Kris Jurka wrote:
> So the fix is probably to check that we aren't setting type
> information from stale results, but this is a pretty complicated area,
> so I'd like to study it a little more.
>
> In any case, thanks for your help in tracking this one down.
No problem, thanks to you for your responsiveness and for your
commitment to this project. Let me know if there's anything more I can
do to assist; testing a patch, more logging, etc.
Regards,
Eric Faulhaber