&ce->oid)); | |
173 | } | |
174 | } | |
175 | if (funny) | |
176 | return -1; | |
177 | ||
178 | /* Also verify that the cache does not have path and path/file | |
179 | * at the same time. At this point we know the cache has only | |
180 | * stage 0 entries. | |
181 | */ | |
182 | funny = 0; | |
183 | for (i = 0; i + 1 < istate->cache_nr; i++) { | |
184 | /* path/file always comes after path because of the way | |
185 | * the cache is sorted. Also path can appear only once, | |
186 | * which means conflicting one would immediately follow. | |
187 | */ | |
188 | const struct cache_entry *this_ce = istate->cache[i]; | |
189 | const struct cache_entry *next_ce = istate->cache[i + 1]; | |
190 | const char *this_name = this_ce->name; | |
191 | const char *next_name = next_ce->name; | |
192 | int this_len = ce_namelen(this_ce); | |
193 | if (this_len < ce_namelen(next_ce) && | |
194 | next_name[this_len] == '/' && | |
195 | strncmp(this_name, next_name, this_len) == 0) { | |
196 | if (10 < ++funny) { | |
197 | fprintf(stderr, "...\n"); | |
198 | break; | |
199 | } | |
200 | fprintf(stderr, "You have both %s and %s\n", | |
201 | this_name, next_name); | |
202 | } | |
203 | } | |
204 | if (funny) | |
205 | return -1; | |
206 | return 0; | |
207 | } | |
208 | ||
209 | static void discard_unused_subtrees(struct cache_tree *it) | |
210 | { | |
211 | struct cache_tree_sub **down = it->down; | |
212 | int nr = it->subtree_nr; | |
213 | int dst, src; | |
214 | for (dst = src = 0; src < nr; src++) { | |
215 | struct cache_tree_sub *s = down[src]; | |
216 | if (s->used) | |
217 | down[dst++] = s; | |
218 | else { | |
219 | cache_tree_free(&s->cache_tree); | |
220 | free(s); | |
221 | it->subtree_nr--; | |
222 | } | |
223 | } | |
224 | } | |
225 | ||
226 | int cache_tree_fully_valid(struct cache_tree *it) | |
227 | { | |
228 | int i; | |
229 | if (!it) | |
230 | return 0; | |
231 | if (it->entry_count < 0 || !has_object_file(&it->oid)) | |
232 | return 0; | |
233 | for (i = 0; i < it->subtree_nr; i++) { | |
234 | if (!cache_tree_fully_valid(it->down[i]->cache_tree)) | |
235 | return 0; | |
236 | } | |
237 | return 1; | |
238 | } | |
239 | ||
240 | static int must_check_existence(const struct cache_entry *ce) | |
241 | { | |
242 | return !(has_promisor_remote() && ce_skip_worktree(ce)); | |
243 | } | |
244 | ||
245 | static int update_one(struct cache_tree *it, | |
246 | struct cache_entry **cache, | |
247 | int entries, | |
248 | const char *base, | |
249 | int baselen, | |
250 | int *skip_count, | |
251 | int flags) | |
252 | { | |
253 | struct strbuf buffer; | |
254 | int missing_ok = flags & WRITE_TREE_MISSING_OK; | |
255 | int dryrun = flags & WRITE_TREE_DRY_RUN; | |
256 | int repair = flags & WRITE_TREE_REPAIR; | |
257 | int to_invalidate = 0; | |
258 | int i; | |
259 | ||
260 | assert(!(dryrun && repair)); | |
261 | ||
262 | *skip_count = 0; | |
263 | ||
264 | /* | |
265 | * If the first entry of this region is a sparse directory | |
266 | * entry corresponding exactly to 'base', then this cache_tree | |
267 | * struct is a "leaf" in the data structure, pointing to the | |
268 | * tree OID specified in the entry. | |
269 | */ | |
270 | if (entries > 0) { | |
271 | const struct cache_entry *ce = cache[0]; | |
272 | ||
273 | if (S_ISSPARSEDIR(ce->ce_mode) && | |
274 | ce->ce_namelen == baselen && | |
275 | !strncmp(ce->name, base, baselen)) { | |
276 | it->entry_count = 1; | |
277 | oidcpy(&it->oid, &ce->oid); | |
278 | return 1; | |
279 | } | |
280 | } | |
281 | ||
282 | if (0 <= it->entry_count && has_object_file(&it->oid)) | |
283 | return it->entry_count; | |
284 | ||
285 | /* | |
286 | * We first scan for subtrees and update them; we start by | |
287 | * marking existing subtrees -- the ones that are unmarked | |
288 | * should not be in the result. | |
289 | */ | |
290 | for (i = 0; i < it->subtree_nr; i++) | |
291 | it->down[i]->used = 0; | |
292 | ||
293 | /* | |
294 | * Find the subtrees and update them. | |
295 | */ | |
296 | i = 0; | |
297 | while (i < entries) { | |
298 | const struct cache_entry *ce = cache[i]; | |
299 | struct cache_tree_sub *sub; | |
300 | const char *path, *slash; | |
301 | int pathlen, sublen, subcnt, subskip; | |
302 | ||
303 | path = ce->name; | |
304 | pathlen = ce_namelen(ce); | |
305 | if (pathlen <= baselen || memcmp(base, path, baselen)) | |
306 | break; /* at the end of this level */ | |
307 | ||
308 | slash = strchr(path + baselen, '/'); | |
309 | if (!slash) { | |
310 | i++; | |
311 | continue; | |
312 | } | |
313 | /* | |
314 | * a/bbb/c (base = a/, slash = /c) | |
315 | * ==> | |
316 | * path+baselen = bbb/c, sublen = 3 | |
317 | */ | |
318 | sublen = slash - (path + baselen); | |
319 | sub = find_subtree(it, path + baselen, sublen, 1); | |
320 | if (!sub->cache_tree) | |
321 | sub->cache_tree = cache_tree(); | |
322 | subcnt = update_one(sub->cache_tree, | |
323 | cache + i, entries - i, | |
324 | path, | |
325 | baselen + sublen + 1, | |
326 | &subskip, | |
327 | flags); | |
328 | if (subcnt < 0) | |
329 | return subcnt; | |
330 | if (!subcnt) | |
331 | die("index cache-tree records empty sub-tree"); | |
332 | i += subcnt; | |
333 | sub->count = subcnt; /* to be used in the next loop */ | |
334 | *skip_count += subskip; | |
335 | sub->used = 1; | |
336 | } | |
337 | ||
338 | discard_unused_subtrees(it); | |
339 | ||
340 | /* | |
341 | * Then write out the tree object for this level. | |
342 | */ | |
343 | strbuf_init(&buffer, 8192); | |
344 | ||
345 | i = 0; | |
346 | while (i < entries) { | |
347 | const struct cache_entry *ce = cache[i]; | |
348 | struct cache_tree_sub *sub = NULL; | |
349 | const char *path, *slash; | |
350 | int pathlen, entlen; | |
351 | const struct object_id *oid; | |
352 | unsigned mode; | |
353 | int expected_missing = 0; | |
354 | int contains_ita = 0; | |
355 | int ce_missing_ok; | |
356 | ||
357 | path = ce->name; | |
358 | pathlen = ce_namelen(ce); | |
359 | if (pathlen <= baselen || memcmp(base, path, baselen)) | |
360 | break; /* at the end of this level */ | |
361 | ||
362 | slash = strchr(path + baselen, '/'); | |
363 | if (slash) { | |
364 | entlen = slash - (path + baselen); | |
365 | sub = find_subtree(it, path + baselen, entlen, 0); | |
366 | if (!sub) | |
367 | die("cache-tree.c: '%.*s' in '%s' not found", | |
368 | entlen, path + baselen, path); | |
369 | i += sub->count; | |
370 | oid = &sub->cache_tree->oid; | |
371 | mode = S_IFDIR; | |
372 | contains_ita = sub->cache_tree->entry_count < 0; | |
373 | if (contains_ita) { | |
374 | to_invalidate = 1; | |
375 | expected_missing = 1; | |
376 | } | |
377 | } | |
378 | else { | |
379 | oid = &ce->oid; | |
380 | mode = ce->ce_mode; | |
381 | entlen = pathlen - baselen; | |
382 | i++; | |
383 | } | |
384 | ||
385 | ce_missing_ok = mode == S_IFGITLINK || missing_ok || | |
386 | !must_check_existence(ce); | |
387 | if (is_null_oid(oid) || | |
388 | (!ce_missing_ok && !has_object_file(oid))) { | |
389 | strbuf_release(&buffer); | |
390 | if (expected_missing) | |
391 | return -1; | |
392 | return error("invalid object %06o %s for '%.*s'", | |
393 | mode, oid_to_hex(oid), entlen+baselen, path); | |
394 | } | |
395 | ||
396 | /* | |
397 | * CE_REMOVE entries are removed before the index is | |
398 | * written to disk. Skip them to remain consistent | |
399 | * with the future on-disk index. | |
400 | */ | |
401 | if (ce->ce_flags & CE_REMOVE) { | |
402 | *skip_count = *skip_count + 1; | |
403 | continue; | |
404 | } | |
405 | ||
406 | /* | |
407 | * CE_INTENT_TO_ADD entries exist on on-disk index but | |
408 | * they are not part of generated trees. Invalidate up | |
409 | * to root to force cache-tree users to read elsewhere. | |
410 | */ | |
411 | if (!sub && ce_intent_to_add(ce)) { | |
412 | to_invalidate = 1; | |
413 | continue; | |
414 | } | |
415 | ||
416 | /* | |
417 | * "sub" can be an empty tree if all subentries are i-t-a. | |
418 | */ | |
419 | if (contains_ita && is_empty_tree_oid(oid)) | |
420 | continue; | |
421 | ||
422 | strbuf_grow(&buffer, entlen + 100); | |
423 | strbuf_addf(&buffer, "%o %.*s%c", mode, entlen, path + baselen, '\0'); | |
424 | strbuf_add(&buffer, oid->hash, the_hash_algo->rawsz); | |
425 | ||
426 | #if DEBUG_CACHE_TREE | |
427 | fprintf(stderr, "cache-tree update-one %o %.*s\n", | |
428 | mode, entlen, path + baselen); | |
429 | #endif | |
430 | } | |
431 | ||
432 | if (repair) { | |
433 | struct object_id oid; | |
434 | hash_object_file(the_hash_algo, buffer.buf, buffer.len, | |
435 | tree_type, &oid); | |
436 | if (has_object_file_with_flags(&oid, OBJECT_INFO_SKIP_FETCH_OBJECT)) | |
437 | oidcpy(&it->oid, &oid); | |
438 | else | |
439 | to_invalidate = 1; | |
440 | } else if (dryrun) { | |
441 | hash_object_file(the_hash_algo, buffer.buf, buffer.len, | |
442 | tree_type, &it->oid); | |
443 | } else if (write_object_file_flags(buffer.buf, buffer.len, tree_type, | |
444 | &it->oid, flags & WRITE_TREE_SILENT | |
445 | ? HASH_SILENT : 0)) { | |
446 | strbuf_release(&buffer); | |
447 | return -1; | |
448 | } | |
449 | ||
450 | strbuf_release(&buffer); | |
451 | it->entry_count = to_invalidate ? -1 : i - *skip_count; | |
452 | #if DEBUG_CACHE_TREE | |
453 | fprintf(stderr, "cache-tree update-one (%d ent, %d subtree) %s\n", | |
454 | it->entry_count, it->subtree_nr, | |
455 | oid_to_hex(&it->oid)); | |
456 | #endif | |
457 | return i; | |
458 | } | |
459 | ||
460 | int cache_tree_update(struct index_state *istate, int flags) | |
461 | { | |
462 | int skip, i; | |
463 | ||
464 | i = verify_cache(istate, flags); | |
465 | ||
466 | if (i) | |
467 | return i; | |
468 | ||
469 | if (!istate->cache_tree) | |
470 | istate->cache_tree = cache_tree(); | |
471 | ||
472 | if (!(flags & WRITE_TREE_MISSING_OK) && has_promisor_remote()) | |
473 | prefetch_cache_entries(istate, must_check_existence); | |
474 | ||
475 | trace_performance_enter(); | |
476 | trace2_region_enter("cache_tree", "update", the_repository); | |
477 | i = update_one(istate->cache_tree, istate->cache, istate->cache_nr, | |
478 | "", 0, &skip, flags); | |
479 | trace2_region_leave("cache_tree", "update", the_repository); | |
480 | trace_performance_leave("cache_tree_update"); | |
481 | if (i < 0) | |
482 | return i; | |
483 | istate->cache_changed |= CACHE_TREE_CHANGED; | |
484 | return 0; | |
485 | } | |
486 | ||
487 | static void write_one(struct strbuf *buffer, struct cache_tree *it, | |
488 | const char *path, int pathlen) | |
489 | { | |
490 | int i; | |
491 | ||
492 | /* One "cache-tree" entry consists of the following: | |
493 | * path (NUL terminated) | |
494 | * entry_count, subtree_nr ("%d %d\n") | |
495 | * tree-sha1 (missing if invalid) | |
496 | * subtree_nr "cache-tree" entries for subtrees. | |
497 | */ | |
498 | strbuf_grow(buffer, pathlen + 100); | |
499 | strbuf_add(buffer, path, pathlen); | |
500 | strbuf_addf(buffer, "%c%d %d\n", 0, it->entry_count, it->subtree_nr); | |
501 | ||
502 | #if DEBUG_CACHE_TREE | |
503 | if (0 <= it->entry_count) | |
504 | fprintf(stderr, "cache-tree <%.*s> (%d ent, %d subtree) %s\n", | |
505 | pathlen, path, it->entry_count, it->subtree_nr, | |
506 | oid_to_hex(&it->oid)); | |
507 | else | |
508 | fprintf(stderr, "cache-tree <%.*s> (%d subtree) invalid\n", | |
509 | pathlen, path, it->subtree_nr); | |
510 | #endif | |
511 | ||
512 | if (0 <= it->entry_count) { | |
513 | strbuf_add(buffer, it->oid.hash, the_hash_algo->rawsz); | |
514 | } | |
515 | for (i = 0; i < it->subtree_nr; i++) { | |
516 | struct cache_tree_sub *down = it->down[i]; | |
517 | if (i) { | |
518 | struct cache_tree_sub *prev = it->down[i-1]; | |
519 | if (subtree_name_cmp(down->name, down->namelen, | |
520 | prev->name, prev->namelen) <= 0) | |
521 | die("fatal - unsorted cache subtree"); | |
522 | } | |
523 | write_one(buffer, down->cache_tree, down->name, down->namelen); | |
524 | } | |
525 | } | |
526 | ||
527 | void cache_tree_write(struct strbuf *sb, struct cache_tree *root) | |
528 | { | |
529 | trace2_region_enter("cache_tree", "write", the_repository); | |
530 | write_one(sb, root, "", 0); | |
531 | trace2_region_leave("cache_tree", "write", the_repository); | |
532 | } | |
533 | ||
534 | static struct cache_tree *read_one(const char **buffer, unsigned long *size_p) | |
535 | { | |
536 | const char *buf = *buffer; | |
537 | unsigned long size = *size_p; | |
538 | const char *cp; | |
539 | char *ep; | |
540 | struct cache_tree *it; | |
541 | int i, subtree_nr; | |
542 | const unsigned rawsz = the_hash_algo->rawsz; | |
543 | ||
544 | it = NULL; | |
545 | /* skip name, but make sure name exists */ | |
546 | while (size && *buf) { | |
547 | size--; | |
548 | buf++; | |
549 | } | |
550 | if (!size) | |
551 | goto free_return; | |
552 | buf++; size--; | |
553 | it = cache_tree(); | |
554 | ||
555 | cp = buf; | |
556 | it->entry_count = strtol(cp, &ep, 10); | |
557 | if (cp == ep) | |
558 | goto free_return; | |
559 | cp = ep; | |
560 | subtree_nr = strtol(cp, &ep, 10); | |
561 | if (cp == ep) | |
562 | goto free_return; | |
563 | while (size && *buf && *buf != '\n') { | |
564 | size--; | |
565 | buf++; | |
566 | } | |
567 | if (!size) | |
568 | goto free_return; | |
569 | buf++; size--; | |
570 | if (0 <= it->entry_count) { | |
571 | if (size < rawsz) | |
572 | goto free_return; | |
573 | oidread(&it->oid, (const unsigned char *)buf); | |
574 | buf += rawsz; | |
575 | size -= rawsz; | |
576 | } | |
577 | ||
578 | #if DEBUG_CACHE_TREE | |
579 | if (0 <= it->entry_count) | |
580 | fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n", | |
581 | *buffer, it->entry_count, subtree_nr, | |
582 | oid_to_hex(&it->oid)); | |
583 | else | |
584 | fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n", | |
585 | *buffer, subtree_nr); | |
586 | #endif | |
587 | ||
588 | /* | |
589 | * Just a heuristic -- we do not add directories that often but | |
590 | * we do not want to have to extend it immediately when we do, | |
591 | * hence +2. | |
592 | */ | |
593 | it->subtree_alloc = subtree_nr + 2; | |
594 | CALLOC_ARRAY(it->down, it->subtree_alloc); | |
595 | for (i = 0; i < subtree_nr; i++) { | |
596 | /* read each subtree */ | |
597 | struct cache_tree *sub; | |
598 | struct cache_tree_sub *subtree; | |
599 | const char *name = buf; | |
600 | ||
601 | sub = read_one(&buf, &size); | |
602 | if (!sub) | |
603 | goto free_return; | |
604 | subtree = cache_tree_sub(it, name); | |
605 | subtree->cache_tree = sub; | |
606 | } | |
607 | if (subtree_nr != it->subtree_nr) | |
608 | die("cache-tree: internal error"); | |
609 | *buffer = buf; | |
610 | *size_p = size; | |
611 | return it; | |
612 | ||
613 | free_return: | |
614 | cache_tree_free(&it); | |
615 | return NULL; | |
616 | } | |
617 | ||
618 | struct cache_tree *cache_tree_read(const char *buffer, unsigned long size) | |
619 | { | |
620 | struct cache_tree *result; | |
621 | ||
622 | if (buffer[0]) | |
623 | return NULL; /* not the whole tree */ | |
624 | ||
625 | trace2_region_enter("cache_tree", "read", the_repository); | |
626 | result = read_one(&buffer, &size); | |
627 | trace2_region_leave("cache_tree", "read", the_repository); | |
628 | ||
629 | return result; | |
630 | } | |
631 | ||
632 | static struct cache_tree *cache_tree_find(struct cache_tree *it, const char *path) | |
633 | { | |
634 | if (!it) | |
635 | return NULL; | |
636 | while (*path) { | |
637 | const char *slash; | |
638 | struct cache_tree_sub *sub; | |
639 | ||
640 | slash = strchrnul(path, '/'); | |
641 | /* | |
642 | * Between path and slash is the name of the subtree | |
643 | * to look for. | |
644 | */ | |
645 | sub = find_subtree(it, path, slash - path, 0); | |
646 | if (!sub) | |
647 | return NULL; | |
648 | it = sub->cache_tree; | |
649 | ||
650 | path = slash; | |
651 | while (*path == '/') | |
652 | path++; | |
653 | } | |
654 | return it; | |
655 | } | |
656 | ||
657 | static int write_index_as_tree_internal(struct object_id *oid, | |
658 | struct index_state *index_state, | |
659 | int cache_tree_valid, | |
660 | int flags, | |
661 | const char *prefix) | |
662 | { | |
663 | if (flags & WRITE_TREE_IGNORE_CACHE_TREE) { | |
664 | cache_tree_free(&index_state->cache_tree); | |
665 | cache_tree_valid = 0; | |
666 | } | |
667 | ||
668 | if (!cache_tree_valid && cache_tree_update(index_state, flags) < 0) | |
669 | return WRITE_TREE_UNMERGED_INDEX; | |
670 | ||
671 | if (prefix) { | |
672 | struct cache_tree *subtree; | |
673 | subtree = cache_tree_find(index_state->cache_tree, prefix); | |
674 | if (!subtree) | |
675 | return WRITE_TREE_PREFIX_ERROR; | |
676 | oidcpy(oid, &subtree->oid); | |
677 | } | |
678 | else | |
679 | oidcpy(oid, &index_state->cache_tree->oid); | |
680 | ||
681 | return 0; | |
682 | } | |
683 | ||
684 | struct tree* write_in_core_index_as_tree(struct repository *repo) { | |
685 | struct object_id o; | |
686 | int was_valid, ret; | |
687 | ||
688 | struct index_state *index_state = repo->index; | |
689 | was_valid = index_state->cache_tree && | |
690 | cache_tree_fully_valid(index_state->cache_tree); | |
691 | ||
692 | ret = write_index_as_tree_internal(&o, index_state, was_valid, 0, NULL); | |
693 | if (ret == WRITE_TREE_UNMERGED_INDEX) { | |
694 | int i; | |
695 | fprintf(stderr, "BUG: There are unmerged index entries:\n"); | |
696 | for (i = 0; i < index_state->cache_nr; i++) { | |
697 | const struct cache_entry *ce = index_state->cache[i]; | |
698 | if (ce_stage(ce)) | |
699 | fprintf(stderr, "BUG: %d %.*s\n", ce_stage(ce), | |
700 | (int)ce_namelen(ce), ce->name); | |
701 | } | |
702 | BUG("unmerged index entries when writing inmemory index"); | |
703 | } | |
704 | ||
705 | return lookup_tree(repo, &index_state->cache_tree->oid); | |
706 | } | |
707 | ||
708 | ||
709 | int write_index_as_tree(struct object_id *oid, struct index_state *index_state, const char *index_path, int flags, const char *prefix) | |
710 | { | |
711 | int entries, was_valid; | |
712 | struct lock_file lock_file = LOCK_INIT; | |
713 | int ret; | |
714 | ||
715 | hold_lock_file_for_update(&lock_file, index_path, LOCK_DIE_ON_ERROR); | |
716 | ||
717 | entries = read_index_from(index_state, index_path, get_git_dir()); | |
718 | if (entries < 0) { | |
719 | ret = WRITE_TREE_UNREADABLE_INDEX; | |
720 | goto out; | |
721 | } | |
722 | ||
723 | was_valid = !(flags & WRITE_TREE_IGNORE_CACHE_TREE) && | |
724 | index_state->cache_tree && | |
725 | cache_tree_fully_valid(index_state->cache_tree); | |
726 | ||
727 | ret = write_index_as_tree_internal(oid, index_state, was_valid, flags, | |
728 | prefix); | |
729 | if (!ret && !was_valid) { | |
730 | write_locked_index(index_state, &lock_file, COMMIT_LOCK); | |
731 | /* Not being able to write is fine -- we are only interested | |
732 | * in updating the cache-tree part, and if the next caller | |
733 | * ends up using the old index with unupdated cache-tree part | |
734 | * it misses the work we did here, but that is just a | |
735 | * performance penalty and not a big deal. | |
736 | */ | |
737 | } | |
738 | ||
739 | out: | |
740 | rollback_lock_file(&lock_file); | |
741 | return ret; | |
742 | } | |
743 | ||
744 | static void prime_cache_tree_sparse_dir(struct cache_tree *it, | |
745 | struct tree *tree) | |
746 | { | |
747 | ||
748 | oidcpy(&it->oid, &tree->object.oid); | |
749 | it->entry_count = 1; | |
750 | } | |
751 | ||
752 | static void prime_cache_tree_rec(struct repository *r, | |
753 | struct cache_tree *it, | |
754 | struct tree *tree, | |
755 | struct strbuf *tree_path) | |
756 | { | |
757 | struct tree_desc desc; | |
758 | struct name_entry entry; | |
759 | int cnt; | |
760 | int base_path_len = tree_path->len; | |
761 | ||
762 | oidcpy(&it->oid, &tree->object.oid); | |
763 | ||
764 | init_tree_desc(&desc, tree->buffer, tree->size); | |
765 | cnt = 0; | |
766 | while (tree_entry(&desc, &entry)) { | |
767 | if (!S_ISDIR(entry.mode)) | |
768 | cnt++; | |
769 | else { | |
770 | struct cache_tree_sub *sub; | |
771 | struct tree *subtree = lookup_tree(r, &entry.oid); | |
772 | ||
773 | if (!subtree->object.parsed) | |
774 | parse_tree(subtree); | |
775 | sub = cache_tree_sub(it, entry.path); | |
776 | sub->cache_tree = cache_tree(); | |
777 | ||
778 | /* | |
779 | * Recursively-constructed subtree path is only needed when working | |
780 | * in a sparse index (where it's used to determine whether the | |
781 | * subtree is a sparse directory in the index). | |
782 | */ | |
783 | if (r->index->sparse_index) { | |
784 | strbuf_setlen(tree_path, base_path_len); | |
785 | strbuf_grow(tree_path, base_path_len + entry.pathlen + 1); | |
786 | strbuf_add(tree_path, entry.path, entry.pathlen); | |
787 | strbuf_addch(tree_path, '/'); | |
788 | } | |
789 | ||
790 | /* | |
791 | * If a sparse index is in use, the directory being processed may be | |
792 | * sparse. To confirm that, we can check whether an entry with that | |
793 | * exact name exists in the index. If it does, the created subtree | |
794 | * should be sparse. Otherwise, cache tree expansion should continue | |
795 | * as normal. | |
796 | */ | |
797 | if (r->index->sparse_index && | |
798 | index_entry_exists(r->index, tree_path->buf, tree_path->len)) | |
799 | prime_cache_tree_sparse_dir(sub->cache_tree, subtree); | |
800 | else | |
801 | prime_cache_tree_rec(r, sub->cache_tree, subtree, tree_path); | |
802 | cnt += sub->cache_tree->entry_count; | |
803 | } | |
804 | } | |
805 | ||
806 | it->entry_count = cnt; | |
807 | } | |
808 | ||
809 | void prime_cache_tree(struct repository *r, | |
810 | struct index_state *istate, | |
811 | struct tree *tree) | |
812 | { | |
813 | struct strbuf tree_path = STRBUF_INIT; | |
814 | ||
815 | trace2_region_enter("cache-tree", "prime_cache_tree", the_repository); | |
816 | cache_tree_free(&istate->cache_tree); | |
817 | istate->cache_tree = cache_tree(); | |
818 | ||
819 | prime_cache_tree_rec(r, istate->cache_tree, tree, &tree_path); | |
820 | strbuf_release(&tree_path); | |
821 | istate->cache_changed |= CACHE_TREE_CHANGED; | |
822 | trace2_region_leave("cache-tree", "prime_cache_tree", the_repository); | |
823 | } | |
824 | ||
825 | /* | |
826 | * find the cache_tree that corresponds to the current level without | |
827 | * exploding the full path into textual form. The root of the | |
828 | * cache tree is given as "root", and our current level is "info". | |
829 | * (1) When at root level, info->prev is NULL, so it is "root" itself. | |
830 | * (2) Otherwise, find the cache_tree that corresponds to one level | |
831 | * above us, and find ourselves in there. | |
832 | */ | |
833 | static struct cache_tree *find_cache_tree_from_traversal(struct cache_tree *root, | |
834 | struct traverse_info *info) | |
835 | { | |
836 | struct cache_tree *our_parent; | |
837 | ||
838 | if (!info->prev) | |
839 | return root; | |
840 | our_parent = find_cache_tree_from_traversal(root, info->prev); | |
841 | return cache_tree_find(our_parent, info->name); | |
842 | } | |
843 | ||
844 | int cache_tree_matches_traversal(struct cache_tree *root, | |
845 | struct name_entry *ent, | |
846 | struct traverse_info *info) | |
847 | { | |
848 | struct cache_tree *it; | |
849 | ||
850 | it = find_cache_tree_from_traversal(root, info); | |
851 | it = cache_tree_find(it, ent->path); | |
852 | if (it && it->entry_count > 0 && oideq(&ent->oid, &it->oid)) | |
853 | return it->entry_count; | |
854 | return 0; | |
855 | } | |
856 | ||
857 | static void verify_one_sparse(struct repository *r, | |
858 | struct index_state *istate, | |
859 | struct cache_tree *it, | |
860 | struct strbuf *path, | |
861 | int pos) | |
862 | { | |
863 | struct cache_entry *ce = istate->cache[pos]; | |
864 | ||
865 | if (!S_ISSPARSEDIR(ce->ce_mode)) | |
866 | BUG("directory '%s' is present in index, but not sparse", | |
867 | path->buf); | |
868 | } | |
869 | ||
870 | /* | |
871 | * Returns: | |
872 | * 0 - Verification completed. | |
873 | * 1 - Restart verification - a call to ensure_full_index() freed the cache | |
874 | * tree that is being verified and verification needs to be restarted from | |
875 | * the new toplevel cache tree. | |
876 | */ | |
877 | static int verify_one(struct repository *r, | |
878 | struct index_state *istate, | |
879 | struct cache_tree *it, | |
880 | struct strbuf *path) | |
881 | { | |
882 | int i, pos, len = path->len; | |
883 | struct strbuf tree_buf = STRBUF_INIT; | |
884 | struct object_id new_oid; | |
885 | ||
886 | for (i = 0; i < it->subtree_nr; i++) { | |
887 | strbuf_addf(path, "%s/", it->down[i]->name); | |
888 | if (verify_one(r, istate, it->down[i]->cache_tree, path)) | |
889 | return 1; | |
890 | strbuf_setlen(path, len); | |
891 | } | |
892 | ||
893 | if (it->entry_count < 0 || | |
894 | /* no verification on tests (t7003) that replace trees */ | |
895 | lookup_replace_object(r, &it->oid) != &it->oid) | |
896 | return 0; | |
897 | ||
898 | if (path->len) { | |
899 | /* | |
900 | * If the index is sparse and the cache tree is not | |
901 | * index_name_pos() may trigger ensure_full_index() which will | |
902 | * free the tree that is being verified. | |
903 | */ | |
904 | int is_sparse = istate->sparse_index; | |
905 | pos = index_name_pos(istate, path->buf, path->len); | |
906 | if (is_sparse && !istate->sparse_index) | |
907 | return 1; | |
908 | ||
909 | if (pos >= 0) { | |
910 | verify_one_sparse(r, istate, it, path, pos); | |
911 | return 0; | |
912 | } | |
913 | ||
914 | pos = -pos - 1; | |
915 | } else { | |
916 | pos = 0; | |
917 | } | |
918 | ||
919 | i = 0; | |
920 | while (i < it->entry_count) { | |
921 | struct cache_entry *ce = istate->cache[pos + i]; | |
922 | const char *slash; | |
923 | struct cache_tree_sub *sub = NULL; | |
924 | const struct object_id *oid; | |
925 | const char *name; | |
926 | unsigned mode; | |
927 | int entlen; | |
928 | ||
929 | if (ce->ce_flags & (CE_STAGEMASK | CE_INTENT_TO_ADD | CE_REMOVE)) | |
930 | BUG("%s with flags 0x%x should not be in cache-tree", | |
931 | ce->name, ce->ce_flags); | |
932 | name = ce->name + path->len; | |
933 | slash = strchr(name, '/'); | |
934 | if (slash) { | |
935 | entlen = slash - name; | |
936 | sub = find_subtree(it, ce->name + path->len, entlen, 0); | |
937 | if (!sub || sub->cache_tree->entry_count < 0) | |
938 | BUG("bad subtree '%.*s'", entlen, name); | |
939 | oid = &sub->cache_tree->oid; | |
940 | mode = S_IFDIR; | |
941 | i += sub->cache_tree->entry_count; | |
942 | } else { | |
943 | oid = &ce->oid; | |
944 | mode = ce->ce_mode; | |
945 | entlen = ce_namelen(ce) - path->len; | |
946 | i++; | |
947 | } | |
948 | strbuf_addf(&tree_buf, "%o %.*s%c", mode, entlen, name, '\0'); | |
949 | strbuf_add(&tree_buf, oid->hash, r->hash_algo->rawsz); | |
950 | } | |
951 | hash_object_file(r->hash_algo, tree_buf.buf, tree_buf.len, tree_type, | |
952 | &new_oid); | |
953 | if (!oideq(&new_oid, &it->oid)) | |
954 | BUG("cache-tree for path %.*s does not match. " | |
955 | "Expected %s got %s", len, path->buf, | |
956 | oid_to_hex(&new_oid), oid_to_hex(&it->oid)); | |
957 | strbuf_setlen(path, len); | |
958 | strbuf_release(&tree_buf); | |
959 | return 0; | |
960 | } | |
961 | ||
962 | void cache_tree_verify(struct repository *r, struct index_state *istate) | |
963 | { | |
964 | struct strbuf path = STRBUF_INIT; | |
965 | ||
966 | if (!istate->cache_tree) | |
967 | return; | |
968 | if (verify_one(r, istate, istate->cache_tree, &path)) { | |
969 | strbuf_reset(&path); | |
970 | if (verify_one(r, istate, istate->cache_tree, &path)) | |
971 | BUG("ensure_full_index() called twice while verifying cache tree"); | |
972 | } | |
973 | strbuf_release(&path); | |
974 | } |