From: "Tom Lane" <tgl@sss.pgh.pa.us>
> In the first query, the MemoryContextReset is nearly free since there's
> nothing to free (and we've special-cased that path in aset.c). It's
> certainly a measurement artifact that it measures out faster, which says
> to me that these numbers can only be trusted within a couple percent;
> but at least we're not taking much hit in cases where the patch isn't
> actually conferring any benefit. For the second query, losing 1% to avoid
> memory bloat seems well worthwhile.
>
> Barring objections I'll apply and back-patch this.
So this patch would solve memory leak issues if other modules had similar
bugs, in addition to the original dblink problem, wouldn't this? Definitely
+1. The original OP wants to use 9.2. I'll report to him when you've
committed this nce patch. Thanks, Tom san.
Regards
MauMau