On Mon, Mar 24, 2008 at 9:24 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> "Jeffrey Baker" <jwbaker@gmail.com> writes:
> > On Sun, Mar 23, 2008 at 8:35 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> >> If it works over TCP and not over Unix socket, it's a kernel bug.
> >> The libpq code doesn't really know the difference after connection
> >> setup.
>
> > The same thought occurred to me, but it could also be a race condition
> > which the unix socket is fast enough to trigger but the TCP socket is
> > not fast enough to trigger. I'm peeking around in the code but
> > nothing jumps out yet.
>
> Fairly hard to believe given that you're talking about communication
> between two sequential processes. Anyway I'd suggest that the first
> thing to do is extract a reproducible test case. It'd be useful
> to see if it hangs on other platforms...
The stack trace doesn't actually make sense, does it? I think that
(at least) the PQmblen frame is spurious.
-jwb