On lör, 2011-02-26 at 09:43 +0100, Jan Urbański wrote:
> I'm officially at a loss on how to fix that bug without some serious
> gutting of how PL/Python arguments work. If someone comes up with a
> brilliant way to solve this problem, we can commit it after beta, or
> even during the 9.2 cycle (should the brilliant solution be
> backpatcheable).
We'd essentially be trading off freeing something too soon with freeing
it not at all. I'm not sure how good that tradeoff is.