On 2014-07-28 15:29:57 -0400, Tom Lane wrote:
> Andres Freund <andres@2ndquadrant.com> writes:
> > One thing I am wondering about around this is: Why are we only
> > processing catchup events when DoingCommandRead? There's other paths
> > where we can wait for data from the client for a long time. Obviously we
> > don't want to process async.c stuff from inside copy, but I don't see
> > why that's the case for sinval.c.
>
> It might be all right to do it during copy, but I'd just as soon treat
> that as a separate issue. If you merge it into the basic patch then it
> might be hard to get rid of if there are problems.
Yea, not planning to merge it. Just wondering to make sure I understand
all the implications.
Another thing I'm wondering about - also not for the basic patch - is
accepting termination while writing to the client. It's rather annoying
that we currently don't allow to pg_terminate_backend() when writing to
the client.
Greetings,
Andres Freund
-- Andres Freund http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services