On 10/03/2012 09:23 PM, Tom Lane wrote:
> A bigger problem with this is that it only fixes the issue for cases in
> which somebody makes new threads of control with fork(). I believe that
> issues involving multiple threads trying to use the same PGconn are at
> least as widespread. I'm not terribly excited about removing
> functionality and adding overhead to protect against just one variant of
> the problem.
>
>
I had the same thought re threads.
cheers
andrew