Explanations should take around 40 words. Some slack will be given to 50 or 60 word explanations (however, don't press your luck too many times on this); explanations a page or more long will lose five points.

1. A colleague of yours had the following bright idea: rather than test for faults, test for failures. However, your colleague is unsure which of the two testing methods - structural or functional - is more appropriate for failure testing. Recognizing your expertise in software engineering, your colleague comes to you for your opinion. What do you say?

   ---

   Your colleague is apparently trying to save testing time by skipping the search for errors and looking directly for failures. Because failures are program-internal manifestations of faults, structural testing would be the appropriate method to try out your colleague's ideas.

   ---

2. Give two explanations about why it may be a good idea to test code that has known errors in it.

   ---

   One reason is to test the quality of the tests designed for that code. If the tests can't find any of the errors, then it's probably a good idea to reconsider the criteria and generation methods associated with the test data.

   A second reason to test for known errors is that testing for a specific type of known error may reveal unknown errors of the same type. For example, testing for known buffer-overflow errors may reveal other, unknown buffer overflows.

   A third reason, related to the first, is that testing for a known number of errors can provide an estimate of the latent errors remaining after testing. If the test suite found 80% of the known errors, then it would be reasonable (or at least not unreasonable) to use 20% when estimating the latent-error count.

   ---

3. As a member of the test department, you have been asked to sit in on a system-design review. Explain what kind of coupling are you looking for and why.

   ---

   You are looking for low coupling among modules (of whatever kind) because low coupling leads to easier testing. The lower the coupling, the greater the independence among modules. A set of relatively independent modules is easer to test than is a set of inter-dependent modules.

   ---

4. You are part of a team developing a system that plays tic-tac-toe, and it is your responsibility to come up with a test plan for the following subsystem:

   bool winning_move(tic_tac_toe_board b, player p)
   

   that returns true if player p (either X or O) has winning move given the board b and false otherwise. Explain the test method and test cases you would use (and don't forget the 40 word limit).

   ---

   The key is to note that there are at most (3⁹)*2, or around 40,000, possible inputs to winning_move(). This suggests that exhaustive testing is appropriate, and because the tests would involve only inputs and outputs, the tests would be functional.

   ---

This page last modified on 14 April 2001.