Re: pg_rewind race condition just after promotion

On 22/02/2023 16:00, Heikki Linnakangas wrote:
> On 11/12/2022 02:01, Ian Lawrence Barwick wrote:
>> 2021年11月9日(火) 20:31 Daniel Gustafsson <daniel(at)yesql(dot)se>:
>>>
>>>> On 14 Jul 2021, at 14:03, Aleksander Alekseev <aleksander(at)timescale(dot)com> wrote:
>>>>
>>>> The following review has been posted through the commitfest application:
>>>> make installcheck-world: tested, passed
>>>> Implements feature: tested, passed
>>>> Spec compliant: tested, passed
>>>> Documentation: tested, passed
>>>>
>>>> The v3 patch LGTM. I wonder if we should explicitly say in pg_rewind tests that
>>>> they _don't_ have to call `checkpoint`, or otherwise, we will lose the test
>>>> coverage for this scenario. But I don't have a strong opinion on this one.
>>>>
>>>> The new status of this patch is: Ready for Committer
>>>
>>> Heikki, do you have plans to address this patch during this CF?
>>
>> Friendly reminder ping one year on; I haven't looked at this patch in
>> detail but going by the thread contents it seems it should be marked
>> "Ready for Committer"? Moved to the next CF anyway.
>
> Here's an updated version of the patch.
>
> I renamed the arguments to findCommonAncestorTimeline() so that the
> 'targetHistory' argument doesn't shadow the global 'targetHistory'
> variable. No other changes, and this still looks good to me, so I'll
> wait for the cfbot to run on this and commit in the next few days.

Pushed. I decided not to backpatch this, after all. We haven't really
been treating this as a bug so far, and the patch didn't apply cleanly
to v13 and before.

- Heikki