Re: BUG #13725: Logical Decoding - wrong results with large transactions and unfortunate timing

On Mon, Oct 26, 2015 at 12:05 PM, Andres Freund <andres@anarazel.de> wrote:

> >
> > So we compute end_of_wal before entering the loop, but the WAL keeps
> > growing as we read through it.
>
> So?
>
> > If we do it correctly, there's potential that the loop will never finish
> if
> > the WAL grows faster than we can decode it.
>
> > Shouldn't we also try to re-write this SRF to support
> > SFRM_ValuePerCall?
>
> Why? ValuePercall gets materialized into a tuplestore as well, unless
> you call it from the select list.
>

Ah, yes: good catch.  Then it makes less sense to change this indeed.

--
Alex