Thread: Yet another regression issue with Fedora-10 + PG 8.4 beta1

Hi,

I built PG 8.4 beta1 on 2 different Fedora-10 boxes. One of them was a
local box, which runs under QEMU. The other one is Fedora's build
servers. (I did a scratch build on Fedora build server)

On the first machine, I got a regression failure:

http://www.gunduz.org/temp/regression.out
http://www.gunduz.org/temp/regression.diffs

I could not repeat this with 2nd one (Fedora build servers)

Any ideas what might have caused it?

Regards,

--
Devrim GÜNDÜZ, RHCE
devrim~gunduz.org, devrim~PostgreSQL.org, devrim.gunduz~linux.org.tr                  http://www.gunduz.org

Devrim GÜNDÜZ <devrim@gunduz.org> writes:
> I built PG 8.4 beta1 on 2 different Fedora-10 boxes. One of them was a
> local box, which runs under QEMU. The other one is Fedora's build
> servers. (I did a scratch build on Fedora build server)

> On the first machine, I got a regression failure:

> http://www.gunduz.org/temp/regression.out
> http://www.gunduz.org/temp/regression.diffs

> I could not repeat this with 2nd one (Fedora build servers)

> Any ideas what might have caused it?

QEMU screwing up the handling of floating-point traps, perhaps?
        regards, tom lane

On Tue, 2009-04-14 at 18:32 -0400, Tom Lane wrote:
> Any ideas what might have caused it?
>
> QEMU screwing up the handling of floating-point traps, perhaps?

Probably. Let me dig it.

Thanks.
--
Devrim GÜNDÜZ, RHCE
devrim~gunduz.org, devrim~PostgreSQL.org, devrim.gunduz~linux.org.tr                  http://www.gunduz.org