Sync with maint
[git/mjg.git] / cache-tree.c
blob75a54fdc7232e5693b042aa06540d51a99f01551
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 repair = flags & WRITE_TREE_REPAIR;
250 int to_invalidate = 0;
251 int i;
253 assert(!(dryrun && repair));
255 *skip_count = 0;
257 if (0 <= it->entry_count && has_sha1_file(it->sha1))
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.
271 i = 0;
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;
278 path = ce->name;
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, '/');
284 if (!slash) {
285 i++;
286 continue;
289 * a/bbb/c (base = a/, slash = /c)
290 * ==>
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,
299 path,
300 baselen + sublen + 1,
301 &subskip,
302 flags);
303 if (subcnt < 0)
304 return subcnt;
305 i += subcnt;
306 sub->count = subcnt; /* to be used in the next loop */
307 *skip_count += subskip;
308 sub->used = 1;
311 discard_unused_subtrees(it);
314 * Then write out the tree object for this level.
316 strbuf_init(&buffer, 8192);
318 i = 0;
319 while (i < entries) {
320 const struct cache_entry *ce = cache[i];
321 struct cache_tree_sub *sub;
322 const char *path, *slash;
323 int pathlen, entlen;
324 const unsigned char *sha1;
325 unsigned mode;
326 int expected_missing = 0;
328 path = ce->name;
329 pathlen = ce_namelen(ce);
330 if (pathlen <= baselen || memcmp(base, path, baselen))
331 break; /* at the end of this level */
333 slash = strchr(path + baselen, '/');
334 if (slash) {
335 entlen = slash - (path + baselen);
336 sub = find_subtree(it, path + baselen, entlen, 0);
337 if (!sub)
338 die("cache-tree.c: '%.*s' in '%s' not found",
339 entlen, path + baselen, path);
340 i += sub->count;
341 sha1 = sub->cache_tree->sha1;
342 mode = S_IFDIR;
343 if (sub->cache_tree->entry_count < 0) {
344 to_invalidate = 1;
345 expected_missing = 1;
348 else {
349 sha1 = ce->sha1;
350 mode = ce->ce_mode;
351 entlen = pathlen - baselen;
352 i++;
354 if (mode != S_IFGITLINK && !missing_ok && !has_sha1_file(sha1)) {
355 strbuf_release(&buffer);
356 if (expected_missing)
357 return -1;
358 return error("invalid object %06o %s for '%.*s'",
359 mode, sha1_to_hex(sha1), entlen+baselen, path);
363 * CE_REMOVE entries are removed before the index is
364 * written to disk. Skip them to remain consistent
365 * with the future on-disk index.
367 if (ce->ce_flags & CE_REMOVE) {
368 *skip_count = *skip_count + 1;
369 continue;
373 * CE_INTENT_TO_ADD entries exist on on-disk index but
374 * they are not part of generated trees. Invalidate up
375 * to root to force cache-tree users to read elsewhere.
377 if (ce->ce_flags & CE_INTENT_TO_ADD) {
378 to_invalidate = 1;
379 continue;
382 strbuf_grow(&buffer, entlen + 100);
383 strbuf_addf(&buffer, "%o %.*s%c", mode, entlen, path + baselen, '\0');
384 strbuf_add(&buffer, sha1, 20);
386 #if DEBUG
387 fprintf(stderr, "cache-tree update-one %o %.*s\n",
388 mode, entlen, path + baselen);
389 #endif
392 if (repair) {
393 unsigned char sha1[20];
394 hash_sha1_file(buffer.buf, buffer.len, tree_type, sha1);
395 if (has_sha1_file(sha1))
396 hashcpy(it->sha1, sha1);
397 else
398 to_invalidate = 1;
399 } else if (dryrun)
400 hash_sha1_file(buffer.buf, buffer.len, tree_type, it->sha1);
401 else if (write_sha1_file(buffer.buf, buffer.len, tree_type, it->sha1)) {
402 strbuf_release(&buffer);
403 return -1;
406 strbuf_release(&buffer);
407 it->entry_count = to_invalidate ? -1 : i - *skip_count;
408 #if DEBUG
409 fprintf(stderr, "cache-tree update-one (%d ent, %d subtree) %s\n",
410 it->entry_count, it->subtree_nr,
411 sha1_to_hex(it->sha1));
412 #endif
413 return i;
416 int cache_tree_update(struct index_state *istate, int flags)
418 struct cache_tree *it = istate->cache_tree;
419 struct cache_entry **cache = istate->cache;
420 int entries = istate->cache_nr;
421 int skip, i = verify_cache(cache, entries, flags);
423 if (i)
424 return i;
425 i = update_one(it, cache, entries, "", 0, &skip, flags);
426 if (i < 0)
427 return i;
428 istate->cache_changed |= CACHE_TREE_CHANGED;
429 return 0;
432 static void write_one(struct strbuf *buffer, struct cache_tree *it,
433 const char *path, int pathlen)
435 int i;
437 /* One "cache-tree" entry consists of the following:
438 * path (NUL terminated)
439 * entry_count, subtree_nr ("%d %d\n")
440 * tree-sha1 (missing if invalid)
441 * subtree_nr "cache-tree" entries for subtrees.
443 strbuf_grow(buffer, pathlen + 100);
444 strbuf_add(buffer, path, pathlen);
445 strbuf_addf(buffer, "%c%d %d\n", 0, it->entry_count, it->subtree_nr);
447 #if DEBUG
448 if (0 <= it->entry_count)
449 fprintf(stderr, "cache-tree <%.*s> (%d ent, %d subtree) %s\n",
450 pathlen, path, it->entry_count, it->subtree_nr,
451 sha1_to_hex(it->sha1));
452 else
453 fprintf(stderr, "cache-tree <%.*s> (%d subtree) invalid\n",
454 pathlen, path, it->subtree_nr);
455 #endif
457 if (0 <= it->entry_count) {
458 strbuf_add(buffer, it->sha1, 20);
460 for (i = 0; i < it->subtree_nr; i++) {
461 struct cache_tree_sub *down = it->down[i];
462 if (i) {
463 struct cache_tree_sub *prev = it->down[i-1];
464 if (subtree_name_cmp(down->name, down->namelen,
465 prev->name, prev->namelen) <= 0)
466 die("fatal - unsorted cache subtree");
468 write_one(buffer, down->cache_tree, down->name, down->namelen);
472 void cache_tree_write(struct strbuf *sb, struct cache_tree *root)
474 write_one(sb, root, "", 0);
477 static struct cache_tree *read_one(const char **buffer, unsigned long *size_p)
479 const char *buf = *buffer;
480 unsigned long size = *size_p;
481 const char *cp;
482 char *ep;
483 struct cache_tree *it;
484 int i, subtree_nr;
486 it = NULL;
487 /* skip name, but make sure name exists */
488 while (size && *buf) {
489 size--;
490 buf++;
492 if (!size)
493 goto free_return;
494 buf++; size--;
495 it = cache_tree();
497 cp = buf;
498 it->entry_count = strtol(cp, &ep, 10);
499 if (cp == ep)
500 goto free_return;
501 cp = ep;
502 subtree_nr = strtol(cp, &ep, 10);
503 if (cp == ep)
504 goto free_return;
505 while (size && *buf && *buf != '\n') {
506 size--;
507 buf++;
509 if (!size)
510 goto free_return;
511 buf++; size--;
512 if (0 <= it->entry_count) {
513 if (size < 20)
514 goto free_return;
515 hashcpy(it->sha1, (const unsigned char*)buf);
516 buf += 20;
517 size -= 20;
520 #if DEBUG
521 if (0 <= it->entry_count)
522 fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n",
523 *buffer, it->entry_count, subtree_nr,
524 sha1_to_hex(it->sha1));
525 else
526 fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n",
527 *buffer, subtree_nr);
528 #endif
531 * Just a heuristic -- we do not add directories that often but
532 * we do not want to have to extend it immediately when we do,
533 * hence +2.
535 it->subtree_alloc = subtree_nr + 2;
536 it->down = xcalloc(it->subtree_alloc, sizeof(struct cache_tree_sub *));
537 for (i = 0; i < subtree_nr; i++) {
538 /* read each subtree */
539 struct cache_tree *sub;
540 struct cache_tree_sub *subtree;
541 const char *name = buf;
543 sub = read_one(&buf, &size);
544 if (!sub)
545 goto free_return;
546 subtree = cache_tree_sub(it, name);
547 subtree->cache_tree = sub;
549 if (subtree_nr != it->subtree_nr)
550 die("cache-tree: internal error");
551 *buffer = buf;
552 *size_p = size;
553 return it;
555 free_return:
556 cache_tree_free(&it);
557 return NULL;
560 struct cache_tree *cache_tree_read(const char *buffer, unsigned long size)
562 if (buffer[0])
563 return NULL; /* not the whole tree */
564 return read_one(&buffer, &size);
567 static struct cache_tree *cache_tree_find(struct cache_tree *it, const char *path)
569 if (!it)
570 return NULL;
571 while (*path) {
572 const char *slash;
573 struct cache_tree_sub *sub;
575 slash = strchrnul(path, '/');
577 * Between path and slash is the name of the subtree
578 * to look for.
580 sub = find_subtree(it, path, slash - path, 0);
581 if (!sub)
582 return NULL;
583 it = sub->cache_tree;
585 path = slash;
586 while (*path == '/')
587 path++;
589 return it;
592 int write_cache_as_tree(unsigned char *sha1, int flags, const char *prefix)
594 int entries, was_valid, newfd;
595 struct lock_file *lock_file;
598 * We can't free this memory, it becomes part of a linked list
599 * parsed atexit()
601 lock_file = xcalloc(1, sizeof(struct lock_file));
603 newfd = hold_locked_index(lock_file, 1);
605 entries = read_cache();
606 if (entries < 0)
607 return WRITE_TREE_UNREADABLE_INDEX;
608 if (flags & WRITE_TREE_IGNORE_CACHE_TREE)
609 cache_tree_free(&(active_cache_tree));
611 if (!active_cache_tree)
612 active_cache_tree = cache_tree();
614 was_valid = cache_tree_fully_valid(active_cache_tree);
615 if (!was_valid) {
616 if (cache_tree_update(&the_index, flags) < 0)
617 return WRITE_TREE_UNMERGED_INDEX;
618 if (0 <= newfd) {
619 if (!write_locked_index(&the_index, lock_file, COMMIT_LOCK))
620 newfd = -1;
622 /* Not being able to write is fine -- we are only interested
623 * in updating the cache-tree part, and if the next caller
624 * ends up using the old index with unupdated cache-tree part
625 * it misses the work we did here, but that is just a
626 * performance penalty and not a big deal.
630 if (prefix) {
631 struct cache_tree *subtree =
632 cache_tree_find(active_cache_tree, prefix);
633 if (!subtree)
634 return WRITE_TREE_PREFIX_ERROR;
635 hashcpy(sha1, subtree->sha1);
637 else
638 hashcpy(sha1, active_cache_tree->sha1);
640 if (0 <= newfd)
641 rollback_lock_file(lock_file);
643 return 0;
646 static void prime_cache_tree_rec(struct cache_tree *it, struct tree *tree)
648 struct tree_desc desc;
649 struct name_entry entry;
650 int cnt;
652 hashcpy(it->sha1, tree->object.sha1);
653 init_tree_desc(&desc, tree->buffer, tree->size);
654 cnt = 0;
655 while (tree_entry(&desc, &entry)) {
656 if (!S_ISDIR(entry.mode))
657 cnt++;
658 else {
659 struct cache_tree_sub *sub;
660 struct tree *subtree = lookup_tree(entry.sha1);
661 if (!subtree->object.parsed)
662 parse_tree(subtree);
663 sub = cache_tree_sub(it, entry.path);
664 sub->cache_tree = cache_tree();
665 prime_cache_tree_rec(sub->cache_tree, subtree);
666 cnt += sub->cache_tree->entry_count;
669 it->entry_count = cnt;
672 void prime_cache_tree(struct index_state *istate, struct tree *tree)
674 cache_tree_free(&istate->cache_tree);
675 istate->cache_tree = cache_tree();
676 prime_cache_tree_rec(istate->cache_tree, tree);
677 istate->cache_changed |= CACHE_TREE_CHANGED;
681 * find the cache_tree that corresponds to the current level without
682 * exploding the full path into textual form. The root of the
683 * cache tree is given as "root", and our current level is "info".
684 * (1) When at root level, info->prev is NULL, so it is "root" itself.
685 * (2) Otherwise, find the cache_tree that corresponds to one level
686 * above us, and find ourselves in there.
688 static struct cache_tree *find_cache_tree_from_traversal(struct cache_tree *root,
689 struct traverse_info *info)
691 struct cache_tree *our_parent;
693 if (!info->prev)
694 return root;
695 our_parent = find_cache_tree_from_traversal(root, info->prev);
696 return cache_tree_find(our_parent, info->name.path);
699 int cache_tree_matches_traversal(struct cache_tree *root,
700 struct name_entry *ent,
701 struct traverse_info *info)
703 struct cache_tree *it;
705 it = find_cache_tree_from_traversal(root, info);
706 it = cache_tree_find(it, ent->path);
707 if (it && it->entry_count > 0 && !hashcmp(ent->sha1, it->sha1))
708 return it->entry_count;
709 return 0;
712 int update_main_cache_tree(int flags)
714 if (!the_index.cache_tree)
715 the_index.cache_tree = cache_tree();
716 return cache_tree_update(&the_index, flags);