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 int cache_tree_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 = cache_tree_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 = cache_tree_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 index_state *istate, int flags)
157 int silent = flags & WRITE_TREE_SILENT;
159 /* Verify that the tree is merged */
161 for (i = 0; i < istate->cache_nr; i++) {
162 const struct cache_entry *ce = istate->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 + 1 < istate->cache_nr; 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 struct cache_entry *this_ce = istate->cache[i];
188 const struct cache_entry *next_ce = istate->cache[i + 1];
189 const char *this_name = this_ce->name;
190 const char *next_name = next_ce->name;
191 int this_len = ce_namelen(this_ce);
192 if (this_len < ce_namelen(next_ce) &&
193 next_name[this_len] == '/' &&
194 strncmp(this_name, next_name, this_len) == 0) {
196 fprintf(stderr, "...\n");
199 fprintf(stderr, "You have both %s and %s\n",
200 this_name, next_name);
208 static void discard_unused_subtrees(struct cache_tree *it)
210 struct cache_tree_sub **down = it->down;
211 int nr = it->subtree_nr;
213 for (dst = src = 0; src < nr; src++) {
214 struct cache_tree_sub *s = down[src];
218 cache_tree_free(&s->cache_tree);
225 int cache_tree_fully_valid(struct cache_tree *it)
230 if (it->entry_count < 0 || !has_object_file(&it->oid))
232 for (i = 0; i < it->subtree_nr; i++) {
233 if (!cache_tree_fully_valid(it->down[i]->cache_tree))
239 static int update_one(struct cache_tree *it,
240 struct cache_entry **cache,
247 struct strbuf buffer;
248 int missing_ok = flags & WRITE_TREE_MISSING_OK;
249 int dryrun = flags & WRITE_TREE_DRY_RUN;
250 int repair = flags & WRITE_TREE_REPAIR;
251 int to_invalidate = 0;
254 assert(!(dryrun && repair));
258 if (0 <= it->entry_count && has_object_file(&it->oid))
259 return it->entry_count;
262 * We first scan for subtrees and update them; we start by
263 * marking existing subtrees -- the ones that are unmarked
264 * should not be in the result.
266 for (i = 0; i < it->subtree_nr; i++)
267 it->down[i]->used = 0;
270 * Find the subtrees and update them.
273 while (i < entries) {
274 const struct cache_entry *ce = cache[i];
275 struct cache_tree_sub *sub;
276 const char *path, *slash;
277 int pathlen, sublen, subcnt, subskip;
280 pathlen = ce_namelen(ce);
281 if (pathlen <= baselen || memcmp(base, path, baselen))
282 break; /* at the end of this level */
284 slash = strchr(path + baselen, '/');
290 * a/bbb/c (base = a/, slash = /c)
292 * path+baselen = bbb/c, sublen = 3
294 sublen = slash - (path + baselen);
295 sub = find_subtree(it, path + baselen, sublen, 1);
296 if (!sub->cache_tree)
297 sub->cache_tree = cache_tree();
298 subcnt = update_one(sub->cache_tree,
299 cache + i, entries - i,
301 baselen + sublen + 1,
307 die("index cache-tree records empty sub-tree");
309 sub->count = subcnt; /* to be used in the next loop */
310 *skip_count += subskip;
314 discard_unused_subtrees(it);
317 * Then write out the tree object for this level.
319 strbuf_init(&buffer, 8192);
322 while (i < entries) {
323 const struct cache_entry *ce = cache[i];
324 struct cache_tree_sub *sub = NULL;
325 const char *path, *slash;
327 const struct object_id *oid;
329 int expected_missing = 0;
330 int contains_ita = 0;
334 pathlen = ce_namelen(ce);
335 if (pathlen <= baselen || memcmp(base, path, baselen))
336 break; /* at the end of this level */
338 slash = strchr(path + baselen, '/');
340 entlen = slash - (path + baselen);
341 sub = find_subtree(it, path + baselen, entlen, 0);
343 die("cache-tree.c: '%.*s' in '%s' not found",
344 entlen, path + baselen, path);
346 oid = &sub->cache_tree->oid;
348 contains_ita = sub->cache_tree->entry_count < 0;
351 expected_missing = 1;
357 entlen = pathlen - baselen;
361 ce_missing_ok = mode == S_IFGITLINK || missing_ok ||
362 (has_promisor_remote() &&
363 ce_skip_worktree(ce));
364 if (is_null_oid(oid) ||
365 (!ce_missing_ok && !has_object_file(oid))) {
366 strbuf_release(&buffer);
367 if (expected_missing)
369 return error("invalid object %06o %s for '%.*s'",
370 mode, oid_to_hex(oid), entlen+baselen, path);
374 * CE_REMOVE entries are removed before the index is
375 * written to disk. Skip them to remain consistent
376 * with the future on-disk index.
378 if (ce->ce_flags & CE_REMOVE) {
379 *skip_count = *skip_count + 1;
384 * CE_INTENT_TO_ADD entries exist on on-disk index but
385 * they are not part of generated trees. Invalidate up
386 * to root to force cache-tree users to read elsewhere.
388 if (!sub && ce_intent_to_add(ce)) {
394 * "sub" can be an empty tree if all subentries are i-t-a.
396 if (contains_ita && is_empty_tree_oid(oid))
399 strbuf_grow(&buffer, entlen + 100);
400 strbuf_addf(&buffer, "%o %.*s%c", mode, entlen, path + baselen, '\0');
401 strbuf_add(&buffer, oid->hash, the_hash_algo->rawsz);
404 fprintf(stderr, "cache-tree update-one %o %.*s\n",
405 mode, entlen, path + baselen);
410 struct object_id oid;
411 hash_object_file(the_hash_algo, buffer.buf, buffer.len,
413 if (has_object_file_with_flags(&oid, OBJECT_INFO_SKIP_FETCH_OBJECT))
414 oidcpy(&it->oid, &oid);
418 hash_object_file(the_hash_algo, buffer.buf, buffer.len,
419 tree_type, &it->oid);
420 } else if (write_object_file(buffer.buf, buffer.len, tree_type,
422 strbuf_release(&buffer);
426 strbuf_release(&buffer);
427 it->entry_count = to_invalidate ? -1 : i - *skip_count;
429 fprintf(stderr, "cache-tree update-one (%d ent, %d subtree) %s\n",
430 it->entry_count, it->subtree_nr,
431 oid_to_hex(&it->oid));
436 int cache_tree_update(struct index_state *istate, int flags)
440 i = verify_cache(istate, flags);
445 if (!istate->cache_tree)
446 istate->cache_tree = cache_tree();
448 trace_performance_enter();
449 trace2_region_enter("cache_tree", "update", the_repository);
450 i = update_one(istate->cache_tree, istate->cache, istate->cache_nr,
451 "", 0, &skip, flags);
452 trace2_region_leave("cache_tree", "update", the_repository);
453 trace_performance_leave("cache_tree_update");
456 istate->cache_changed |= CACHE_TREE_CHANGED;
460 static void write_one(struct strbuf *buffer, struct cache_tree *it,
461 const char *path, int pathlen)
465 /* One "cache-tree" entry consists of the following:
466 * path (NUL terminated)
467 * entry_count, subtree_nr ("%d %d\n")
468 * tree-sha1 (missing if invalid)
469 * subtree_nr "cache-tree" entries for subtrees.
471 strbuf_grow(buffer, pathlen + 100);
472 strbuf_add(buffer, path, pathlen);
473 strbuf_addf(buffer, "%c%d %d\n", 0, it->entry_count, it->subtree_nr);
476 if (0 <= it->entry_count)
477 fprintf(stderr, "cache-tree <%.*s> (%d ent, %d subtree) %s\n",
478 pathlen, path, it->entry_count, it->subtree_nr,
479 oid_to_hex(&it->oid));
481 fprintf(stderr, "cache-tree <%.*s> (%d subtree) invalid\n",
482 pathlen, path, it->subtree_nr);
485 if (0 <= it->entry_count) {
486 strbuf_add(buffer, it->oid.hash, the_hash_algo->rawsz);
488 for (i = 0; i < it->subtree_nr; i++) {
489 struct cache_tree_sub *down = it->down[i];
491 struct cache_tree_sub *prev = it->down[i-1];
492 if (subtree_name_cmp(down->name, down->namelen,
493 prev->name, prev->namelen) <= 0)
494 die("fatal - unsorted cache subtree");
496 write_one(buffer, down->cache_tree, down->name, down->namelen);
500 void cache_tree_write(struct strbuf *sb, struct cache_tree *root)
502 trace2_region_enter("cache_tree", "write", the_repository);
503 write_one(sb, root, "", 0);
504 trace2_region_leave("cache_tree", "write", the_repository);
507 static struct cache_tree *read_one(const char **buffer, unsigned long *size_p)
509 const char *buf = *buffer;
510 unsigned long size = *size_p;
513 struct cache_tree *it;
515 const unsigned rawsz = the_hash_algo->rawsz;
518 /* skip name, but make sure name exists */
519 while (size && *buf) {
529 it->entry_count = strtol(cp, &ep, 10);
533 subtree_nr = strtol(cp, &ep, 10);
536 while (size && *buf && *buf != '\n') {
543 if (0 <= it->entry_count) {
546 oidread(&it->oid, (const unsigned char *)buf);
552 if (0 <= it->entry_count)
553 fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n",
554 *buffer, it->entry_count, subtree_nr,
555 oid_to_hex(&it->oid));
557 fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n",
558 *buffer, subtree_nr);
562 * Just a heuristic -- we do not add directories that often but
563 * we do not want to have to extend it immediately when we do,
566 it->subtree_alloc = subtree_nr + 2;
567 CALLOC_ARRAY(it->down, it->subtree_alloc);
568 for (i = 0; i < subtree_nr; i++) {
569 /* read each subtree */
570 struct cache_tree *sub;
571 struct cache_tree_sub *subtree;
572 const char *name = buf;
574 sub = read_one(&buf, &size);
577 subtree = cache_tree_sub(it, name);
578 subtree->cache_tree = sub;
580 if (subtree_nr != it->subtree_nr)
581 die("cache-tree: internal error");
587 cache_tree_free(&it);
591 struct cache_tree *cache_tree_read(const char *buffer, unsigned long size)
593 struct cache_tree *result;
596 return NULL; /* not the whole tree */
598 trace2_region_enter("cache_tree", "read", the_repository);
599 result = read_one(&buffer, &size);
600 trace2_region_leave("cache_tree", "read", the_repository);
605 static struct cache_tree *cache_tree_find(struct cache_tree *it, const char *path)
611 struct cache_tree_sub *sub;
613 slash = strchrnul(path, '/');
615 * Between path and slash is the name of the subtree
618 sub = find_subtree(it, path, slash - path, 0);
621 it = sub->cache_tree;
630 static int write_index_as_tree_internal(struct object_id *oid,
631 struct index_state *index_state,
632 int cache_tree_valid,
636 if (flags & WRITE_TREE_IGNORE_CACHE_TREE) {
637 cache_tree_free(&index_state->cache_tree);
638 cache_tree_valid = 0;
641 if (!cache_tree_valid && cache_tree_update(index_state, flags) < 0)
642 return WRITE_TREE_UNMERGED_INDEX;
645 struct cache_tree *subtree;
646 subtree = cache_tree_find(index_state->cache_tree, prefix);
648 return WRITE_TREE_PREFIX_ERROR;
649 oidcpy(oid, &subtree->oid);
652 oidcpy(oid, &index_state->cache_tree->oid);
657 struct tree* write_in_core_index_as_tree(struct repository *repo) {
661 struct index_state *index_state = repo->index;
662 was_valid = index_state->cache_tree &&
663 cache_tree_fully_valid(index_state->cache_tree);
665 ret = write_index_as_tree_internal(&o, index_state, was_valid, 0, NULL);
666 if (ret == WRITE_TREE_UNMERGED_INDEX) {
668 fprintf(stderr, "BUG: There are unmerged index entries:\n");
669 for (i = 0; i < index_state->cache_nr; i++) {
670 const struct cache_entry *ce = index_state->cache[i];
672 fprintf(stderr, "BUG: %d %.*s\n", ce_stage(ce),
673 (int)ce_namelen(ce), ce->name);
675 BUG("unmerged index entries when writing inmemory index");
678 return lookup_tree(repo, &index_state->cache_tree->oid);
682 int write_index_as_tree(struct object_id *oid, struct index_state *index_state, const char *index_path, int flags, const char *prefix)
684 int entries, was_valid;
685 struct lock_file lock_file = LOCK_INIT;
688 hold_lock_file_for_update(&lock_file, index_path, LOCK_DIE_ON_ERROR);
690 entries = read_index_from(index_state, index_path, get_git_dir());
692 ret = WRITE_TREE_UNREADABLE_INDEX;
696 was_valid = !(flags & WRITE_TREE_IGNORE_CACHE_TREE) &&
697 index_state->cache_tree &&
698 cache_tree_fully_valid(index_state->cache_tree);
700 ret = write_index_as_tree_internal(oid, index_state, was_valid, flags,
702 if (!ret && !was_valid) {
703 write_locked_index(index_state, &lock_file, COMMIT_LOCK);
704 /* Not being able to write is fine -- we are only interested
705 * in updating the cache-tree part, and if the next caller
706 * ends up using the old index with unupdated cache-tree part
707 * it misses the work we did here, but that is just a
708 * performance penalty and not a big deal.
713 rollback_lock_file(&lock_file);
717 static void prime_cache_tree_rec(struct repository *r,
718 struct cache_tree *it,
721 struct tree_desc desc;
722 struct name_entry entry;
725 oidcpy(&it->oid, &tree->object.oid);
726 init_tree_desc(&desc, tree->buffer, tree->size);
728 while (tree_entry(&desc, &entry)) {
729 if (!S_ISDIR(entry.mode))
732 struct cache_tree_sub *sub;
733 struct tree *subtree = lookup_tree(r, &entry.oid);
734 if (!subtree->object.parsed)
736 sub = cache_tree_sub(it, entry.path);
737 sub->cache_tree = cache_tree();
738 prime_cache_tree_rec(r, sub->cache_tree, subtree);
739 cnt += sub->cache_tree->entry_count;
742 it->entry_count = cnt;
745 void prime_cache_tree(struct repository *r,
746 struct index_state *istate,
749 trace2_region_enter("cache-tree", "prime_cache_tree", the_repository);
750 cache_tree_free(&istate->cache_tree);
751 istate->cache_tree = cache_tree();
753 prime_cache_tree_rec(r, istate->cache_tree, tree);
754 istate->cache_changed |= CACHE_TREE_CHANGED;
755 trace2_region_leave("cache-tree", "prime_cache_tree", the_repository);
759 * find the cache_tree that corresponds to the current level without
760 * exploding the full path into textual form. The root of the
761 * cache tree is given as "root", and our current level is "info".
762 * (1) When at root level, info->prev is NULL, so it is "root" itself.
763 * (2) Otherwise, find the cache_tree that corresponds to one level
764 * above us, and find ourselves in there.
766 static struct cache_tree *find_cache_tree_from_traversal(struct cache_tree *root,
767 struct traverse_info *info)
769 struct cache_tree *our_parent;
773 our_parent = find_cache_tree_from_traversal(root, info->prev);
774 return cache_tree_find(our_parent, info->name);
777 int cache_tree_matches_traversal(struct cache_tree *root,
778 struct name_entry *ent,
779 struct traverse_info *info)
781 struct cache_tree *it;
783 it = find_cache_tree_from_traversal(root, info);
784 it = cache_tree_find(it, ent->path);
785 if (it && it->entry_count > 0 && oideq(&ent->oid, &it->oid))
786 return it->entry_count;
790 static void verify_one(struct repository *r,
791 struct index_state *istate,
792 struct cache_tree *it,
795 int i, pos, len = path->len;
796 struct strbuf tree_buf = STRBUF_INIT;
797 struct object_id new_oid;
799 for (i = 0; i < it->subtree_nr; i++) {
800 strbuf_addf(path, "%s/", it->down[i]->name);
801 verify_one(r, istate, it->down[i]->cache_tree, path);
802 strbuf_setlen(path, len);
805 if (it->entry_count < 0 ||
806 /* no verification on tests (t7003) that replace trees */
807 lookup_replace_object(r, &it->oid) != &it->oid)
811 pos = index_name_pos(istate, path->buf, path->len);
818 while (i < it->entry_count) {
819 struct cache_entry *ce = istate->cache[pos + i];
821 struct cache_tree_sub *sub = NULL;
822 const struct object_id *oid;
827 if (ce->ce_flags & (CE_STAGEMASK | CE_INTENT_TO_ADD | CE_REMOVE))
828 BUG("%s with flags 0x%x should not be in cache-tree",
829 ce->name, ce->ce_flags);
830 name = ce->name + path->len;
831 slash = strchr(name, '/');
833 entlen = slash - name;
834 sub = find_subtree(it, ce->name + path->len, entlen, 0);
835 if (!sub || sub->cache_tree->entry_count < 0)
836 BUG("bad subtree '%.*s'", entlen, name);
837 oid = &sub->cache_tree->oid;
839 i += sub->cache_tree->entry_count;
843 entlen = ce_namelen(ce) - path->len;
846 strbuf_addf(&tree_buf, "%o %.*s%c", mode, entlen, name, '\0');
847 strbuf_add(&tree_buf, oid->hash, r->hash_algo->rawsz);
849 hash_object_file(r->hash_algo, tree_buf.buf, tree_buf.len, tree_type,
851 if (!oideq(&new_oid, &it->oid))
852 BUG("cache-tree for path %.*s does not match. "
853 "Expected %s got %s", len, path->buf,
854 oid_to_hex(&new_oid), oid_to_hex(&it->oid));
855 strbuf_setlen(path, len);
856 strbuf_release(&tree_buf);
859 void cache_tree_verify(struct repository *r, struct index_state *istate)
861 struct strbuf path = STRBUF_INIT;
863 if (!istate->cache_tree)
865 verify_one(r, istate, istate->cache_tree, &path);
866 strbuf_release(&path);