Alfred Perlstein <bright(at)wintelcom(dot)net> writes:
> I have a reliable way to make postgresql crash after a
> couple of hours over here and a backtrace that looks like a good
> catch.
I'm interested in pursuing this, but the backtrace doesn't give enough
info to debug it. It looks like the backend is crashing because of
a previously-corrupted tuple, so what we'll need to do is work backwards
to find where the data corruption is occurring.
Can you boil down the test sequence to something that could be
reproduced by other people? The most convenient way to work on it
would be to see it happen here...
regards, tom lane