Tom Lane wrote:
> Alvaro Herrera <alvherre@commandprompt.com> writes:
> > SIBackendInit returns a flag indicating whether it worked or not. Since
> > there is only one caller and it dies with a FATAL error when
> > SIBackendInit failed, it seems better to move the elog and remove the
> > return value, per this patch.
>
> The reason for the existing coding is to release the SInvalLock before
> throwing the error. Now proc_exit cleanup should release the lock
> anyway, but this proposed change will mean that a failing backend holds
> the lock a bit longer before releasing, which might be a bad thing.
I'll leave this alone until early 8.4. Thanks for the insight.
--
Alvaro Herrera http://www.CommandPrompt.com/
PostgreSQL Replication, Consulting, Custom Development, 24x7 support