Jim Blandy <jimb@zwingli.cygnus.com> writes:
> Great! I'll check it over tomorrow...
One more patch. This fixes the same bug GregS pointed out
to similar code.
> > I have one question with regard to svn_fs__dag_close which
> > was just deleted. It's still referenced by tree.c and I'd
> > like to resurrect it. It will call just apr_pool_destroy
> > (dag->pool) if I understand it correctly. The problem of
> > current implementation is that dag->pool points to
> > trail->pool. Shouldn't dag->pool be subpool of trail->pool?
>
> Your understanding is correct. We decided that it wasn't important
> enough to free dag nodes before their underlying pools were freed.
> What would you like to do with svn_fs__dag_close, other than free
> memory?
Hmm. I'd say none. It seems somehow I thought dag_node_t
can be big. But it only mainains reference to node-revision
skel...
--- dag.c.1 Thu Feb 22 13:15:11 2001
+++ dag.c Thu Feb 22 13:22:27 2001
@@ -313,7 +313,8 @@
if (entry->is_atom || ! entry->children->is_atom)
abort ();
- if (! memcmp (entry->children->data, name, entry->children->len))
+ if (entry->children->len == strlen (name)
+ && ! memcmp (entry->children->data, name, entry->children->len))
{
skel_t *id = entry->children->next;
child->id = svn_fs_parse_id (id->data, id->len, trail->pool);
--
Yoshiki Hayashi
Received on Sat Oct 21 14:36:23 2006