Because it's the test which is failing, not the code. Unfortunately, it can't be reliably tested :-(.