> On Sep 9, 2019, at 9:45 AM, Vlad Khorsun <hv...@optima.com.ua> wrote:
>
> 07.09.2019 19:35, Ann Harrison wrote:
> ...
>> It's also an old bug, probably dating to the implementation of foreign keys
>> in
>> InterBase.
>> ...it's certainly legitimate.
>
> So, to fix this we need to repeat validation of master record in context of
> client transaction,
> correct ?
Yes, I think so. Only in the case of an insert by a snapshot transaction
inserting a child record. The current test returns an error in the case of a
recently deleted parent record. We need a second test to verify that th master
is visible to the inserting transaction.
> Of course, it should be made for concurrency (snapshot) transaction only and
> if current
> state of master record is not visible for client transaction.
>
Right. A small number of inserts will fail that might have succeeded, but the
problem Carlos demonstrated will be gone. (I think)
Cheers,
Ann
>
>
>
> Firebird-Devel mailing list, web interface at
> https://lists.sourceforge.net/lists/listinfo/firebird-devel
Firebird-Devel mailing list, web interface at
https://lists.sourceforge.net/lists/listinfo/firebird-devel