On Thu, Jan 25, 2007 at 04:29:58PM -0500, Tom Lane wrote:
> Alvaro Herrera <alvherre@commandprompt.com> writes:
> > Jeremy Haile wrote:
> >> If anyone else is experiencing similar problems, please post your
> >> situation.
>
> > All the Windows buildfarm machines are, apparently.
>
> Can't anyone with a debugger duplicate this and get a stack trace for
> us? If the stats collector is indeed freezing up, a stack trace showing
> where it's stuck would be exceedingly helpful.
Maybe I should finish testing before I send my emails.
Apparantly there is a bug lurking somewhere in pgwin32_select(). Because
if I put a #undef select right before the select in pgstat.c, the
regression tests pass.
I guess the bug is shown because with row level stats we simply have
more data to process. And it appears only to happen on UDP sockets from
what I can tell.
Now, what Iwould *like* to do is to re-implement that part of the code
using the Win32 APIs instead of going through select(). Since it's very
isolated code. I'm going to try that and see how invasive it is, then
see if it'll get accepted :-)
(This would of course give us better performance in general in that
codepath, since all the emulation stuff wouldn't be needed, so there's a
point to doing that other than finding the obscure UDP-related bug in
pgwin23_select)
//Magnus