Re: -DCLOBBER_CACHE_ALWAYS shows COPY FREEZE regression problem

On 2012-12-06 23:59:06 +0000, Simon Riggs wrote:
> On 6 December 2012 23:28, Andrew Dunstan <andrew(at)dunslane(dot)net> wrote:
> > On a new buildfarm member friarbird
> > <http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=friarbird&dt=2012-12-06%2020%3A55%3A31>,
> > configured with _DCLOBBER_CACHE_ALWAYS:
> >
> > BEGIN;
> > TRUNCATE vistest;
> > COPY vistest FROM stdin CSV FREEZE;
> > + NOTICE: FREEZE option specified but pre-conditions not met
>
> I don't understand why that build option would produce different
> output for simple logic.
>
> I'll look again in the morning.

It clears the relcache entry as soon as AcceptInvalidationMessages() is
done seems to loose rd_createSubid and rd_newRelfilenodeSubid.

Apparently the magic to preserve those values across cache resets isn't
strong enough for this. Seems bad, because that seems to mean a sinval
overflow leads to this and related optimizations being lost?

Greetings,

Andres Freund

--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services