On 2014-05-15 18:52:45 -0300, Euler Taveira wrote:
> On 15-05-2014 18:09, Andres Freund wrote:
> > I have some preliminary tests for the pg_recvlogical binary using the
> > infrastructure Peter added. I am wondering if somebody has a good idea
> > about how to make the tests more meaningful. Currently all that's tested
> > are simple commands. Not the main functionality namely the actual
> > streaming of changes.
> >
> Could you post your preliminary patch?
It's not particularly interesting yet, but attached.
> > I wonder if somebody has a great idea for cancelling pg_recvlogical
> > after a while. Right now my best idea is to add a new
> > --stop-after-messages parameter. Does anybody have a better idea?
> >
> pgbench uses two stop conditions: (i) number of transactions and (ii)
> time. These could be applied to pg_recvlogical too.
>
> --transactions=NUM
> --time=NUM
Sounds too complicated for what I need it for. The former isn't easily
implementable because pg_recvlogical doesn't know about transactions and
the latter isn't that interesting for the tests because it'll have
timing issues...
> > If that's the way, would somebody object to tests and the parameter
> > being added now? The potential harm seems pretty low and the additional
> > tests would cover the walsender interface that's not covered by any
> > tests right now...
> >
> beta is not the right time to add even a *minor* functionality. Let's do
> it for 9.5.
I am not interested in the feature iself at all. I am interested into
adding more tests for new functionality that's currently not tested in
an automated fassion. And I do think beta isn't a bad time for that. We
didn't *have* the infrastructure for the tests of binaries until the end
of the last CF, so I couldn't have added it during development.
Greetings,
Andres Freund
--
Andres Freund http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services
