On Thu, Apr 8, 2021 at 5:56 AM Andres Freund <andres@anarazel.de> wrote:
>
> Hi,
>
> On 2021-04-08 09:17:42 +0900, Michael Paquier wrote:
> > On Wed, Apr 07, 2021 at 11:09:41AM -0300, Euler Taveira wrote:
> > > On Wed, Apr 7, 2021, at 10:25 AM, Bharath Rupireddy wrote:
> > >> I agree to not remove "with (oids = false)". At least shouldn't we fix
> > >> the "create table ... with (oids = false, oids = false ....)" case,
> > >> just to be consistent with other options?
> > >
> > > It would be weird to error out while parsing a no-op option, no?
> >
> > There is an argument to be made both ways here.
>
> > >> But, why do we need to allow specifying oids = false multiple times(see
> > >> below)? Shouldn't we throw an error for consistency with other options?
> > >>
> > >
> > > If you look at transformReloptions(), the no-op code is just a hack. Such a
> > > patch should add 'oids' as a reloption to test for multiple occurrences.
> > > Although, CREATE TABLE says you can use 'oids=false', Storage Parameters
> > > section does not mention it as a parameter. The code is fine as is.
> >
> > But I agree with letting what we have here as it is, per the same
> > argument of upthread that this could just break stuff for free, and
> > that's not a maintenance burden either.
>
> Agreed.
>
> Given that this case didn't error out before the OIDs removal, it seems
> like it'd be really strange to make it error out in the compat code...
Agreed to not error out for a no-op case i.e. with (oids = false, oids
= false). Thank you all for providing thoughts. I'm ending the
discussion here.
With Regards,
Bharath Rupireddy.
EnterpriseDB: http://www.enterprisedb.com
