David Newton wrote:You'd have to make sure the tests for the tests were working.
I'd never unit-tested a test case--that strikes me as weird. Don't get me wrong--test cases *can* be wrong, but you have to stop somewhere.
I have, but only on rare occasions. When I'm writing a generic test that is meant to be extended and I don't have any code to run it against. In this case, the test case is the product though, so I am testing the "real code."