Sync with maint
[git.git] / cache-tree.c
blobc53f7de2b13acfc1b97c60b04552e57791377442
1 #include "cache.h"
2 #include "tree.h"
3 #include "tree-walk.h"
4 #include "cache-tree.h"
6 #ifndef DEBUG
7 #define DEBUG 0
8 #endif
10 struct cache_tree *cache_tree(void)
12 struct cache_tree *it = xcalloc(1, sizeof(struct cache_tree));
13 it->entry_count = -1;
14 return it;
17 void cache_tree_free(struct cache_tree **it_p)
19 int i;
20 struct cache_tree *it = *it_p;
22 if (!it)
23 return;
24 for (i = 0; i < it->subtree_nr; i++)
25 if (it->down[i]) {
26 cache_tree_free(&it->down[i]->cache_tree);
27 free(it->down[i]);
29 free(it->down);
30 free(it);
31 *it_p = NULL;
34 static int subtree_name_cmp(const char *one, int onelen,
35 const char *two, int twolen)
37 if (onelen < twolen)
38 return -1;
39 if (twolen < onelen)
40 return 1;
41 return memcmp(one, two, onelen);
44 static int subtree_pos(struct cache_tree *it, const char *path, int pathlen)
46 struct cache_tree_sub **down = it->down;
47 int lo, hi;
48 lo = 0;
49 hi = it->subtree_nr;
50 while (lo < hi) {
51 int mi = (lo + hi) / 2;
52 struct cache_tree_sub *mdl = down[mi];
53 int cmp = subtree_name_cmp(path, pathlen,
54 mdl->name, mdl->namelen);
55 if (!cmp)
56 return mi;
57 if (cmp < 0)
58 hi = mi;
59 else
60 lo = mi + 1;
62 return -lo-1;
65 static struct cache_tree_sub *find_subtree(struct cache_tree *it,
66 const char *path,
67 int pathlen,
68 int create)
70 struct cache_tree_sub *down;
71 int pos = subtree_pos(it, path, pathlen);
72 if (0 <= pos)
73 return it->down[pos];
74 if (!create)
75 return NULL;
77 pos = -pos-1;
78 ALLOC_GROW(it->down, it->subtree_nr + 1, it->subtree_alloc);
79 it->subtree_nr++;
81 down = xmalloc(sizeof(*down) + pathlen + 1);
82 down->cache_tree = NULL;
83 down->namelen = pathlen;
84 memcpy(down->name, path, pathlen);
85 down->name[pathlen] = 0;
87 if (pos < it->subtree_nr)
88 memmove(it->down + pos + 1,
89 it->down + pos,
90 sizeof(down) * (it->subtree_nr - pos - 1));
91 it->down[pos] = down;
92 return down;
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)
103 /* a/b/c
104 * ==> invalidate self
105 * ==> find "a", have it invalidate "b/c"
107 * ==> invalidate self
108 * ==> if "a" exists as a subtree, remove it.
110 const char *slash;
111 int namelen;
112 struct cache_tree_sub *down;
114 #if DEBUG
115 fprintf(stderr, "cache-tree invalidate <%s>\n", path);
116 #endif
118 if (!it)
119 return 0;
120 slash = strchrnul(path, '/');
121 namelen = slash - path;
122 it->entry_count = -1;
123 if (!*slash) {
124 int pos;
125 pos = subtree_pos(it, path, namelen);
126 if (0 <= pos) {
127 cache_tree_free(&it->down[pos]->cache_tree);
128 free(it->down[pos]);
129 /* 0 1 2 3 4 5
130 * ^ ^subtree_nr = 6
131 * pos
132 * move 4 and 5 up one place (2 entries)
133 * 2 = 6 - 3 - 1 = subtree_nr - pos - 1
135 memmove(it->down+pos, it->down+pos+1,
136 sizeof(struct cache_tree_sub *) *
137 (it->subtree_nr - pos - 1));
138 it->subtree_nr--;
140 return 1;
142 down = find_subtree(it, path, namelen, 0);
143 if (down)
144 do_invalidate_path(down->cache_tree, slash + 1);
145 return 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)
157 int i, funny;
158 int silent = flags & WRITE_TREE_SILENT;
160 /* Verify that the tree is merged */
161 funny = 0;
162 for (i = 0; i < entries; i++) {
163 const struct cache_entry *ce = cache[i];
164 if (ce_stage(ce)) {
165 if (silent)
166 return -1;
167 if (10 < ++funny) {
168 fprintf(stderr, "...\n");
169 break;
171 fprintf(stderr, "%s: unmerged (%s)\n",
172 ce->name, sha1_to_hex(ce->sha1));
175 if (funny)
176 return -1;
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
180 * stage 0 entries.
182 funny = 0;
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] == '/') {
194 if (10 < ++funny) {
195 fprintf(stderr, "...\n");
196 break;
198 fprintf(stderr, "You have both %s and %s\n",
199 this_name, next_name);
202 if (funny)
203 return -1;
204 return 0;
207 static void discard_unused_subtrees(struct cache_tree *it)
209 struct cache_tree_sub **down = it->down;
210 int nr = it->subtree_nr;
211 int dst, src;
212 for (dst = src = 0; src < nr; src++) {
213 struct cache_tree_sub *s = down[src];
214 if (s->used)
215 down[dst++] = s;
216 else {
217 cache_tree_free(&s->cache_tree);
218 free(s);
219 it->subtree_nr--;
224 int cache_tree_fully_valid(struct cache_tree *it)
226 int i;
227 if (!it)
228 return 0;
229 if (it->entry_count < 0 || !has_sha1_file(it->sha1))
230 return 0;
231 for (i = 0; i < it->subtree_nr; i++) {
232 if (!cache_tree_fully_valid(it->down[i]->cache_tree))
233 return 0;
235 return 1;
238 static int update_one(struct cache_tree *it,
239 struct cache_entry **cache,
240 int entries,
241 const char *base,
242 int baselen,
243 int *skip_count,
244 int flags)
246 struct strbuf buffer;
247 int missing_ok = flags & WRITE_TREE_MISSING_OK;
248 int dryrun = flags & WRITE_TREE_DRY_RUN;
249 int to_invalidate = 0;
250 int i;
252 *skip_count = 0;
254 if (0 <= it->entry_count && has_sha1_file(it->sha1))
255 return it->entry_count;
258 * We first scan for subtrees and update them; we start by
259 * marking existing subtrees -- the ones that are unmarked
260 * should not be in the result.
262 for (i = 0; i < it->subtree_nr; i++)
263 it->down[i]->used = 0;
266 * Find the subtrees and update them.
268 i = 0;
269 while (i < entries) {
270 const struct cache_entry *ce = cache[i];
271 struct cache_tree_sub *sub;
272 const char *path, *slash;
273 int pathlen, sublen, subcnt, subskip;
275 path = ce->name;
276 pathlen = ce_namelen(ce);
277 if (pathlen <= baselen || memcmp(base, path, baselen))
278 break; /* at the end of this level */
280 slash = strchr(path + baselen, '/');
281 if (!slash) {
282 i++;
283 continue;
286 * a/bbb/c (base = a/, slash = /c)
287 * ==>
288 * path+baselen = bbb/c, sublen = 3
290 sublen = slash - (path + baselen);
291 sub = find_subtree(it, path + baselen, sublen, 1);
292 if (!sub->cache_tree)
293 sub->cache_tree = cache_tree();
294 subcnt = update_one(sub->cache_tree,
295 cache + i, entries - i,
296 path,
297 baselen + sublen + 1,
298 &subskip,
299 flags);
300 if (subcnt < 0)
301 return subcnt;
302 i += subcnt;
303 sub->count = subcnt; /* to be used in the next loop */
304 *skip_count += subskip;
305 sub->used = 1;
308 discard_unused_subtrees(it);
311 * Then write out the tree object for this level.
313 strbuf_init(&buffer, 8192);
315 i = 0;
316 while (i < entries) {
317 const struct cache_entry *ce = cache[i];
318 struct cache_tree_sub *sub;
319 const char *path, *slash;
320 int pathlen, entlen;
321 const unsigned char *sha1;
322 unsigned mode;
324 path = ce->name;
325 pathlen = ce_namelen(ce);
326 if (pathlen <= baselen || memcmp(base, path, baselen))
327 break; /* at the end of this level */
329 slash = strchr(path + baselen, '/');
330 if (slash) {
331 entlen = slash - (path + baselen);
332 sub = find_subtree(it, path + baselen, entlen, 0);
333 if (!sub)
334 die("cache-tree.c: '%.*s' in '%s' not found",
335 entlen, path + baselen, path);
336 i += sub->count;
337 sha1 = sub->cache_tree->sha1;
338 mode = S_IFDIR;
339 if (sub->cache_tree->entry_count < 0)
340 to_invalidate = 1;
342 else {
343 sha1 = ce->sha1;
344 mode = ce->ce_mode;
345 entlen = pathlen - baselen;
346 i++;
348 if (mode != S_IFGITLINK && !missing_ok && !has_sha1_file(sha1)) {
349 strbuf_release(&buffer);
350 return error("invalid object %06o %s for '%.*s'",
351 mode, sha1_to_hex(sha1), entlen+baselen, path);
355 * CE_REMOVE entries are removed before the index is
356 * written to disk. Skip them to remain consistent
357 * with the future on-disk index.
359 if (ce->ce_flags & CE_REMOVE) {
360 *skip_count = *skip_count + 1;
361 continue;
365 * CE_INTENT_TO_ADD entries exist on on-disk index but
366 * they are not part of generated trees. Invalidate up
367 * to root to force cache-tree users to read elsewhere.
369 if (ce->ce_flags & CE_INTENT_TO_ADD) {
370 to_invalidate = 1;
371 continue;
374 strbuf_grow(&buffer, entlen + 100);
375 strbuf_addf(&buffer, "%o %.*s%c", mode, entlen, path + baselen, '\0');
376 strbuf_add(&buffer, sha1, 20);
378 #if DEBUG
379 fprintf(stderr, "cache-tree update-one %o %.*s\n",
380 mode, entlen, path + baselen);
381 #endif
384 if (dryrun)
385 hash_sha1_file(buffer.buf, buffer.len, tree_type, it->sha1);
386 else if (write_sha1_file(buffer.buf, buffer.len, tree_type, it->sha1)) {
387 strbuf_release(&buffer);
388 return -1;
391 strbuf_release(&buffer);
392 it->entry_count = to_invalidate ? -1 : i - *skip_count;
393 #if DEBUG
394 fprintf(stderr, "cache-tree update-one (%d ent, %d subtree) %s\n",
395 it->entry_count, it->subtree_nr,
396 sha1_to_hex(it->sha1));
397 #endif
398 return i;
401 int cache_tree_update(struct index_state *istate, int flags)
403 struct cache_tree *it = istate->cache_tree;
404 struct cache_entry **cache = istate->cache;
405 int entries = istate->cache_nr;
406 int skip, i = verify_cache(cache, entries, flags);
408 if (i)
409 return i;
410 i = update_one(it, cache, entries, "", 0, &skip, flags);
411 if (i < 0)
412 return i;
413 istate->cache_changed |= CACHE_TREE_CHANGED;
414 return 0;
417 static void write_one(struct strbuf *buffer, struct cache_tree *it,
418 const char *path, int pathlen)
420 int i;
422 /* One "cache-tree" entry consists of the following:
423 * path (NUL terminated)
424 * entry_count, subtree_nr ("%d %d\n")
425 * tree-sha1 (missing if invalid)
426 * subtree_nr "cache-tree" entries for subtrees.
428 strbuf_grow(buffer, pathlen + 100);
429 strbuf_add(buffer, path, pathlen);
430 strbuf_addf(buffer, "%c%d %d\n", 0, it->entry_count, it->subtree_nr);
432 #if DEBUG
433 if (0 <= it->entry_count)
434 fprintf(stderr, "cache-tree <%.*s> (%d ent, %d subtree) %s\n",
435 pathlen, path, it->entry_count, it->subtree_nr,
436 sha1_to_hex(it->sha1));
437 else
438 fprintf(stderr, "cache-tree <%.*s> (%d subtree) invalid\n",
439 pathlen, path, it->subtree_nr);
440 #endif
442 if (0 <= it->entry_count) {
443 strbuf_add(buffer, it->sha1, 20);
445 for (i = 0; i < it->subtree_nr; i++) {
446 struct cache_tree_sub *down = it->down[i];
447 if (i) {
448 struct cache_tree_sub *prev = it->down[i-1];
449 if (subtree_name_cmp(down->name, down->namelen,
450 prev->name, prev->namelen) <= 0)
451 die("fatal - unsorted cache subtree");
453 write_one(buffer, down->cache_tree, down->name, down->namelen);
457 void cache_tree_write(struct strbuf *sb, struct cache_tree *root)
459 write_one(sb, root, "", 0);
462 static struct cache_tree *read_one(const char **buffer, unsigned long *size_p)
464 const char *buf = *buffer;
465 unsigned long size = *size_p;
466 const char *cp;
467 char *ep;
468 struct cache_tree *it;
469 int i, subtree_nr;
471 it = NULL;
472 /* skip name, but make sure name exists */
473 while (size && *buf) {
474 size--;
475 buf++;
477 if (!size)
478 goto free_return;
479 buf++; size--;
480 it = cache_tree();
482 cp = buf;
483 it->entry_count = strtol(cp, &ep, 10);
484 if (cp == ep)
485 goto free_return;
486 cp = ep;
487 subtree_nr = strtol(cp, &ep, 10);
488 if (cp == ep)
489 goto free_return;
490 while (size && *buf && *buf != '\n') {
491 size--;
492 buf++;
494 if (!size)
495 goto free_return;
496 buf++; size--;
497 if (0 <= it->entry_count) {
498 if (size < 20)
499 goto free_return;
500 hashcpy(it->sha1, (const unsigned char*)buf);
501 buf += 20;
502 size -= 20;
505 #if DEBUG
506 if (0 <= it->entry_count)
507 fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n",
508 *buffer, it->entry_count, subtree_nr,
509 sha1_to_hex(it->sha1));
510 else
511 fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n",
512 *buffer, subtree_nr);
513 #endif
516 * Just a heuristic -- we do not add directories that often but
517 * we do not want to have to extend it immediately when we do,
518 * hence +2.
520 it->subtree_alloc = subtree_nr + 2;
521 it->down = xcalloc(it->subtree_alloc, sizeof(struct cache_tree_sub *));
522 for (i = 0; i < subtree_nr; i++) {
523 /* read each subtree */
524 struct cache_tree *sub;
525 struct cache_tree_sub *subtree;
526 const char *name = buf;
528 sub = read_one(&buf, &size);
529 if (!sub)
530 goto free_return;
531 subtree = cache_tree_sub(it, name);
532 subtree->cache_tree = sub;
534 if (subtree_nr != it->subtree_nr)
535 die("cache-tree: internal error");
536 *buffer = buf;
537 *size_p = size;
538 return it;
540 free_return:
541 cache_tree_free(&it);
542 return NULL;
545 struct cache_tree *cache_tree_read(const char *buffer, unsigned long size)
547 if (buffer[0])
548 return NULL; /* not the whole tree */
549 return read_one(&buffer, &size);
552 static struct cache_tree *cache_tree_find(struct cache_tree *it, const char *path)
554 if (!it)
555 return NULL;
556 while (*path) {
557 const char *slash;
558 struct cache_tree_sub *sub;
560 slash = strchrnul(path, '/');
562 * Between path and slash is the name of the subtree
563 * to look for.
565 sub = find_subtree(it, path, slash - path, 0);
566 if (!sub)
567 return NULL;
568 it = sub->cache_tree;
570 path = slash;
571 while (*path == '/')
572 path++;
574 return it;
577 int write_cache_as_tree(unsigned char *sha1, int flags, const char *prefix)
579 int entries, was_valid, newfd;
580 struct lock_file *lock_file;
583 * We can't free this memory, it becomes part of a linked list
584 * parsed atexit()
586 lock_file = xcalloc(1, sizeof(struct lock_file));
588 newfd = hold_locked_index(lock_file, 1);
590 entries = read_cache();
591 if (entries < 0)
592 return WRITE_TREE_UNREADABLE_INDEX;
593 if (flags & WRITE_TREE_IGNORE_CACHE_TREE)
594 cache_tree_free(&(active_cache_tree));
596 if (!active_cache_tree)
597 active_cache_tree = cache_tree();
599 was_valid = cache_tree_fully_valid(active_cache_tree);
600 if (!was_valid) {
601 if (cache_tree_update(&the_index, flags) < 0)
602 return WRITE_TREE_UNMERGED_INDEX;
603 if (0 <= newfd) {
604 if (!write_locked_index(&the_index, lock_file, COMMIT_LOCK))
605 newfd = -1;
607 /* Not being able to write is fine -- we are only interested
608 * in updating the cache-tree part, and if the next caller
609 * ends up using the old index with unupdated cache-tree part
610 * it misses the work we did here, but that is just a
611 * performance penalty and not a big deal.
615 if (prefix) {
616 struct cache_tree *subtree =
617 cache_tree_find(active_cache_tree, prefix);
618 if (!subtree)
619 return WRITE_TREE_PREFIX_ERROR;
620 hashcpy(sha1, subtree->sha1);
622 else
623 hashcpy(sha1, active_cache_tree->sha1);
625 if (0 <= newfd)
626 rollback_lock_file(lock_file);
628 return 0;
631 static void prime_cache_tree_rec(struct cache_tree *it, struct tree *tree)
633 struct tree_desc desc;
634 struct name_entry entry;
635 int cnt;
637 hashcpy(it->sha1, tree->object.sha1);
638 init_tree_desc(&desc, tree->buffer, tree->size);
639 cnt = 0;
640 while (tree_entry(&desc, &entry)) {
641 if (!S_ISDIR(entry.mode))
642 cnt++;
643 else {
644 struct cache_tree_sub *sub;
645 struct tree *subtree = lookup_tree(entry.sha1);
646 if (!subtree->object.parsed)
647 parse_tree(subtree);
648 sub = cache_tree_sub(it, entry.path);
649 sub->cache_tree = cache_tree();
650 prime_cache_tree_rec(sub->cache_tree, subtree);
651 cnt += sub->cache_tree->entry_count;
654 it->entry_count = cnt;
657 void prime_cache_tree(struct index_state *istate, struct tree *tree)
659 cache_tree_free(&istate->cache_tree);
660 istate->cache_tree = cache_tree();
661 prime_cache_tree_rec(istate->cache_tree, tree);
662 istate->cache_changed |= CACHE_TREE_CHANGED;
666 * find the cache_tree that corresponds to the current level without
667 * exploding the full path into textual form. The root of the
668 * cache tree is given as "root", and our current level is "info".
669 * (1) When at root level, info->prev is NULL, so it is "root" itself.
670 * (2) Otherwise, find the cache_tree that corresponds to one level
671 * above us, and find ourselves in there.
673 static struct cache_tree *find_cache_tree_from_traversal(struct cache_tree *root,
674 struct traverse_info *info)
676 struct cache_tree *our_parent;
678 if (!info->prev)
679 return root;
680 our_parent = find_cache_tree_from_traversal(root, info->prev);
681 return cache_tree_find(our_parent, info->name.path);
684 int cache_tree_matches_traversal(struct cache_tree *root,
685 struct name_entry *ent,
686 struct traverse_info *info)
688 struct cache_tree *it;
690 it = find_cache_tree_from_traversal(root, info);
691 it = cache_tree_find(it, ent->path);
692 if (it && it->entry_count > 0 && !hashcmp(ent->sha1, it->sha1))
693 return it->entry_count;
694 return 0;
697 int update_main_cache_tree(int flags)
699 if (!the_index.cache_tree)
700 the_index.cache_tree = cache_tree();
701 return cache_tree_update(&the_index, flags);