5 #include "cache-tree.h"
6 #include "object-store.h"
7 #include "replace-object.h"
8 #include "promisor-remote.h"
10 #ifndef DEBUG_CACHE_TREE
11 #define DEBUG_CACHE_TREE 0
14 struct cache_tree *cache_tree(void)
16 struct cache_tree *it = xcalloc(1, sizeof(struct cache_tree));
21 void cache_tree_free(struct cache_tree **it_p)
24 struct cache_tree *it = *it_p;
28 for (i = 0; i < it->subtree_nr; i++)
30 cache_tree_free(&it->down[i]->cache_tree);
38 static int subtree_name_cmp(const char *one, int onelen,
39 const char *two, int twolen)
45 return memcmp(one, two, onelen);
48 static int subtree_pos(struct cache_tree *it, const char *path, int pathlen)
50 struct cache_tree_sub **down = it->down;
55 int mi = lo + (hi - lo) / 2;
56 struct cache_tree_sub *mdl = down[mi];
57 int cmp = subtree_name_cmp(path, pathlen,
58 mdl->name, mdl->namelen);
69 static struct cache_tree_sub *find_subtree(struct cache_tree *it,
74 struct cache_tree_sub *down;
75 int pos = subtree_pos(it, path, pathlen);
82 ALLOC_GROW(it->down, it->subtree_nr + 1, it->subtree_alloc);
85 FLEX_ALLOC_MEM(down, name, path, pathlen);
86 down->cache_tree = NULL;
87 down->namelen = pathlen;
89 if (pos < it->subtree_nr)
90 MOVE_ARRAY(it->down + pos + 1, it->down + pos,
91 it->subtree_nr - pos - 1);
96 struct cache_tree_sub *cache_tree_sub(struct cache_tree *it, const char *path)
98 int pathlen = strlen(path);
99 return find_subtree(it, path, pathlen, 1);
102 static int do_invalidate_path(struct cache_tree *it, const char *path)
105 * ==> invalidate self
106 * ==> find "a", have it invalidate "b/c"
108 * ==> invalidate self
109 * ==> if "a" exists as a subtree, remove it.
113 struct cache_tree_sub *down;
116 fprintf(stderr, "cache-tree invalidate <%s>\n", path);
121 slash = strchrnul(path, '/');
122 namelen = slash - path;
123 it->entry_count = -1;
126 pos = subtree_pos(it, path, namelen);
128 cache_tree_free(&it->down[pos]->cache_tree);
133 * move 4 and 5 up one place (2 entries)
134 * 2 = 6 - 3 - 1 = subtree_nr - pos - 1
136 MOVE_ARRAY(it->down + pos, it->down + pos + 1,
137 it->subtree_nr - pos - 1);
142 down = find_subtree(it, path, namelen, 0);
144 do_invalidate_path(down->cache_tree, slash + 1);
148 void cache_tree_invalidate_path(struct index_state *istate, const char *path)
150 if (do_invalidate_path(istate->cache_tree, path))
151 istate->cache_changed |= CACHE_TREE_CHANGED;
154 static int verify_cache(struct cache_entry **cache,
155 int entries, int flags)
158 int silent = flags & WRITE_TREE_SILENT;
160 /* Verify that the tree is merged */
162 for (i = 0; i < entries; i++) {
163 const struct cache_entry *ce = cache[i];
168 fprintf(stderr, "...\n");
171 fprintf(stderr, "%s: unmerged (%s)\n",
172 ce->name, oid_to_hex(&ce->oid));
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
183 for (i = 0; i < entries - 1; 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.
188 const char *this_name = cache[i]->name;
189 const char *next_name = cache[i+1]->name;
190 int this_len = strlen(this_name);
191 if (this_len < strlen(next_name) &&
192 strncmp(this_name, next_name, this_len) == 0 &&
193 next_name[this_len] == '/') {
195 fprintf(stderr, "...\n");
198 fprintf(stderr, "You have both %s and %s\n",
199 this_name, next_name);
207 static void discard_unused_subtrees(struct cache_tree *it)
209 struct cache_tree_sub **down = it->down;
210 int nr = it->subtree_nr;
212 for (dst = src = 0; src < nr; src++) {
213 struct cache_tree_sub *s = down[src];
217 cache_tree_free(&s->cache_tree);
224 int cache_tree_fully_valid(struct cache_tree *it)
229 if (it->entry_count < 0 || !has_object_file(&it->oid))
231 for (i = 0; i < it->subtree_nr; i++) {
232 if (!cache_tree_fully_valid(it->down[i]->cache_tree))
238 static int update_one(struct cache_tree *it,
239 struct cache_entry **cache,
246 struct strbuf buffer;
247 int missing_ok = flags & WRITE_TREE_MISSING_OK;
248 int dryrun = flags & WRITE_TREE_DRY_RUN;
249 int repair = flags & WRITE_TREE_REPAIR;
250 int to_invalidate = 0;
253 assert(!(dryrun && repair));
257 if (0 <= it->entry_count && has_object_file(&it->oid))
258 return it->entry_count;
261 * We first scan for subtrees and update them; we start by
262 * marking existing subtrees -- the ones that are unmarked
263 * should not be in the result.
265 for (i = 0; i < it->subtree_nr; i++)
266 it->down[i]->used = 0;
269 * Find the subtrees and update them.
272 while (i < entries) {
273 const struct cache_entry *ce = cache[i];
274 struct cache_tree_sub *sub;
275 const char *path, *slash;
276 int pathlen, sublen, subcnt, subskip;
279 pathlen = ce_namelen(ce);
280 if (pathlen <= baselen || memcmp(base, path, baselen))
281 break; /* at the end of this level */
283 slash = strchr(path + baselen, '/');
289 * a/bbb/c (base = a/, slash = /c)
291 * path+baselen = bbb/c, sublen = 3
293 sublen = slash - (path + baselen);
294 sub = find_subtree(it, path + baselen, sublen, 1);
295 if (!sub->cache_tree)
296 sub->cache_tree = cache_tree();
297 subcnt = update_one(sub->cache_tree,
298 cache + i, entries - i,
300 baselen + sublen + 1,
306 die("index cache-tree records empty sub-tree");
308 sub->count = subcnt; /* to be used in the next loop */
309 *skip_count += subskip;
313 discard_unused_subtrees(it);
316 * Then write out the tree object for this level.
318 strbuf_init(&buffer, 8192);
321 while (i < entries) {
322 const struct cache_entry *ce = cache[i];
323 struct cache_tree_sub *sub = NULL;
324 const char *path, *slash;
326 const struct object_id *oid;
328 int expected_missing = 0;
329 int contains_ita = 0;
333 pathlen = ce_namelen(ce);
334 if (pathlen <= baselen || memcmp(base, path, baselen))
335 break; /* at the end of this level */
337 slash = strchr(path + baselen, '/');
339 entlen = slash - (path + baselen);
340 sub = find_subtree(it, path + baselen, entlen, 0);
342 die("cache-tree.c: '%.*s' in '%s' not found",
343 entlen, path + baselen, path);
345 oid = &sub->cache_tree->oid;
347 contains_ita = sub->cache_tree->entry_count < 0;
350 expected_missing = 1;
356 entlen = pathlen - baselen;
360 ce_missing_ok = mode == S_IFGITLINK || missing_ok ||
361 (has_promisor_remote() &&
362 ce_skip_worktree(ce));
363 if (is_null_oid(oid) ||
364 (!ce_missing_ok && !has_object_file(oid))) {
365 strbuf_release(&buffer);
366 if (expected_missing)
368 return error("invalid object %06o %s for '%.*s'",
369 mode, oid_to_hex(oid), entlen+baselen, path);
373 * CE_REMOVE entries are removed before the index is
374 * written to disk. Skip them to remain consistent
375 * with the future on-disk index.
377 if (ce->ce_flags & CE_REMOVE) {
378 *skip_count = *skip_count + 1;
383 * CE_INTENT_TO_ADD entries exist on on-disk index but
384 * they are not part of generated trees. Invalidate up
385 * to root to force cache-tree users to read elsewhere.
387 if (!sub && ce_intent_to_add(ce)) {
393 * "sub" can be an empty tree if all subentries are i-t-a.
395 if (contains_ita && is_empty_tree_oid(oid))
398 strbuf_grow(&buffer, entlen + 100);
399 strbuf_addf(&buffer, "%o %.*s%c", mode, entlen, path + baselen, '\0');
400 strbuf_add(&buffer, oid->hash, the_hash_algo->rawsz);
403 fprintf(stderr, "cache-tree update-one %o %.*s\n",
404 mode, entlen, path + baselen);
409 struct object_id oid;
410 hash_object_file(the_hash_algo, buffer.buf, buffer.len,
412 if (has_object_file_with_flags(&oid, OBJECT_INFO_SKIP_FETCH_OBJECT))
413 oidcpy(&it->oid, &oid);
417 hash_object_file(the_hash_algo, buffer.buf, buffer.len,
418 tree_type, &it->oid);
419 } else if (write_object_file(buffer.buf, buffer.len, tree_type,
421 strbuf_release(&buffer);
425 strbuf_release(&buffer);
426 it->entry_count = to_invalidate ? -1 : i - *skip_count;
428 fprintf(stderr, "cache-tree update-one (%d ent, %d subtree) %s\n",
429 it->entry_count, it->subtree_nr,
430 oid_to_hex(&it->oid));
435 int cache_tree_update(struct index_state *istate, int flags)
437 struct cache_tree *it = istate->cache_tree;
438 struct cache_entry **cache = istate->cache;
439 int entries = istate->cache_nr;
440 int skip, i = verify_cache(cache, entries, flags);
444 trace_performance_enter();
445 i = update_one(it, cache, entries, "", 0, &skip, flags);
446 trace_performance_leave("cache_tree_update");
449 istate->cache_changed |= CACHE_TREE_CHANGED;
453 static void write_one(struct strbuf *buffer, struct cache_tree *it,
454 const char *path, int pathlen)
458 /* One "cache-tree" entry consists of the following:
459 * path (NUL terminated)
460 * entry_count, subtree_nr ("%d %d\n")
461 * tree-sha1 (missing if invalid)
462 * subtree_nr "cache-tree" entries for subtrees.
464 strbuf_grow(buffer, pathlen + 100);
465 strbuf_add(buffer, path, pathlen);
466 strbuf_addf(buffer, "%c%d %d\n", 0, it->entry_count, it->subtree_nr);
469 if (0 <= it->entry_count)
470 fprintf(stderr, "cache-tree <%.*s> (%d ent, %d subtree) %s\n",
471 pathlen, path, it->entry_count, it->subtree_nr,
472 oid_to_hex(&it->oid));
474 fprintf(stderr, "cache-tree <%.*s> (%d subtree) invalid\n",
475 pathlen, path, it->subtree_nr);
478 if (0 <= it->entry_count) {
479 strbuf_add(buffer, it->oid.hash, the_hash_algo->rawsz);
481 for (i = 0; i < it->subtree_nr; i++) {
482 struct cache_tree_sub *down = it->down[i];
484 struct cache_tree_sub *prev = it->down[i-1];
485 if (subtree_name_cmp(down->name, down->namelen,
486 prev->name, prev->namelen) <= 0)
487 die("fatal - unsorted cache subtree");
489 write_one(buffer, down->cache_tree, down->name, down->namelen);
493 void cache_tree_write(struct strbuf *sb, struct cache_tree *root)
495 write_one(sb, root, "", 0);
498 static struct cache_tree *read_one(const char **buffer, unsigned long *size_p)
500 const char *buf = *buffer;
501 unsigned long size = *size_p;
504 struct cache_tree *it;
506 const unsigned rawsz = the_hash_algo->rawsz;
509 /* skip name, but make sure name exists */
510 while (size && *buf) {
520 it->entry_count = strtol(cp, &ep, 10);
524 subtree_nr = strtol(cp, &ep, 10);
527 while (size && *buf && *buf != '\n') {
534 if (0 <= it->entry_count) {
537 oidread(&it->oid, (const unsigned char *)buf);
543 if (0 <= it->entry_count)
544 fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n",
545 *buffer, it->entry_count, subtree_nr,
546 oid_to_hex(&it->oid));
548 fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n",
549 *buffer, subtree_nr);
553 * Just a heuristic -- we do not add directories that often but
554 * we do not want to have to extend it immediately when we do,
557 it->subtree_alloc = subtree_nr + 2;
558 it->down = xcalloc(it->subtree_alloc, sizeof(struct cache_tree_sub *));
559 for (i = 0; i < subtree_nr; i++) {
560 /* read each subtree */
561 struct cache_tree *sub;
562 struct cache_tree_sub *subtree;
563 const char *name = buf;
565 sub = read_one(&buf, &size);
568 subtree = cache_tree_sub(it, name);
569 subtree->cache_tree = sub;
571 if (subtree_nr != it->subtree_nr)
572 die("cache-tree: internal error");
578 cache_tree_free(&it);
582 struct cache_tree *cache_tree_read(const char *buffer, unsigned long size)
585 return NULL; /* not the whole tree */
586 return read_one(&buffer, &size);
589 static struct cache_tree *cache_tree_find(struct cache_tree *it, const char *path)
595 struct cache_tree_sub *sub;
597 slash = strchrnul(path, '/');
599 * Between path and slash is the name of the subtree
602 sub = find_subtree(it, path, slash - path, 0);
605 it = sub->cache_tree;
614 static int write_index_as_tree_internal(struct object_id *oid,
615 struct index_state *index_state,
616 int cache_tree_valid,
620 if (flags & WRITE_TREE_IGNORE_CACHE_TREE) {
621 cache_tree_free(&index_state->cache_tree);
622 cache_tree_valid = 0;
625 if (!index_state->cache_tree)
626 index_state->cache_tree = cache_tree();
628 if (!cache_tree_valid && cache_tree_update(index_state, flags) < 0)
629 return WRITE_TREE_UNMERGED_INDEX;
632 struct cache_tree *subtree;
633 subtree = cache_tree_find(index_state->cache_tree, prefix);
635 return WRITE_TREE_PREFIX_ERROR;
636 oidcpy(oid, &subtree->oid);
639 oidcpy(oid, &index_state->cache_tree->oid);
644 struct tree* write_in_core_index_as_tree(struct repository *repo) {
648 struct index_state *index_state = repo->index;
649 was_valid = index_state->cache_tree &&
650 cache_tree_fully_valid(index_state->cache_tree);
652 ret = write_index_as_tree_internal(&o, index_state, was_valid, 0, NULL);
653 if (ret == WRITE_TREE_UNMERGED_INDEX) {
655 fprintf(stderr, "BUG: There are unmerged index entries:\n");
656 for (i = 0; i < index_state->cache_nr; i++) {
657 const struct cache_entry *ce = index_state->cache[i];
659 fprintf(stderr, "BUG: %d %.*s\n", ce_stage(ce),
660 (int)ce_namelen(ce), ce->name);
662 BUG("unmerged index entries when writing inmemory index");
665 return lookup_tree(repo, &index_state->cache_tree->oid);
669 int write_index_as_tree(struct object_id *oid, struct index_state *index_state, const char *index_path, int flags, const char *prefix)
671 int entries, was_valid;
672 struct lock_file lock_file = LOCK_INIT;
675 hold_lock_file_for_update(&lock_file, index_path, LOCK_DIE_ON_ERROR);
677 entries = read_index_from(index_state, index_path, get_git_dir());
679 ret = WRITE_TREE_UNREADABLE_INDEX;
683 was_valid = !(flags & WRITE_TREE_IGNORE_CACHE_TREE) &&
684 index_state->cache_tree &&
685 cache_tree_fully_valid(index_state->cache_tree);
687 ret = write_index_as_tree_internal(oid, index_state, was_valid, flags,
689 if (!ret && !was_valid) {
690 write_locked_index(index_state, &lock_file, COMMIT_LOCK);
691 /* Not being able to write is fine -- we are only interested
692 * in updating the cache-tree part, and if the next caller
693 * ends up using the old index with unupdated cache-tree part
694 * it misses the work we did here, but that is just a
695 * performance penalty and not a big deal.
700 rollback_lock_file(&lock_file);
704 static void prime_cache_tree_rec(struct repository *r,
705 struct cache_tree *it,
708 struct tree_desc desc;
709 struct name_entry entry;
712 oidcpy(&it->oid, &tree->object.oid);
713 init_tree_desc(&desc, tree->buffer, tree->size);
715 while (tree_entry(&desc, &entry)) {
716 if (!S_ISDIR(entry.mode))
719 struct cache_tree_sub *sub;
720 struct tree *subtree = lookup_tree(r, &entry.oid);
721 if (!subtree->object.parsed)
723 sub = cache_tree_sub(it, entry.path);
724 sub->cache_tree = cache_tree();
725 prime_cache_tree_rec(r, sub->cache_tree, subtree);
726 cnt += sub->cache_tree->entry_count;
729 it->entry_count = cnt;
732 void prime_cache_tree(struct repository *r,
733 struct index_state *istate,
736 cache_tree_free(&istate->cache_tree);
737 istate->cache_tree = cache_tree();
738 prime_cache_tree_rec(r, istate->cache_tree, tree);
739 istate->cache_changed |= CACHE_TREE_CHANGED;
743 * find the cache_tree that corresponds to the current level without
744 * exploding the full path into textual form. The root of the
745 * cache tree is given as "root", and our current level is "info".
746 * (1) When at root level, info->prev is NULL, so it is "root" itself.
747 * (2) Otherwise, find the cache_tree that corresponds to one level
748 * above us, and find ourselves in there.
750 static struct cache_tree *find_cache_tree_from_traversal(struct cache_tree *root,
751 struct traverse_info *info)
753 struct cache_tree *our_parent;
757 our_parent = find_cache_tree_from_traversal(root, info->prev);
758 return cache_tree_find(our_parent, info->name);
761 int cache_tree_matches_traversal(struct cache_tree *root,
762 struct name_entry *ent,
763 struct traverse_info *info)
765 struct cache_tree *it;
767 it = find_cache_tree_from_traversal(root, info);
768 it = cache_tree_find(it, ent->path);
769 if (it && it->entry_count > 0 && oideq(&ent->oid, &it->oid))
770 return it->entry_count;
774 static void verify_one(struct repository *r,
775 struct index_state *istate,
776 struct cache_tree *it,
779 int i, pos, len = path->len;
780 struct strbuf tree_buf = STRBUF_INIT;
781 struct object_id new_oid;
783 for (i = 0; i < it->subtree_nr; i++) {
784 strbuf_addf(path, "%s/", it->down[i]->name);
785 verify_one(r, istate, it->down[i]->cache_tree, path);
786 strbuf_setlen(path, len);
789 if (it->entry_count < 0 ||
790 /* no verification on tests (t7003) that replace trees */
791 lookup_replace_object(r, &it->oid) != &it->oid)
795 pos = index_name_pos(istate, path->buf, path->len);
802 while (i < it->entry_count) {
803 struct cache_entry *ce = istate->cache[pos + i];
805 struct cache_tree_sub *sub = NULL;
806 const struct object_id *oid;
811 if (ce->ce_flags & (CE_STAGEMASK | CE_INTENT_TO_ADD | CE_REMOVE))
812 BUG("%s with flags 0x%x should not be in cache-tree",
813 ce->name, ce->ce_flags);
814 name = ce->name + path->len;
815 slash = strchr(name, '/');
817 entlen = slash - name;
818 sub = find_subtree(it, ce->name + path->len, entlen, 0);
819 if (!sub || sub->cache_tree->entry_count < 0)
820 BUG("bad subtree '%.*s'", entlen, name);
821 oid = &sub->cache_tree->oid;
823 i += sub->cache_tree->entry_count;
827 entlen = ce_namelen(ce) - path->len;
830 strbuf_addf(&tree_buf, "%o %.*s%c", mode, entlen, name, '\0');
831 strbuf_add(&tree_buf, oid->hash, r->hash_algo->rawsz);
833 hash_object_file(r->hash_algo, tree_buf.buf, tree_buf.len, tree_type,
835 if (!oideq(&new_oid, &it->oid))
836 BUG("cache-tree for path %.*s does not match. "
837 "Expected %s got %s", len, path->buf,
838 oid_to_hex(&new_oid), oid_to_hex(&it->oid));
839 strbuf_setlen(path, len);
840 strbuf_release(&tree_buf);
843 void cache_tree_verify(struct repository *r, struct index_state *istate)
845 struct strbuf path = STRBUF_INIT;
847 if (!istate->cache_tree)
849 verify_one(r, istate, istate->cache_tree, &path);
850 strbuf_release(&path);