The examples in this appendix have been necessarily simple. In practice, we can write tests that ensure the correct operation of quite complicated behaviors.
Ideally, we keep our tests as brief and simple as possible, even when the behaviors they are testing are intricate. By writing tests for a few specific scenarios, we can be reasonably certain that we are fully testing the behavior, even though we do not have a test for every possible set of inputs.
However, it is possible that an error is observed in our code even though we have written tests for it. When tests pass and yet an error occurs, the correct response is not to immediately fix the problem, but rather to first write a new test for the behavior that fails. This way, we can not only verify that the problem is solved when we correct the code, but also introduce an additional test that will help us avoid regressions in the future...
