5 #include "cache-tree.h"
11 struct cache_tree *cache_tree(void)
13 struct cache_tree *it = xcalloc(1, sizeof(struct cache_tree));
18 void cache_tree_free(struct cache_tree **it_p)
21 struct cache_tree *it = *it_p;
25 for (i = 0; i < it->subtree_nr; i++)
27 cache_tree_free(&it->down[i]->cache_tree);
35 static int subtree_name_cmp(const char *one, int onelen,
36 const char *two, int twolen)
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;
52 int mi = (lo + hi) / 2;
53 struct cache_tree_sub *mdl = down[mi];
54 int cmp = subtree_name_cmp(path, pathlen,
55 mdl->name, mdl->namelen);
66 static struct cache_tree_sub *find_subtree(struct cache_tree *it,
71 struct cache_tree_sub *down;
72 int pos = subtree_pos(it, path, pathlen);
79 ALLOC_GROW(it->down, it->subtree_nr + 1, it->subtree_alloc);
82 FLEX_ALLOC_MEM(down, name, path, pathlen);
83 down->cache_tree = NULL;
84 down->namelen = pathlen;
86 if (pos < it->subtree_nr)
87 memmove(it->down + pos + 1,
89 sizeof(down) * (it->subtree_nr - pos - 1));
94 struct cache_tree_sub *cache_tree_sub(struct cache_tree *it, const char *path)
96 int pathlen = strlen(path);
97 return find_subtree(it, path, pathlen, 1);
100 static int do_invalidate_path(struct cache_tree *it, const char *path)
103 * ==> invalidate self
104 * ==> find "a", have it invalidate "b/c"
106 * ==> invalidate self
107 * ==> if "a" exists as a subtree, remove it.
111 struct cache_tree_sub *down;
114 fprintf(stderr, "cache-tree invalidate <%s>\n", path);
119 slash = strchrnul(path, '/');
120 namelen = slash - path;
121 it->entry_count = -1;
124 pos = subtree_pos(it, path, namelen);
126 cache_tree_free(&it->down[pos]->cache_tree);
131 * move 4 and 5 up one place (2 entries)
132 * 2 = 6 - 3 - 1 = subtree_nr - pos - 1
134 memmove(it->down+pos, it->down+pos+1,
135 sizeof(struct cache_tree_sub *) *
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, sha1_to_hex(ce->sha1));
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->sha1))
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->sha1))
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 unsigned char *sha1;
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 sha1 = sub->cache_tree->sha1;
345 contains_ita = sub->cache_tree->entry_count < 0;
348 expected_missing = 1;
354 entlen = pathlen - baselen;
357 if (mode != S_IFGITLINK && !missing_ok && !has_sha1_file(sha1)) {
358 strbuf_release(&buffer);
359 if (expected_missing)
361 return error("invalid object %06o %s for '%.*s'",
362 mode, sha1_to_hex(sha1), 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;
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)) {
386 * "sub" can be an empty tree if all subentries are i-t-a.
388 if (contains_ita && !hashcmp(sha1, EMPTY_TREE_SHA1_BIN))
391 strbuf_grow(&buffer, entlen + 100);
392 strbuf_addf(&buffer, "%o %.*s%c", mode, entlen, path + baselen, '\0');
393 strbuf_add(&buffer, sha1, 20);
396 fprintf(stderr, "cache-tree update-one %o %.*s\n",
397 mode, entlen, path + baselen);
402 unsigned char sha1[20];
403 hash_sha1_file(buffer.buf, buffer.len, tree_type, sha1);
404 if (has_sha1_file(sha1))
405 hashcpy(it->sha1, sha1);
409 hash_sha1_file(buffer.buf, buffer.len, tree_type, it->sha1);
410 else if (write_sha1_file(buffer.buf, buffer.len, tree_type, it->sha1)) {
411 strbuf_release(&buffer);
415 strbuf_release(&buffer);
416 it->entry_count = to_invalidate ? -1 : i - *skip_count;
418 fprintf(stderr, "cache-tree update-one (%d ent, %d subtree) %s\n",
419 it->entry_count, it->subtree_nr,
420 sha1_to_hex(it->sha1));
425 int cache_tree_update(struct index_state *istate, int flags)
427 struct cache_tree *it = istate->cache_tree;
428 struct cache_entry **cache = istate->cache;
429 int entries = istate->cache_nr;
430 int skip, i = verify_cache(cache, entries, flags);
434 i = update_one(it, cache, entries, "", 0, &skip, flags);
437 istate->cache_changed |= CACHE_TREE_CHANGED;
441 static void write_one(struct strbuf *buffer, struct cache_tree *it,
442 const char *path, int pathlen)
446 /* One "cache-tree" entry consists of the following:
447 * path (NUL terminated)
448 * entry_count, subtree_nr ("%d %d\n")
449 * tree-sha1 (missing if invalid)
450 * subtree_nr "cache-tree" entries for subtrees.
452 strbuf_grow(buffer, pathlen + 100);
453 strbuf_add(buffer, path, pathlen);
454 strbuf_addf(buffer, "%c%d %d\n", 0, it->entry_count, it->subtree_nr);
457 if (0 <= it->entry_count)
458 fprintf(stderr, "cache-tree <%.*s> (%d ent, %d subtree) %s\n",
459 pathlen, path, it->entry_count, it->subtree_nr,
460 sha1_to_hex(it->sha1));
462 fprintf(stderr, "cache-tree <%.*s> (%d subtree) invalid\n",
463 pathlen, path, it->subtree_nr);
466 if (0 <= it->entry_count) {
467 strbuf_add(buffer, it->sha1, 20);
469 for (i = 0; i < it->subtree_nr; i++) {
470 struct cache_tree_sub *down = it->down[i];
472 struct cache_tree_sub *prev = it->down[i-1];
473 if (subtree_name_cmp(down->name, down->namelen,
474 prev->name, prev->namelen) <= 0)
475 die("fatal - unsorted cache subtree");
477 write_one(buffer, down->cache_tree, down->name, down->namelen);
481 void cache_tree_write(struct strbuf *sb, struct cache_tree *root)
483 write_one(sb, root, "", 0);
486 static struct cache_tree *read_one(const char **buffer, unsigned long *size_p)
488 const char *buf = *buffer;
489 unsigned long size = *size_p;
492 struct cache_tree *it;
496 /* skip name, but make sure name exists */
497 while (size && *buf) {
507 it->entry_count = strtol(cp, &ep, 10);
511 subtree_nr = strtol(cp, &ep, 10);
514 while (size && *buf && *buf != '\n') {
521 if (0 <= it->entry_count) {
524 hashcpy(it->sha1, (const unsigned char*)buf);
530 if (0 <= it->entry_count)
531 fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n",
532 *buffer, it->entry_count, subtree_nr,
533 sha1_to_hex(it->sha1));
535 fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n",
536 *buffer, subtree_nr);
540 * Just a heuristic -- we do not add directories that often but
541 * we do not want to have to extend it immediately when we do,
544 it->subtree_alloc = subtree_nr + 2;
545 it->down = xcalloc(it->subtree_alloc, sizeof(struct cache_tree_sub *));
546 for (i = 0; i < subtree_nr; i++) {
547 /* read each subtree */
548 struct cache_tree *sub;
549 struct cache_tree_sub *subtree;
550 const char *name = buf;
552 sub = read_one(&buf, &size);
555 subtree = cache_tree_sub(it, name);
556 subtree->cache_tree = sub;
558 if (subtree_nr != it->subtree_nr)
559 die("cache-tree: internal error");
565 cache_tree_free(&it);
569 struct cache_tree *cache_tree_read(const char *buffer, unsigned long size)
572 return NULL; /* not the whole tree */
573 return read_one(&buffer, &size);
576 static struct cache_tree *cache_tree_find(struct cache_tree *it, const char *path)
582 struct cache_tree_sub *sub;
584 slash = strchrnul(path, '/');
586 * Between path and slash is the name of the subtree
589 sub = find_subtree(it, path, slash - path, 0);
592 it = sub->cache_tree;
601 int write_index_as_tree(unsigned char *sha1, struct index_state *index_state, const char *index_path, int flags, const char *prefix)
603 int entries, was_valid, newfd;
604 struct lock_file *lock_file;
607 * We can't free this memory, it becomes part of a linked list
610 lock_file = xcalloc(1, sizeof(struct lock_file));
612 newfd = hold_lock_file_for_update(lock_file, index_path, LOCK_DIE_ON_ERROR);
614 entries = read_index_from(index_state, index_path);
616 return WRITE_TREE_UNREADABLE_INDEX;
617 if (flags & WRITE_TREE_IGNORE_CACHE_TREE)
618 cache_tree_free(&index_state->cache_tree);
620 if (!index_state->cache_tree)
621 index_state->cache_tree = cache_tree();
623 was_valid = cache_tree_fully_valid(index_state->cache_tree);
625 if (cache_tree_update(index_state, flags) < 0)
626 return WRITE_TREE_UNMERGED_INDEX;
628 if (!write_locked_index(index_state, lock_file, COMMIT_LOCK))
631 /* Not being able to write is fine -- we are only interested
632 * in updating the cache-tree part, and if the next caller
633 * ends up using the old index with unupdated cache-tree part
634 * it misses the work we did here, but that is just a
635 * performance penalty and not a big deal.
640 struct cache_tree *subtree;
641 subtree = cache_tree_find(index_state->cache_tree, prefix);
643 return WRITE_TREE_PREFIX_ERROR;
644 hashcpy(sha1, subtree->sha1);
647 hashcpy(sha1, index_state->cache_tree->sha1);
650 rollback_lock_file(lock_file);
655 int write_cache_as_tree(unsigned char *sha1, int flags, const char *prefix)
657 return write_index_as_tree(sha1, &the_index, get_index_file(), flags, prefix);
660 static void prime_cache_tree_rec(struct cache_tree *it, struct tree *tree)
662 struct tree_desc desc;
663 struct name_entry entry;
666 hashcpy(it->sha1, tree->object.oid.hash);
667 init_tree_desc(&desc, tree->buffer, tree->size);
669 while (tree_entry(&desc, &entry)) {
670 if (!S_ISDIR(entry.mode))
673 struct cache_tree_sub *sub;
674 struct tree *subtree = lookup_tree(entry.oid->hash);
675 if (!subtree->object.parsed)
677 sub = cache_tree_sub(it, entry.path);
678 sub->cache_tree = cache_tree();
679 prime_cache_tree_rec(sub->cache_tree, subtree);
680 cnt += sub->cache_tree->entry_count;
683 it->entry_count = cnt;
686 void prime_cache_tree(struct index_state *istate, struct tree *tree)
688 cache_tree_free(&istate->cache_tree);
689 istate->cache_tree = cache_tree();
690 prime_cache_tree_rec(istate->cache_tree, tree);
691 istate->cache_changed |= CACHE_TREE_CHANGED;
695 * find the cache_tree that corresponds to the current level without
696 * exploding the full path into textual form. The root of the
697 * cache tree is given as "root", and our current level is "info".
698 * (1) When at root level, info->prev is NULL, so it is "root" itself.
699 * (2) Otherwise, find the cache_tree that corresponds to one level
700 * above us, and find ourselves in there.
702 static struct cache_tree *find_cache_tree_from_traversal(struct cache_tree *root,
703 struct traverse_info *info)
705 struct cache_tree *our_parent;
709 our_parent = find_cache_tree_from_traversal(root, info->prev);
710 return cache_tree_find(our_parent, info->name.path);
713 int cache_tree_matches_traversal(struct cache_tree *root,
714 struct name_entry *ent,
715 struct traverse_info *info)
717 struct cache_tree *it;
719 it = find_cache_tree_from_traversal(root, info);
720 it = cache_tree_find(it, ent->path);
721 if (it && it->entry_count > 0 && !hashcmp(ent->oid->hash, it->sha1))
722 return it->entry_count;
726 int update_main_cache_tree(int flags)
728 if (!the_index.cache_tree)
729 the_index.cache_tree = cache_tree();
730 return cache_tree_update(&the_index, flags);