On Sun, 20 Jul 2003, Paul Thomas wrote:
>
> On 20/07/2003 14:15 Nigel J. Andrews wrote:
> >
> > As usual I forgot to include the version number. It's 7.3.3
> > >
> > > I'd really appreciate an explanation, since this test is based on
> > queries
> > > extracted from the db log, is only one specific example of this sort of
> > > operation from many in the driving program and most significantly it
> > seems I
> > > can't even write sql statments hardcoding these values as the test
> > script shows
> > > they still get the ref. int. error.
>
> A bit more detail about the tables might be helpful, constraints, triggers
> etc... How about doing a select of site_membership immediately before the
> delete. What does that show? Have you got a trigger somewhere that would
> insert a record into site_membership? This could cause the RI failure but,
> without the transaction being committed, the inserted record could be
> discarded and the table would still appear empty after the error. I'm very
> much clutching at straws here and am probably way off but without more
> details, wild guesses are the best I can do.
I understand, and that's all I'm hoping for at this stage.
No, there's no triggers on any of these tables. In the test script
site_membership is still empty immediately before the delete.
Nigel Andrews