Joe Uhl <joeuhl(at)gmail(dot)com> writes:
> We have a query that returns a different result in 8.2.5 and 8.3.0.
> ...
> The result in 8.3.0 is what we expect.
I guess I read a little too fast, because I overlooked that comment.
It's the 8.3.0 plan that is broken in my opinion --- it shouldn't be
pushing down the IS NULL conditions. If you think that 8.2.5 is giving
a wrong answer then you need to provide a test case demonstrating that.
regards, tom lane