revision: rename add_pending_sha1 to add_pending_oid
[git.git] / cache-tree.c
blob48924102123b019f3f6d69f7f07047a3d4a68d50
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) / 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 memmove(it->down + pos + 1,
88 it->down + pos,
89 sizeof(down) * (it->subtree_nr - pos - 1));
90 it->down[pos] = down;
91 return down;
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)
102 /* a/b/c
103 * ==> invalidate self
104 * ==> find "a", have it invalidate "b/c"
106 * ==> invalidate self
107 * ==> if "a" exists as a subtree, remove it.
109 const char *slash;
110 int namelen;
111 struct cache_tree_sub *down;
113 #if DEBUG
114 fprintf(stderr, "cache-tree invalidate <%s>\n", path);
115 #endif
117 if (!it)
118 return 0;
119 slash = strchrnul(path, '/');
120 namelen = slash - path;
121 it->entry_count = -1;
122 if (!*slash) {
123 int pos;
124 pos = subtree_pos(it, path, namelen);
125 if (0 <= pos) {
126 cache_tree_free(&it->down[pos]->cache_tree);
127 free(it->down[pos]);
128 /* 0 1 2 3 4 5
129 * ^ ^subtree_nr = 6
130 * pos
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));
137 it->subtree_nr--;
139 return 1;
141 down = find_subtree(it, path, namelen, 0);
142 if (down)
143 do_invalidate_path(down->cache_tree, slash + 1);
144 return 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)
156 int i, funny;
157 int silent = flags & WRITE_TREE_SILENT;
159 /* Verify that the tree is merged */
160 funny = 0;
161 for (i = 0; i < entries; i++) {
162 const struct cache_entry *ce = cache[i];
163 if (ce_stage(ce)) {
164 if (silent)
165 return -1;
166 if (10 < ++funny) {
167 fprintf(stderr, "...\n");
168 break;
170 fprintf(stderr, "%s: unmerged (%s)\n",
171 ce->name, oid_to_hex(&ce->oid));
174 if (funny)
175 return -1;
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
179 * stage 0 entries.
181 funny = 0;
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] == '/') {
193 if (10 < ++funny) {
194 fprintf(stderr, "...\n");
195 break;
197 fprintf(stderr, "You have both %s and %s\n",
198 this_name, next_name);
201 if (funny)
202 return -1;
203 return 0;
206 static void discard_unused_subtrees(struct cache_tree *it)
208 struct cache_tree_sub **down = it->down;
209 int nr = it->subtree_nr;
210 int dst, src;
211 for (dst = src = 0; src < nr; src++) {
212 struct cache_tree_sub *s = down[src];
213 if (s->used)
214 down[dst++] = s;
215 else {
216 cache_tree_free(&s->cache_tree);
217 free(s);
218 it->subtree_nr--;
223 int cache_tree_fully_valid(struct cache_tree *it)
225 int i;
226 if (!it)
227 return 0;
228 if (it->entry_count < 0 || !has_sha1_file(it->oid.hash))
229 return 0;
230 for (i = 0; i < it->subtree_nr; i++) {
231 if (!cache_tree_fully_valid(it->down[i]->cache_tree))
232 return 0;
234 return 1;
237 static int update_one(struct cache_tree *it,
238 struct cache_entry **cache,
239 int entries,
240 const char *base,
241 int baselen,
242 int *skip_count,
243 int flags)
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;
250 int i;
252 assert(!(dryrun && repair));
254 *skip_count = 0;
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.
270 i = 0;
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;
277 path = ce->name;
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, '/');
283 if (!slash) {
284 i++;
285 continue;
288 * a/bbb/c (base = a/, slash = /c)
289 * ==>
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,
298 path,
299 baselen + sublen + 1,
300 &subskip,
301 flags);
302 if (subcnt < 0)
303 return subcnt;
304 if (!subcnt)
305 die("index cache-tree records empty sub-tree");
306 i += subcnt;
307 sub->count = subcnt; /* to be used in the next loop */
308 *skip_count += subskip;
309 sub->used = 1;
312 discard_unused_subtrees(it);
315 * Then write out the tree object for this level.
317 strbuf_init(&buffer, 8192);
319 i = 0;
320 while (i < entries) {
321 const struct cache_entry *ce = cache[i];
322 struct cache_tree_sub *sub = NULL;
323 const char *path, *slash;
324 int pathlen, entlen;
325 const unsigned char *sha1;
326 unsigned mode;
327 int expected_missing = 0;
328 int contains_ita = 0;
330 path = ce->name;
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, '/');
336 if (slash) {
337 entlen = slash - (path + baselen);
338 sub = find_subtree(it, path + baselen, entlen, 0);
339 if (!sub)
340 die("cache-tree.c: '%.*s' in '%s' not found",
341 entlen, path + baselen, path);
342 i += sub->count;
343 sha1 = sub->cache_tree->oid.hash;
344 mode = S_IFDIR;
345 contains_ita = sub->cache_tree->entry_count < 0;
346 if (contains_ita) {
347 to_invalidate = 1;
348 expected_missing = 1;
351 else {
352 sha1 = ce->oid.hash;
353 mode = ce->ce_mode;
354 entlen = pathlen - baselen;
355 i++;
357 if (mode != S_IFGITLINK && !missing_ok && !has_sha1_file(sha1)) {
358 strbuf_release(&buffer);
359 if (expected_missing)
360 return -1;
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;
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 && !hashcmp(sha1, EMPTY_TREE_SHA1_BIN))
389 continue;
391 strbuf_grow(&buffer, entlen + 100);
392 strbuf_addf(&buffer, "%o %.*s%c", mode, entlen, path + baselen, '\0');
393 strbuf_add(&buffer, sha1, 20);
395 #if DEBUG
396 fprintf(stderr, "cache-tree update-one %o %.*s\n",
397 mode, entlen, path + baselen);
398 #endif
401 if (repair) {
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->oid.hash, sha1);
406 else
407 to_invalidate = 1;
408 } else if (dryrun)
409 hash_sha1_file(buffer.buf, buffer.len, tree_type,
410 it->oid.hash);
411 else if (write_sha1_file(buffer.buf, buffer.len, tree_type, it->oid.hash)) {
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, 20);
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;
496 it = NULL;
497 /* skip name, but make sure name exists */
498 while (size && *buf) {
499 size--;
500 buf++;
502 if (!size)
503 goto free_return;
504 buf++; size--;
505 it = cache_tree();
507 cp = buf;
508 it->entry_count = strtol(cp, &ep, 10);
509 if (cp == ep)
510 goto free_return;
511 cp = ep;
512 subtree_nr = strtol(cp, &ep, 10);
513 if (cp == ep)
514 goto free_return;
515 while (size && *buf && *buf != '\n') {
516 size--;
517 buf++;
519 if (!size)
520 goto free_return;
521 buf++; size--;
522 if (0 <= it->entry_count) {
523 if (size < 20)
524 goto free_return;
525 hashcpy(it->oid.hash, (const unsigned char*)buf);
526 buf += 20;
527 size -= 20;
530 #if DEBUG
531 if (0 <= it->entry_count)
532 fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n",
533 *buffer, it->entry_count, subtree_nr,
534 oid_to_hex(&it->oid));
535 else
536 fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n",
537 *buffer, subtree_nr);
538 #endif
541 * Just a heuristic -- we do not add directories that often but
542 * we do not want to have to extend it immediately when we do,
543 * hence +2.
545 it->subtree_alloc = subtree_nr + 2;
546 it->down = xcalloc(it->subtree_alloc, sizeof(struct cache_tree_sub *));
547 for (i = 0; i < subtree_nr; i++) {
548 /* read each subtree */
549 struct cache_tree *sub;
550 struct cache_tree_sub *subtree;
551 const char *name = buf;
553 sub = read_one(&buf, &size);
554 if (!sub)
555 goto free_return;
556 subtree = cache_tree_sub(it, name);
557 subtree->cache_tree = sub;
559 if (subtree_nr != it->subtree_nr)
560 die("cache-tree: internal error");
561 *buffer = buf;
562 *size_p = size;
563 return it;
565 free_return:
566 cache_tree_free(&it);
567 return NULL;
570 struct cache_tree *cache_tree_read(const char *buffer, unsigned long size)
572 if (buffer[0])
573 return NULL; /* not the whole tree */
574 return read_one(&buffer, &size);
577 static struct cache_tree *cache_tree_find(struct cache_tree *it, const char *path)
579 if (!it)
580 return NULL;
581 while (*path) {
582 const char *slash;
583 struct cache_tree_sub *sub;
585 slash = strchrnul(path, '/');
587 * Between path and slash is the name of the subtree
588 * to look for.
590 sub = find_subtree(it, path, slash - path, 0);
591 if (!sub)
592 return NULL;
593 it = sub->cache_tree;
595 path = slash;
596 while (*path == '/')
597 path++;
599 return it;
602 int write_index_as_tree(unsigned char *sha1, struct index_state *index_state, const char *index_path, int flags, const char *prefix)
604 int entries, was_valid, newfd;
605 struct lock_file *lock_file;
608 * We can't free this memory, it becomes part of a linked list
609 * parsed atexit()
611 lock_file = xcalloc(1, sizeof(struct lock_file));
613 newfd = hold_lock_file_for_update(lock_file, index_path, LOCK_DIE_ON_ERROR);
615 entries = read_index_from(index_state, index_path);
616 if (entries < 0)
617 return WRITE_TREE_UNREADABLE_INDEX;
618 if (flags & WRITE_TREE_IGNORE_CACHE_TREE)
619 cache_tree_free(&index_state->cache_tree);
621 if (!index_state->cache_tree)
622 index_state->cache_tree = cache_tree();
624 was_valid = cache_tree_fully_valid(index_state->cache_tree);
625 if (!was_valid) {
626 if (cache_tree_update(index_state, flags) < 0)
627 return WRITE_TREE_UNMERGED_INDEX;
628 if (0 <= newfd) {
629 if (!write_locked_index(index_state, lock_file, COMMIT_LOCK))
630 newfd = -1;
632 /* Not being able to write is fine -- we are only interested
633 * in updating the cache-tree part, and if the next caller
634 * ends up using the old index with unupdated cache-tree part
635 * it misses the work we did here, but that is just a
636 * performance penalty and not a big deal.
640 if (prefix) {
641 struct cache_tree *subtree;
642 subtree = cache_tree_find(index_state->cache_tree, prefix);
643 if (!subtree)
644 return WRITE_TREE_PREFIX_ERROR;
645 hashcpy(sha1, subtree->oid.hash);
647 else
648 hashcpy(sha1, index_state->cache_tree->oid.hash);
650 if (0 <= newfd)
651 rollback_lock_file(lock_file);
653 return 0;
656 int write_cache_as_tree(unsigned char *sha1, int flags, const char *prefix)
658 return write_index_as_tree(sha1, &the_index, get_index_file(), flags, prefix);
661 static void prime_cache_tree_rec(struct cache_tree *it, struct tree *tree)
663 struct tree_desc desc;
664 struct name_entry entry;
665 int cnt;
667 oidcpy(&it->oid, &tree->object.oid);
668 init_tree_desc(&desc, tree->buffer, tree->size);
669 cnt = 0;
670 while (tree_entry(&desc, &entry)) {
671 if (!S_ISDIR(entry.mode))
672 cnt++;
673 else {
674 struct cache_tree_sub *sub;
675 struct tree *subtree = lookup_tree(entry.oid);
676 if (!subtree->object.parsed)
677 parse_tree(subtree);
678 sub = cache_tree_sub(it, entry.path);
679 sub->cache_tree = cache_tree();
680 prime_cache_tree_rec(sub->cache_tree, subtree);
681 cnt += sub->cache_tree->entry_count;
684 it->entry_count = cnt;
687 void prime_cache_tree(struct index_state *istate, struct tree *tree)
689 cache_tree_free(&istate->cache_tree);
690 istate->cache_tree = cache_tree();
691 prime_cache_tree_rec(istate->cache_tree, tree);
692 istate->cache_changed |= CACHE_TREE_CHANGED;
696 * find the cache_tree that corresponds to the current level without
697 * exploding the full path into textual form. The root of the
698 * cache tree is given as "root", and our current level is "info".
699 * (1) When at root level, info->prev is NULL, so it is "root" itself.
700 * (2) Otherwise, find the cache_tree that corresponds to one level
701 * above us, and find ourselves in there.
703 static struct cache_tree *find_cache_tree_from_traversal(struct cache_tree *root,
704 struct traverse_info *info)
706 struct cache_tree *our_parent;
708 if (!info->prev)
709 return root;
710 our_parent = find_cache_tree_from_traversal(root, info->prev);
711 return cache_tree_find(our_parent, info->name.path);
714 int cache_tree_matches_traversal(struct cache_tree *root,
715 struct name_entry *ent,
716 struct traverse_info *info)
718 struct cache_tree *it;
720 it = find_cache_tree_from_traversal(root, info);
721 it = cache_tree_find(it, ent->path);
722 if (it && it->entry_count > 0 && !oidcmp(ent->oid, &it->oid))
723 return it->entry_count;
724 return 0;
727 int update_main_cache_tree(int flags)
729 if (!the_index.cache_tree)
730 the_index.cache_tree = cache_tree();
731 return cache_tree_update(&the_index, flags);