Catversion bumps during beta (was Re: [Proposal] Expose internal MultiXact member count function for efficient monitoring) - Mailing list pgsql-hackers

On 22/09/2026 18:46, Nathan Bossart wrote:
> On Tue, Sep 22, 2026 at 09:58:38AM -0400, Melanie Plageman wrote:
>> On Tue, Sep 22, 2026 at 4:01 AM Michael Paquier <michael@paquier.xyz> wrote:
>>> That would be OK here.  CC-ing Nathan for RMT input, as this would
>>> require a catversion bump.
>>
>> I don't have an issue with a catversion bump as long as it's before GA
>> and it doesn't make it more likely we'll have to change it again after
>> GA.
> 
> Same.  I'd think before RC1 would be preferable.

I'm cool with it. Sooner the better, of course.

I don't really mind catversion bumps during the beta period in general, 
I'm in the camp that if there's a change that would make the release 
better and it's not particularly risky, it's better to still sneak it in.

I feel I might be in the minority with that view though. Historically, 
we've avoided fixes that need catversion bumps late in the beta period, 
so that beta-testers don't need to pg_upgrade. I think that's a poor 
tradeoff, but if we're changing long-standing policy, the RMT probably 
isn't the right group for that. So, how do others feel?

- Heikki