Re: simplifying foreign key/RI checks

On 2021-Nov-11, Tom Lane wrote:

> It appears to me that commit 71f4c8c6f (DETACH PARTITION CONCURRENTLY)
> broke the semantics here, because now things work differently with a
> partitioned PK table than with a plain table, thanks to not bothering
> to distinguish questions of how to handle partition detachment from
> questions of visibility of individual data tuples. We evidently
> haven't got test coverage for this :-(, which is perhaps not so
> surprising because all this behavior long predates the isolationtester
> infrastructure that would've allowed us to test it mechanically.
>
> Anyway, I think that (1) we should write some more test cases around
> this behavior, (2) you need to establish the snapshot to use in two
> different ways for the RI_FKey_check and ri_Check_Pk_Match cases,
> and (3) something's going to have to be done to repair the behavior
> in v14 (unless we want to back-patch this into v14, which seems a
> bit scary).

I think we (I) should definitely pursue fixing whatever was broken by
DETACH CONCURRENTLY, back to pg14, independently of this patch ... but
I would appreciate some insight into what the problem is.

--
Álvaro Herrera 39°49'30"S 73°17'W — https://www.EnterpriseDB.com/
"Find a bug in a program, and fix it, and the program will work today.
Show the program how to find and fix a bug, and the program
will work forever" (Oliver Silfridge)