Conditional Testing

There are several powerful automatic testers available, each with different strengths and weaknesses. To immediately benefit from different strengths of different tools, we need to investigate ways for quick and easy combination of techniques. Until now, research has mostly investigated integrated combinations, which require extra implementation effort. We propose the concept of conditional testing and a set of combination techniques that do not require implementation effort: Different testers can be taken ‘off the shelf’ and combined in a way that they cooperatively solve the problem of test-case generation for a given input program and coverage criterion. This way, the latest advances in test-case generation can be combined without delay. Conditional testing passes the test goals that a first tester has covered to the next tester, so that the next tester does not need to repeat work (as in combinations without information passing) but can focus on the remaining test goals. Our combinations do not require changes to the implementation of a tester, because we leverage a testability transformation (i.e., we reduce the input program to those parts that are relevant to the remaining test goals). To evaluate conditional testing and our proposed combination techniques, we (1) implemented the generic conditional tester , including the required transformations, and (2) ran experiments on a large amount of benchmark tasks; the obtained results are promising.