Re: [BUG] Fix DETACH with FK pointing to a partitioned table fails

Jehan-Guillaume de Rorthais <jgdr(at)dalibo(dot)com> 于2024年9月3日周二 17:26写道：

> Hi Tender,
>
> On Tue, 3 Sep 2024 10:16:44 +0800
> Tender Wang <tndrwang(at)gmail(dot)com> wrote:
>
> > Jehan-Guillaume de Rorthais <jgdr(at)dalibo(dot)com> 于2024年9月3日周二 05:02写道：
> […]
> > > * Constraint & trigger catalog cleanup [1] (this thread)
> > > * FK broken after DETACH [2]
> > > * Maintenance consideration about self referencing FK between
> partitions
> > > [3]
> > >
> >
> > The third issue has been fixed, and codes have been pushed. Because of
> my
> > misunderstanding,
> > It should not be here.
>
> I just retried the SQL scenario Guillaume gave on both master and master
> with
> Alvaro's patch. See:
>
>
> https://www.postgresql.org/message-id/flat/CAECtzeWHCA%2B6tTcm2Oh2%2Bg7fURUJpLZb-%3DpRXgeWJ-Pi%2BVU%3D_w%40mail.gmail.com
>
> It doesn't seem fixed at all. Maybe you are mixing up with another
> thread/issue?
>

Sorry, I mixed up the third issue with the Alexander reported issue.
Please ignore the above noise.

>
> > > 0. Splitting in two commits
> > >
> > > […]
> > >
> > > Unfortunately, this discussion about the first bug slipped to the
> second
> > > one when Tender stumbled on this bug as well and reported it. But,
> both
> > > bugs can be triggered independently, and have distinct fixes.
> >
> > It's ok that these two issues are fixed together. It is because current
> > codes don't handle better when the referenced side is the partition
> table.
>
> I don't feel the same. Mixing two discussions and fixes together in the
> same
> thread and commit makes life harder.
>

Hmm, these two issues have a close relationship. Anyway, I think it's ok to
fix the two issues together.

> Last year, when you found the other bug, I tried to point you to the
> right thread to avoid mixing subjects:
>
> https://www.postgresql.org/message-id/20230810170345.26e41b05%40karst
>
> If I wrote about the third (non fixed) issue yesterday, it's just because
> Alvaro included a reference to it in his commit message. But I think we
> should
> really keep up with this issue on its own, dedicated discussion:
>
>
> https://www.postgresql.org/message-id/flat/CAECtzeWHCA%2B6tTcm2Oh2%2Bg7fURUJpLZb-%3DpRXgeWJ-Pi%2BVU%3D_w%40mail.gmail.com

Thanks for the reminder. I didn't take the time to look into the third
issue. Please give me some to analyze it.

--
Thanks,
Tender Wang