There may come a point in your programming adventures where you are forced to write code against a set of tests which cannot be seen by you. This is most likely in an academic setting, since they don't want students coding to tests.

There may also come a point where you are shown to be passing all test cases of some particular tested feature, but others are failing in a seemingly unrelated area.

No matter what you do or how you rewrite your code or how much you verify that your logic is correct, those tests always fail.

Well, eventually you have to ask yourself, "is the test category that I'm passing really feature-complete?" Maybe not. Maybe the reason you're failing these tests in the first place isn't because of a problem with the current feature, but an already assumed-to-be-working feature.

So, assume the tests aren't feature complete, and create more tests in that area. Think of things outside the box that might not have been covered, but are still within the requirements, and maybe you'll find that bug that once fixed will allow you to pass all cases.