On Wed, 3 Apr 2002 swalker@iglou.com wrote:
>
> So turning intially deferred on is a suitable solution? It seems
> to work OK in this simple test case.
That'll hide this particular locking issue (since the locks are held
for much less time in that case). There are some issues with
initially deferred constraints right now in some cases if you're
doing actions on both tables in one transaction (like deleting a
pk row and then re-adding it).
>
> On Tue, 2 Apr 2002, Stephan Szabo wrote:
> > The locks being grabbed are a bit stronger than they need to be
> > (if you want more info, there was a bit of discussion on -hackers
> > recently). The updates should block updates on parent of the
> > referenced row but don't need to block other child updates but there's
> > no current lock level that is quite right.
>
