Re: START_REPLICATION SLOT causing a crash in an assert build

On Sun, Oct 9, 2022 at 2:42 AM Andres Freund <andres(at)anarazel(dot)de> wrote:
>
> On 2022-10-08 09:53:50 -0700, Andres Freund wrote:
> > On 2022-10-07 19:56:33 -0700, Andres Freund wrote:
> > > I'm planning to push this either later tonight (if I feel up to it after
> > > cooking dinner) or tomorrow morning PST, due to the release wrap deadline.
> >
> > I looked this over again, tested a bit more, and pushed the adjusted 15 and
> > master versions to github to get a CI run. Once that passes, as I expect, I'll
> > push them for real.
>
> Those passed and thus pushed.
>
> Thanks for the report, debugging and review everyone!

Thanks!

>
>
> I think we need at least the following tests for replslots:
> - a reset while decoding is ongoing works correctly
> - replslot stats continue to be accumulated after stats have been removed
>
>
> I wonder how much it'd take to teach isolationtester to handle the replication
> protocol...

I think we can do these tests by using pg_recvlogical in TAP tests.
I've attached a patch to do that.

Regards,

--
Masahiko Sawada
PostgreSQL Contributors Team
RDS Open Source Databases
Amazon Web Services: https://aws.amazon.com