On Tue, Jun 24, 2008 at 05:34:50PM -0400, Tom Lane wrote:
> daveg <daveg@sonic.net> writes:
> > lock-timeout sets statement_timeout to a small value while locks are being
> > taken on all the tables. Then it resets it to default. So it could reset it
> > to whatever the new default is.
>
> "reset to default" is *surely* not the right behavior; resetting to the
> setting that had been in effect might be a bit sane. But the whole
> design sounds pretty broken to start with. The timer management code
> already understands the concept of scheduling the interrupt for the
> nearest of a couple of potentially active timeouts. ISTM any patch
> intended to add a feature like this ought to extend that logic rather
> than hack around changing the values of global variables.
Are we talking about the same patch? Because I don't know what you are
refering to with "timer management code" and "scheduling the interrupt" in
the context of pg_dump.
-dg
--
David Gould daveg@sonic.net 510 536 1443 510 282 0869
If simplicity worked, the world would be overrun with insects.