On 26/05/2026 12:33, Andrey Borodin wrote:
>> On 26 May 2026, at 14:27, Michael Paquier <michael(at)paquier(dot)xyz> wrote:
>>
>> Hmm, okay. I am not sure what you mean here, could you demonstrate
>> your idea with a patch later?
>
> Something like attached, not tested yet, working on an automated test.
Yeah, that looks correct to me. It moves the locking on v16 to where it
happens on v17 and v18. I don't see any reason to hold the lock in the
earlier parts of RecordNewMultiXact() in v16.
- Heikki