5 #include "cache-tree.h"
6 #include "object-store.h"
7 #include "replace-object.h"
13 struct cache_tree *cache_tree(void)
15 struct cache_tree *it = xcalloc(1, sizeof(struct cache_tree));
20 void cache_tree_free(struct cache_tree **it_p)
23 struct cache_tree *it = *it_p;
27 for (i = 0; i < it->subtree_nr; i++)
29 cache_tree_free(&it->down[i]->cache_tree);
37 static int subtree_name_cmp(const char *one, int onelen,
38 const char *two, int twolen)
44 return memcmp(one, two, onelen);
47 static int subtree_pos(struct cache_tree *it, const char *path, int pathlen)
49 struct cache_tree_sub **down = it->down;
54 int mi = lo + (hi - lo) / 2;
55 struct cache_tree_sub *mdl = down[mi];
56 int cmp = subtree_name_cmp(path, pathlen,
57 mdl->name, mdl->namelen);
68 static struct cache_tree_sub *find_subtree(struct cache_tree *it,
73 struct cache_tree_sub *down;
74 int pos = subtree_pos(it, path, pathlen);
81 ALLOC_GROW(it->down, it->subtree_nr + 1, it->subtree_alloc);
84 FLEX_ALLOC_MEM(down, name, path, pathlen);
85 down->cache_tree = NULL;
86 down->namelen = pathlen;
88 if (pos < it->subtree_nr)
89 MOVE_ARRAY(it->down + pos + 1, it->down + pos,
90 it->subtree_nr - pos - 1);
95 struct cache_tree_sub *cache_tree_sub(struct cache_tree *it, const char *path)
97 int pathlen = strlen(path);
98 return find_subtree(it, path, pathlen, 1);
101 static int do_invalidate_path(struct cache_tree *it, const char *path)
104 * ==> invalidate self
105 * ==> find "a", have it invalidate "b/c"
107 * ==> invalidate self
108 * ==> if "a" exists as a subtree, remove it.
112 struct cache_tree_sub *down;
115 fprintf(stderr, "cache-tree invalidate <%s>\n", path);
120 slash = strchrnul(path, '/');
121 namelen = slash - path;
122 it->entry_count = -1;
125 pos = subtree_pos(it, path, namelen);
127 cache_tree_free(&it->down[pos]->cache_tree);
132 * move 4 and 5 up one place (2 entries)
133 * 2 = 6 - 3 - 1 = subtree_nr - pos - 1
135 MOVE_ARRAY(it->down + pos, it->down + pos + 1,
136 it->subtree_nr - pos - 1);
141 down = find_subtree(it, path, namelen, 0);
143 do_invalidate_path(down->cache_tree, slash + 1);
147 void cache_tree_invalidate_path(struct index_state *istate, const char *path)
149 if (do_invalidate_path(istate->cache_tree, path))
150 istate->cache_changed |= CACHE_TREE_CHANGED;
153 static int verify_cache(struct cache_entry **cache,
154 int entries, int flags)
157 int silent = flags & WRITE_TREE_SILENT;
159 /* Verify that the tree is merged */
161 for (i = 0; i < entries; i++) {
162 const struct cache_entry *ce = cache[i];
167 fprintf(stderr, "...\n");
170 fprintf(stderr, "%s: unmerged (%s)\n",
171 ce->name, oid_to_hex(&ce->oid));
177 /* Also verify that the cache does not have path and path/file
178 * at the same time. At this point we know the cache has only
182 for (i = 0; i < entries - 1; i++) {
183 /* path/file always comes after path because of the way
184 * the cache is sorted. Also path can appear only once,
185 * which means conflicting one would immediately follow.
187 const char *this_name = cache[i]->name;
188 const char *next_name = cache[i+1]->name;
189 int this_len = strlen(this_name);
190 if (this_len < strlen(next_name) &&
191 strncmp(this_name, next_name, this_len) == 0 &&
192 next_name[this_len] == '/') {
194 fprintf(stderr, "...\n");
197 fprintf(stderr, "You have both %s and %s\n",
198 this_name, next_name);
206 static void discard_unused_subtrees(struct cache_tree *it)
208 struct cache_tree_sub **down = it->down;
209 int nr = it->subtree_nr;
211 for (dst = src = 0; src < nr; src++) {
212 struct cache_tree_sub *s = down[src];
216 cache_tree_free(&s->cache_tree);
223 int cache_tree_fully_valid(struct cache_tree *it)
228 if (it->entry_count < 0 || !has_sha1_file(it->oid.hash))
230 for (i = 0; i < it->subtree_nr; i++) {
231 if (!cache_tree_fully_valid(it->down[i]->cache_tree))
237 static int update_one(struct cache_tree *it,
238 struct cache_entry **cache,
245 struct strbuf buffer;
246 int missing_ok = flags & WRITE_TREE_MISSING_OK;
247 int dryrun = flags & WRITE_TREE_DRY_RUN;
248 int repair = flags & WRITE_TREE_REPAIR;
249 int to_invalidate = 0;
252 assert(!(dryrun && repair));
256 if (0 <= it->entry_count && has_sha1_file(it->oid.hash))
257 return it->entry_count;
260 * We first scan for subtrees and update them; we start by
261 * marking existing subtrees -- the ones that are unmarked
262 * should not be in the result.
264 for (i = 0; i < it->subtree_nr; i++)
265 it->down[i]->used = 0;
268 * Find the subtrees and update them.
271 while (i < entries) {
272 const struct cache_entry *ce = cache[i];
273 struct cache_tree_sub *sub;
274 const char *path, *slash;
275 int pathlen, sublen, subcnt, subskip;
278 pathlen = ce_namelen(ce);
279 if (pathlen <= baselen || memcmp(base, path, baselen))
280 break; /* at the end of this level */
282 slash = strchr(path + baselen, '/');
288 * a/bbb/c (base = a/, slash = /c)
290 * path+baselen = bbb/c, sublen = 3
292 sublen = slash - (path + baselen);
293 sub = find_subtree(it, path + baselen, sublen, 1);
294 if (!sub->cache_tree)
295 sub->cache_tree = cache_tree();
296 subcnt = update_one(sub->cache_tree,
297 cache + i, entries - i,
299 baselen + sublen + 1,
305 die("index cache-tree records empty sub-tree");
307 sub->count = subcnt; /* to be used in the next loop */
308 *skip_count += subskip;
312 discard_unused_subtrees(it);
315 * Then write out the tree object for this level.
317 strbuf_init(&buffer, 8192);
320 while (i < entries) {
321 const struct cache_entry *ce = cache[i];
322 struct cache_tree_sub *sub = NULL;
323 const char *path, *slash;
325 const struct object_id *oid;
327 int expected_missing = 0;
328 int contains_ita = 0;
331 pathlen = ce_namelen(ce);
332 if (pathlen <= baselen || memcmp(base, path, baselen))
333 break; /* at the end of this level */
335 slash = strchr(path + baselen, '/');
337 entlen = slash - (path + baselen);
338 sub = find_subtree(it, path + baselen, entlen, 0);
340 die("cache-tree.c: '%.*s' in '%s' not found",
341 entlen, path + baselen, path);
343 oid = &sub->cache_tree->oid;
345 contains_ita = sub->cache_tree->entry_count < 0;
348 expected_missing = 1;
354 entlen = pathlen - baselen;
358 if (is_null_oid(oid) ||
359 (mode != S_IFGITLINK && !missing_ok && !has_object_file(oid))) {
360 strbuf_release(&buffer);
361 if (expected_missing)
363 return error("invalid object %06o %s for '%.*s'",
364 mode, oid_to_hex(oid), entlen+baselen, path);
368 * CE_REMOVE entries are removed before the index is
369 * written to disk. Skip them to remain consistent
370 * with the future on-disk index.
372 if (ce->ce_flags & CE_REMOVE) {
373 *skip_count = *skip_count + 1;
378 * CE_INTENT_TO_ADD entries exist on on-disk index but
379 * they are not part of generated trees. Invalidate up
380 * to root to force cache-tree users to read elsewhere.
382 if (!sub && ce_intent_to_add(ce)) {
388 * "sub" can be an empty tree if all subentries are i-t-a.
390 if (contains_ita && is_empty_tree_oid(oid))
393 strbuf_grow(&buffer, entlen + 100);
394 strbuf_addf(&buffer, "%o %.*s%c", mode, entlen, path + baselen, '\0');
395 strbuf_add(&buffer, oid->hash, the_hash_algo->rawsz);
398 fprintf(stderr, "cache-tree update-one %o %.*s\n",
399 mode, entlen, path + baselen);
404 struct object_id oid;
405 hash_object_file(buffer.buf, buffer.len, tree_type, &oid);
406 if (has_object_file(&oid))
407 oidcpy(&it->oid, &oid);
411 hash_object_file(buffer.buf, buffer.len, tree_type, &it->oid);
412 } else if (write_object_file(buffer.buf, buffer.len, tree_type,
414 strbuf_release(&buffer);
418 strbuf_release(&buffer);
419 it->entry_count = to_invalidate ? -1 : i - *skip_count;
421 fprintf(stderr, "cache-tree update-one (%d ent, %d subtree) %s\n",
422 it->entry_count, it->subtree_nr,
423 oid_to_hex(&it->oid));
428 int cache_tree_update(struct index_state *istate, int flags)
430 struct cache_tree *it = istate->cache_tree;
431 struct cache_entry **cache = istate->cache;
432 int entries = istate->cache_nr;
433 int skip, i = verify_cache(cache, entries, flags);
437 trace_performance_enter();
438 i = update_one(it, cache, entries, "", 0, &skip, flags);
439 trace_performance_leave("cache_tree_update");
442 istate->cache_changed |= CACHE_TREE_CHANGED;
446 static void write_one(struct strbuf *buffer, struct cache_tree *it,
447 const char *path, int pathlen)
451 /* One "cache-tree" entry consists of the following:
452 * path (NUL terminated)
453 * entry_count, subtree_nr ("%d %d\n")
454 * tree-sha1 (missing if invalid)
455 * subtree_nr "cache-tree" entries for subtrees.
457 strbuf_grow(buffer, pathlen + 100);
458 strbuf_add(buffer, path, pathlen);
459 strbuf_addf(buffer, "%c%d %d\n", 0, it->entry_count, it->subtree_nr);
462 if (0 <= it->entry_count)
463 fprintf(stderr, "cache-tree <%.*s> (%d ent, %d subtree) %s\n",
464 pathlen, path, it->entry_count, it->subtree_nr,
465 oid_to_hex(&it->oid));
467 fprintf(stderr, "cache-tree <%.*s> (%d subtree) invalid\n",
468 pathlen, path, it->subtree_nr);
471 if (0 <= it->entry_count) {
472 strbuf_add(buffer, it->oid.hash, the_hash_algo->rawsz);
474 for (i = 0; i < it->subtree_nr; i++) {
475 struct cache_tree_sub *down = it->down[i];
477 struct cache_tree_sub *prev = it->down[i-1];
478 if (subtree_name_cmp(down->name, down->namelen,
479 prev->name, prev->namelen) <= 0)
480 die("fatal - unsorted cache subtree");
482 write_one(buffer, down->cache_tree, down->name, down->namelen);
486 void cache_tree_write(struct strbuf *sb, struct cache_tree *root)
488 write_one(sb, root, "", 0);
491 static struct cache_tree *read_one(const char **buffer, unsigned long *size_p)
493 const char *buf = *buffer;
494 unsigned long size = *size_p;
497 struct cache_tree *it;
499 const unsigned rawsz = the_hash_algo->rawsz;
502 /* skip name, but make sure name exists */
503 while (size && *buf) {
513 it->entry_count = strtol(cp, &ep, 10);
517 subtree_nr = strtol(cp, &ep, 10);
520 while (size && *buf && *buf != '\n') {
527 if (0 <= it->entry_count) {
530 oidread(&it->oid, (const unsigned char *)buf);
536 if (0 <= it->entry_count)
537 fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n",
538 *buffer, it->entry_count, subtree_nr,
539 oid_to_hex(&it->oid));
541 fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n",
542 *buffer, subtree_nr);
546 * Just a heuristic -- we do not add directories that often but
547 * we do not want to have to extend it immediately when we do,
550 it->subtree_alloc = subtree_nr + 2;
551 it->down = xcalloc(it->subtree_alloc, sizeof(struct cache_tree_sub *));
552 for (i = 0; i < subtree_nr; i++) {
553 /* read each subtree */
554 struct cache_tree *sub;
555 struct cache_tree_sub *subtree;
556 const char *name = buf;
558 sub = read_one(&buf, &size);
561 subtree = cache_tree_sub(it, name);
562 subtree->cache_tree = sub;
564 if (subtree_nr != it->subtree_nr)
565 die("cache-tree: internal error");
571 cache_tree_free(&it);
575 struct cache_tree *cache_tree_read(const char *buffer, unsigned long size)
578 return NULL; /* not the whole tree */
579 return read_one(&buffer, &size);
582 static struct cache_tree *cache_tree_find(struct cache_tree *it, const char *path)
588 struct cache_tree_sub *sub;
590 slash = strchrnul(path, '/');
592 * Between path and slash is the name of the subtree
595 sub = find_subtree(it, path, slash - path, 0);
598 it = sub->cache_tree;
607 int write_index_as_tree(struct object_id *oid, struct index_state *index_state, const char *index_path, int flags, const char *prefix)
609 int entries, was_valid;
610 struct lock_file lock_file = LOCK_INIT;
613 hold_lock_file_for_update(&lock_file, index_path, LOCK_DIE_ON_ERROR);
615 entries = read_index_from(index_state, index_path, get_git_dir());
617 ret = WRITE_TREE_UNREADABLE_INDEX;
620 if (flags & WRITE_TREE_IGNORE_CACHE_TREE)
621 cache_tree_free(&index_state->cache_tree);
623 if (!index_state->cache_tree)
624 index_state->cache_tree = cache_tree();
626 was_valid = cache_tree_fully_valid(index_state->cache_tree);
628 if (cache_tree_update(index_state, flags) < 0) {
629 ret = WRITE_TREE_UNMERGED_INDEX;
632 write_locked_index(index_state, &lock_file, COMMIT_LOCK);
633 /* Not being able to write is fine -- we are only interested
634 * in updating the cache-tree part, and if the next caller
635 * ends up using the old index with unupdated cache-tree part
636 * it misses the work we did here, but that is just a
637 * performance penalty and not a big deal.
642 struct cache_tree *subtree;
643 subtree = cache_tree_find(index_state->cache_tree, prefix);
645 ret = WRITE_TREE_PREFIX_ERROR;
648 oidcpy(oid, &subtree->oid);
651 oidcpy(oid, &index_state->cache_tree->oid);
654 rollback_lock_file(&lock_file);
658 static void prime_cache_tree_rec(struct cache_tree *it, struct tree *tree)
660 struct tree_desc desc;
661 struct name_entry entry;
664 oidcpy(&it->oid, &tree->object.oid);
665 init_tree_desc(&desc, tree->buffer, tree->size);
667 while (tree_entry(&desc, &entry)) {
668 if (!S_ISDIR(entry.mode))
671 struct cache_tree_sub *sub;
672 struct tree *subtree = lookup_tree(the_repository,
674 if (!subtree->object.parsed)
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;
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 && oideq(ent->oid, &it->oid))
721 return it->entry_count;
725 static void verify_one(struct index_state *istate,
726 struct cache_tree *it,
729 int i, pos, len = path->len;
730 struct strbuf tree_buf = STRBUF_INIT;
731 struct object_id new_oid;
733 for (i = 0; i < it->subtree_nr; i++) {
734 strbuf_addf(path, "%s/", it->down[i]->name);
735 verify_one(istate, it->down[i]->cache_tree, path);
736 strbuf_setlen(path, len);
739 if (it->entry_count < 0 ||
740 /* no verification on tests (t7003) that replace trees */
741 lookup_replace_object(the_repository, &it->oid) != &it->oid)
745 pos = index_name_pos(istate, path->buf, path->len);
752 while (i < it->entry_count) {
753 struct cache_entry *ce = istate->cache[pos + i];
755 struct cache_tree_sub *sub = NULL;
756 const struct object_id *oid;
761 if (ce->ce_flags & (CE_STAGEMASK | CE_INTENT_TO_ADD | CE_REMOVE))
762 BUG("%s with flags 0x%x should not be in cache-tree",
763 ce->name, ce->ce_flags);
764 name = ce->name + path->len;
765 slash = strchr(name, '/');
767 entlen = slash - name;
768 sub = find_subtree(it, ce->name + path->len, entlen, 0);
769 if (!sub || sub->cache_tree->entry_count < 0)
770 BUG("bad subtree '%.*s'", entlen, name);
771 oid = &sub->cache_tree->oid;
773 i += sub->cache_tree->entry_count;
777 entlen = ce_namelen(ce) - path->len;
780 strbuf_addf(&tree_buf, "%o %.*s%c", mode, entlen, name, '\0');
781 strbuf_add(&tree_buf, oid->hash, the_hash_algo->rawsz);
783 hash_object_file(tree_buf.buf, tree_buf.len, tree_type, &new_oid);
784 if (!oideq(&new_oid, &it->oid))
785 BUG("cache-tree for path %.*s does not match. "
786 "Expected %s got %s", len, path->buf,
787 oid_to_hex(&new_oid), oid_to_hex(&it->oid));
788 strbuf_setlen(path, len);
789 strbuf_release(&tree_buf);
792 void cache_tree_verify(struct index_state *istate)
794 struct strbuf path = STRBUF_INIT;
796 if (!istate->cache_tree)
798 verify_one(istate, istate->cache_tree, &path);
799 strbuf_release(&path);