t/test-lib: introduce OID_REGEX
[git/debian.git] / cache-tree.c
blob6a555f4d431f9f6dbf8dad06d75a4ec81a4254fd
1 #include "cache.h"
2 #include "lockfile.h"
3 #include "tree.h"
4 #include "tree-walk.h"
5 #include "cache-tree.h"
7 #ifndef DEBUG
8 #define DEBUG 0
9 #endif
11 struct cache_tree *cache_tree(void)
13 struct cache_tree *it = xcalloc(1, sizeof(struct cache_tree));
14 it->entry_count = -1;
15 return it;
18 void cache_tree_free(struct cache_tree **it_p)
20 int i;
21 struct cache_tree *it = *it_p;
23 if (!it)
24 return;
25 for (i = 0; i < it->subtree_nr; i++)
26 if (it->down[i]) {
27 cache_tree_free(&it->down[i]->cache_tree);
28 free(it->down[i]);
30 free(it->down);
31 free(it);
32 *it_p = NULL;
35 static int subtree_name_cmp(const char *one, int onelen,
36 const char *two, int twolen)
38 if (onelen < twolen)
39 return -1;
40 if (twolen < onelen)
41 return 1;
42 return memcmp(one, two, onelen);
45 static int subtree_pos(struct cache_tree *it, const char *path, int pathlen)
47 struct cache_tree_sub **down = it->down;
48 int lo, hi;
49 lo = 0;
50 hi = it->subtree_nr;
51 while (lo < hi) {
52 int mi = lo + (hi - lo) / 2;
53 struct cache_tree_sub *mdl = down[mi];
54 int cmp = subtree_name_cmp(path, pathlen,
55 mdl->name, mdl->namelen);
56 if (!cmp)
57 return mi;
58 if (cmp < 0)
59 hi = mi;
60 else
61 lo = mi + 1;
63 return -lo-1;
66 static struct cache_tree_sub *find_subtree(struct cache_tree *it,
67 const char *path,
68 int pathlen,
69 int create)
71 struct cache_tree_sub *down;
72 int pos = subtree_pos(it, path, pathlen);
73 if (0 <= pos)
74 return it->down[pos];
75 if (!create)
76 return NULL;
78 pos = -pos-1;
79 ALLOC_GROW(it->down, it->subtree_nr + 1, it->subtree_alloc);
80 it->subtree_nr++;
82 FLEX_ALLOC_MEM(down, name, path, pathlen);
83 down->cache_tree = NULL;
84 down->namelen = pathlen;
86 if (pos < it->subtree_nr)
87 MOVE_ARRAY(it->down + pos + 1, it->down + pos,
88 it->subtree_nr - pos - 1);
89 it->down[pos] = down;
90 return down;
93 struct cache_tree_sub *cache_tree_sub(struct cache_tree *it, const char *path)
95 int pathlen = strlen(path);
96 return find_subtree(it, path, pathlen, 1);
99 static int do_invalidate_path(struct cache_tree *it, const char *path)
101 /* a/b/c
102 * ==> invalidate self
103 * ==> find "a", have it invalidate "b/c"
105 * ==> invalidate self
106 * ==> if "a" exists as a subtree, remove it.
108 const char *slash;
109 int namelen;
110 struct cache_tree_sub *down;
112 #if DEBUG
113 fprintf(stderr, "cache-tree invalidate <%s>\n", path);
114 #endif
116 if (!it)
117 return 0;
118 slash = strchrnul(path, '/');
119 namelen = slash - path;
120 it->entry_count = -1;
121 if (!*slash) {
122 int pos;
123 pos = subtree_pos(it, path, namelen);
124 if (0 <= pos) {
125 cache_tree_free(&it->down[pos]->cache_tree);
126 free(it->down[pos]);
127 /* 0 1 2 3 4 5
128 * ^ ^subtree_nr = 6
129 * pos
130 * move 4 and 5 up one place (2 entries)
131 * 2 = 6 - 3 - 1 = subtree_nr - pos - 1
133 MOVE_ARRAY(it->down + pos, it->down + pos + 1,
134 it->subtree_nr - pos - 1);
135 it->subtree_nr--;
137 return 1;
139 down = find_subtree(it, path, namelen, 0);
140 if (down)
141 do_invalidate_path(down->cache_tree, slash + 1);
142 return 1;
145 void cache_tree_invalidate_path(struct index_state *istate, const char *path)
147 if (do_invalidate_path(istate->cache_tree, path))
148 istate->cache_changed |= CACHE_TREE_CHANGED;
151 static int verify_cache(struct cache_entry **cache,
152 int entries, int flags)
154 int i, funny;
155 int silent = flags & WRITE_TREE_SILENT;
157 /* Verify that the tree is merged */
158 funny = 0;
159 for (i = 0; i < entries; i++) {
160 const struct cache_entry *ce = cache[i];
161 if (ce_stage(ce)) {
162 if (silent)
163 return -1;
164 if (10 < ++funny) {
165 fprintf(stderr, "...\n");
166 break;
168 fprintf(stderr, "%s: unmerged (%s)\n",
169 ce->name, oid_to_hex(&ce->oid));
172 if (funny)
173 return -1;
175 /* Also verify that the cache does not have path and path/file
176 * at the same time. At this point we know the cache has only
177 * stage 0 entries.
179 funny = 0;
180 for (i = 0; i < entries - 1; i++) {
181 /* path/file always comes after path because of the way
182 * the cache is sorted. Also path can appear only once,
183 * which means conflicting one would immediately follow.
185 const char *this_name = cache[i]->name;
186 const char *next_name = cache[i+1]->name;
187 int this_len = strlen(this_name);
188 if (this_len < strlen(next_name) &&
189 strncmp(this_name, next_name, this_len) == 0 &&
190 next_name[this_len] == '/') {
191 if (10 < ++funny) {
192 fprintf(stderr, "...\n");
193 break;
195 fprintf(stderr, "You have both %s and %s\n",
196 this_name, next_name);
199 if (funny)
200 return -1;
201 return 0;
204 static void discard_unused_subtrees(struct cache_tree *it)
206 struct cache_tree_sub **down = it->down;
207 int nr = it->subtree_nr;
208 int dst, src;
209 for (dst = src = 0; src < nr; src++) {
210 struct cache_tree_sub *s = down[src];
211 if (s->used)
212 down[dst++] = s;
213 else {
214 cache_tree_free(&s->cache_tree);
215 free(s);
216 it->subtree_nr--;
221 int cache_tree_fully_valid(struct cache_tree *it)
223 int i;
224 if (!it)
225 return 0;
226 if (it->entry_count < 0 || !has_sha1_file(it->oid.hash))
227 return 0;
228 for (i = 0; i < it->subtree_nr; i++) {
229 if (!cache_tree_fully_valid(it->down[i]->cache_tree))
230 return 0;
232 return 1;
235 static int update_one(struct cache_tree *it,
236 struct cache_entry **cache,
237 int entries,
238 const char *base,
239 int baselen,
240 int *skip_count,
241 int flags)
243 struct strbuf buffer;
244 int missing_ok = flags & WRITE_TREE_MISSING_OK;
245 int dryrun = flags & WRITE_TREE_DRY_RUN;
246 int repair = flags & WRITE_TREE_REPAIR;
247 int to_invalidate = 0;
248 int i;
250 assert(!(dryrun && repair));
252 *skip_count = 0;
254 if (0 <= it->entry_count && has_sha1_file(it->oid.hash))
255 return it->entry_count;
258 * We first scan for subtrees and update them; we start by
259 * marking existing subtrees -- the ones that are unmarked
260 * should not be in the result.
262 for (i = 0; i < it->subtree_nr; i++)
263 it->down[i]->used = 0;
266 * Find the subtrees and update them.
268 i = 0;
269 while (i < entries) {
270 const struct cache_entry *ce = cache[i];
271 struct cache_tree_sub *sub;
272 const char *path, *slash;
273 int pathlen, sublen, subcnt, subskip;
275 path = ce->name;
276 pathlen = ce_namelen(ce);
277 if (pathlen <= baselen || memcmp(base, path, baselen))
278 break; /* at the end of this level */
280 slash = strchr(path + baselen, '/');
281 if (!slash) {
282 i++;
283 continue;
286 * a/bbb/c (base = a/, slash = /c)
287 * ==>
288 * path+baselen = bbb/c, sublen = 3
290 sublen = slash - (path + baselen);
291 sub = find_subtree(it, path + baselen, sublen, 1);
292 if (!sub->cache_tree)
293 sub->cache_tree = cache_tree();
294 subcnt = update_one(sub->cache_tree,
295 cache + i, entries - i,
296 path,
297 baselen + sublen + 1,
298 &subskip,
299 flags);
300 if (subcnt < 0)
301 return subcnt;
302 if (!subcnt)
303 die("index cache-tree records empty sub-tree");
304 i += subcnt;
305 sub->count = subcnt; /* to be used in the next loop */
306 *skip_count += subskip;
307 sub->used = 1;
310 discard_unused_subtrees(it);
313 * Then write out the tree object for this level.
315 strbuf_init(&buffer, 8192);
317 i = 0;
318 while (i < entries) {
319 const struct cache_entry *ce = cache[i];
320 struct cache_tree_sub *sub = NULL;
321 const char *path, *slash;
322 int pathlen, entlen;
323 const struct object_id *oid;
324 unsigned mode;
325 int expected_missing = 0;
326 int contains_ita = 0;
328 path = ce->name;
329 pathlen = ce_namelen(ce);
330 if (pathlen <= baselen || memcmp(base, path, baselen))
331 break; /* at the end of this level */
333 slash = strchr(path + baselen, '/');
334 if (slash) {
335 entlen = slash - (path + baselen);
336 sub = find_subtree(it, path + baselen, entlen, 0);
337 if (!sub)
338 die("cache-tree.c: '%.*s' in '%s' not found",
339 entlen, path + baselen, path);
340 i += sub->count;
341 oid = &sub->cache_tree->oid;
342 mode = S_IFDIR;
343 contains_ita = sub->cache_tree->entry_count < 0;
344 if (contains_ita) {
345 to_invalidate = 1;
346 expected_missing = 1;
349 else {
350 oid = &ce->oid;
351 mode = ce->ce_mode;
352 entlen = pathlen - baselen;
353 i++;
356 if (is_null_oid(oid) ||
357 (mode != S_IFGITLINK && !missing_ok && !has_object_file(oid))) {
358 strbuf_release(&buffer);
359 if (expected_missing)
360 return -1;
361 return error("invalid object %06o %s for '%.*s'",
362 mode, oid_to_hex(oid), entlen+baselen, path);
366 * CE_REMOVE entries are removed before the index is
367 * written to disk. Skip them to remain consistent
368 * with the future on-disk index.
370 if (ce->ce_flags & CE_REMOVE) {
371 *skip_count = *skip_count + 1;
372 continue;
376 * CE_INTENT_TO_ADD entries exist on on-disk index but
377 * they are not part of generated trees. Invalidate up
378 * to root to force cache-tree users to read elsewhere.
380 if (!sub && ce_intent_to_add(ce)) {
381 to_invalidate = 1;
382 continue;
386 * "sub" can be an empty tree if all subentries are i-t-a.
388 if (contains_ita && !oidcmp(oid, &empty_tree_oid))
389 continue;
391 strbuf_grow(&buffer, entlen + 100);
392 strbuf_addf(&buffer, "%o %.*s%c", mode, entlen, path + baselen, '\0');
393 strbuf_add(&buffer, oid->hash, the_hash_algo->rawsz);
395 #if DEBUG
396 fprintf(stderr, "cache-tree update-one %o %.*s\n",
397 mode, entlen, path + baselen);
398 #endif
401 if (repair) {
402 struct object_id oid;
403 hash_object_file(buffer.buf, buffer.len, tree_type, &oid);
404 if (has_object_file(&oid))
405 oidcpy(&it->oid, &oid);
406 else
407 to_invalidate = 1;
408 } else if (dryrun) {
409 hash_object_file(buffer.buf, buffer.len, tree_type, &it->oid);
410 } else if (write_object_file(buffer.buf, buffer.len, tree_type,
411 &it->oid)) {
412 strbuf_release(&buffer);
413 return -1;
416 strbuf_release(&buffer);
417 it->entry_count = to_invalidate ? -1 : i - *skip_count;
418 #if DEBUG
419 fprintf(stderr, "cache-tree update-one (%d ent, %d subtree) %s\n",
420 it->entry_count, it->subtree_nr,
421 oid_to_hex(&it->oid));
422 #endif
423 return i;
426 int cache_tree_update(struct index_state *istate, int flags)
428 struct cache_tree *it = istate->cache_tree;
429 struct cache_entry **cache = istate->cache;
430 int entries = istate->cache_nr;
431 int skip, i = verify_cache(cache, entries, flags);
433 if (i)
434 return i;
435 i = update_one(it, cache, entries, "", 0, &skip, flags);
436 if (i < 0)
437 return i;
438 istate->cache_changed |= CACHE_TREE_CHANGED;
439 return 0;
442 static void write_one(struct strbuf *buffer, struct cache_tree *it,
443 const char *path, int pathlen)
445 int i;
447 /* One "cache-tree" entry consists of the following:
448 * path (NUL terminated)
449 * entry_count, subtree_nr ("%d %d\n")
450 * tree-sha1 (missing if invalid)
451 * subtree_nr "cache-tree" entries for subtrees.
453 strbuf_grow(buffer, pathlen + 100);
454 strbuf_add(buffer, path, pathlen);
455 strbuf_addf(buffer, "%c%d %d\n", 0, it->entry_count, it->subtree_nr);
457 #if DEBUG
458 if (0 <= it->entry_count)
459 fprintf(stderr, "cache-tree <%.*s> (%d ent, %d subtree) %s\n",
460 pathlen, path, it->entry_count, it->subtree_nr,
461 oid_to_hex(&it->oid));
462 else
463 fprintf(stderr, "cache-tree <%.*s> (%d subtree) invalid\n",
464 pathlen, path, it->subtree_nr);
465 #endif
467 if (0 <= it->entry_count) {
468 strbuf_add(buffer, it->oid.hash, the_hash_algo->rawsz);
470 for (i = 0; i < it->subtree_nr; i++) {
471 struct cache_tree_sub *down = it->down[i];
472 if (i) {
473 struct cache_tree_sub *prev = it->down[i-1];
474 if (subtree_name_cmp(down->name, down->namelen,
475 prev->name, prev->namelen) <= 0)
476 die("fatal - unsorted cache subtree");
478 write_one(buffer, down->cache_tree, down->name, down->namelen);
482 void cache_tree_write(struct strbuf *sb, struct cache_tree *root)
484 write_one(sb, root, "", 0);
487 static struct cache_tree *read_one(const char **buffer, unsigned long *size_p)
489 const char *buf = *buffer;
490 unsigned long size = *size_p;
491 const char *cp;
492 char *ep;
493 struct cache_tree *it;
494 int i, subtree_nr;
495 const unsigned rawsz = the_hash_algo->rawsz;
497 it = NULL;
498 /* skip name, but make sure name exists */
499 while (size && *buf) {
500 size--;
501 buf++;
503 if (!size)
504 goto free_return;
505 buf++; size--;
506 it = cache_tree();
508 cp = buf;
509 it->entry_count = strtol(cp, &ep, 10);
510 if (cp == ep)
511 goto free_return;
512 cp = ep;
513 subtree_nr = strtol(cp, &ep, 10);
514 if (cp == ep)
515 goto free_return;
516 while (size && *buf && *buf != '\n') {
517 size--;
518 buf++;
520 if (!size)
521 goto free_return;
522 buf++; size--;
523 if (0 <= it->entry_count) {
524 if (size < rawsz)
525 goto free_return;
526 memcpy(it->oid.hash, (const unsigned char*)buf, rawsz);
527 buf += rawsz;
528 size -= rawsz;
531 #if DEBUG
532 if (0 <= it->entry_count)
533 fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n",
534 *buffer, it->entry_count, subtree_nr,
535 oid_to_hex(&it->oid));
536 else
537 fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n",
538 *buffer, subtree_nr);
539 #endif
542 * Just a heuristic -- we do not add directories that often but
543 * we do not want to have to extend it immediately when we do,
544 * hence +2.
546 it->subtree_alloc = subtree_nr + 2;
547 it->down = xcalloc(it->subtree_alloc, sizeof(struct cache_tree_sub *));
548 for (i = 0; i < subtree_nr; i++) {
549 /* read each subtree */
550 struct cache_tree *sub;
551 struct cache_tree_sub *subtree;
552 const char *name = buf;
554 sub = read_one(&buf, &size);
555 if (!sub)
556 goto free_return;
557 subtree = cache_tree_sub(it, name);
558 subtree->cache_tree = sub;
560 if (subtree_nr != it->subtree_nr)
561 die("cache-tree: internal error");
562 *buffer = buf;
563 *size_p = size;
564 return it;
566 free_return:
567 cache_tree_free(&it);
568 return NULL;
571 struct cache_tree *cache_tree_read(const char *buffer, unsigned long size)
573 if (buffer[0])
574 return NULL; /* not the whole tree */
575 return read_one(&buffer, &size);
578 static struct cache_tree *cache_tree_find(struct cache_tree *it, const char *path)
580 if (!it)
581 return NULL;
582 while (*path) {
583 const char *slash;
584 struct cache_tree_sub *sub;
586 slash = strchrnul(path, '/');
588 * Between path and slash is the name of the subtree
589 * to look for.
591 sub = find_subtree(it, path, slash - path, 0);
592 if (!sub)
593 return NULL;
594 it = sub->cache_tree;
596 path = slash;
597 while (*path == '/')
598 path++;
600 return it;
603 int write_index_as_tree(struct object_id *oid, struct index_state *index_state, const char *index_path, int flags, const char *prefix)
605 int entries, was_valid;
606 struct lock_file lock_file = LOCK_INIT;
607 int ret = 0;
609 hold_lock_file_for_update(&lock_file, index_path, LOCK_DIE_ON_ERROR);
611 entries = read_index_from(index_state, index_path, get_git_dir());
612 if (entries < 0) {
613 ret = WRITE_TREE_UNREADABLE_INDEX;
614 goto out;
616 if (flags & WRITE_TREE_IGNORE_CACHE_TREE)
617 cache_tree_free(&index_state->cache_tree);
619 if (!index_state->cache_tree)
620 index_state->cache_tree = cache_tree();
622 was_valid = cache_tree_fully_valid(index_state->cache_tree);
623 if (!was_valid) {
624 if (cache_tree_update(index_state, flags) < 0) {
625 ret = WRITE_TREE_UNMERGED_INDEX;
626 goto out;
628 write_locked_index(index_state, &lock_file, COMMIT_LOCK);
629 /* Not being able to write is fine -- we are only interested
630 * in updating the cache-tree part, and if the next caller
631 * ends up using the old index with unupdated cache-tree part
632 * it misses the work we did here, but that is just a
633 * performance penalty and not a big deal.
637 if (prefix) {
638 struct cache_tree *subtree;
639 subtree = cache_tree_find(index_state->cache_tree, prefix);
640 if (!subtree) {
641 ret = WRITE_TREE_PREFIX_ERROR;
642 goto out;
644 oidcpy(oid, &subtree->oid);
646 else
647 oidcpy(oid, &index_state->cache_tree->oid);
649 out:
650 rollback_lock_file(&lock_file);
651 return ret;
654 int write_cache_as_tree(struct object_id *oid, int flags, const char *prefix)
656 return write_index_as_tree(oid, &the_index, get_index_file(), flags, prefix);
659 static void prime_cache_tree_rec(struct cache_tree *it, struct tree *tree)
661 struct tree_desc desc;
662 struct name_entry entry;
663 int cnt;
665 oidcpy(&it->oid, &tree->object.oid);
666 init_tree_desc(&desc, tree->buffer, tree->size);
667 cnt = 0;
668 while (tree_entry(&desc, &entry)) {
669 if (!S_ISDIR(entry.mode))
670 cnt++;
671 else {
672 struct cache_tree_sub *sub;
673 struct tree *subtree = lookup_tree(entry.oid);
674 if (!subtree->object.parsed)
675 parse_tree(subtree);
676 sub = cache_tree_sub(it, entry.path);
677 sub->cache_tree = cache_tree();
678 prime_cache_tree_rec(sub->cache_tree, subtree);
679 cnt += sub->cache_tree->entry_count;
682 it->entry_count = cnt;
685 void prime_cache_tree(struct index_state *istate, struct tree *tree)
687 cache_tree_free(&istate->cache_tree);
688 istate->cache_tree = cache_tree();
689 prime_cache_tree_rec(istate->cache_tree, tree);
690 istate->cache_changed |= CACHE_TREE_CHANGED;
694 * find the cache_tree that corresponds to the current level without
695 * exploding the full path into textual form. The root of the
696 * cache tree is given as "root", and our current level is "info".
697 * (1) When at root level, info->prev is NULL, so it is "root" itself.
698 * (2) Otherwise, find the cache_tree that corresponds to one level
699 * above us, and find ourselves in there.
701 static struct cache_tree *find_cache_tree_from_traversal(struct cache_tree *root,
702 struct traverse_info *info)
704 struct cache_tree *our_parent;
706 if (!info->prev)
707 return root;
708 our_parent = find_cache_tree_from_traversal(root, info->prev);
709 return cache_tree_find(our_parent, info->name.path);
712 int cache_tree_matches_traversal(struct cache_tree *root,
713 struct name_entry *ent,
714 struct traverse_info *info)
716 struct cache_tree *it;
718 it = find_cache_tree_from_traversal(root, info);
719 it = cache_tree_find(it, ent->path);
720 if (it && it->entry_count > 0 && !oidcmp(ent->oid, &it->oid))
721 return it->entry_count;
722 return 0;
725 int update_main_cache_tree(int flags)
727 if (!the_index.cache_tree)
728 the_index.cache_tree = cache_tree();
729 return cache_tree_update(&the_index, flags);