An Evaluation of the Minimal-MUMCUT Logic Criterion and Prime Path Coverage.,

The 2010 International Conferenceon Software Engineering Research and Practice, Las Vegas Nevada, USA. To appear.

Garrett Kaminski, Upsorn Praphamontripong, Paul Ammann, Jeff Offutt

Abstract

This paper presents comparisons of the Minimal- MUMCUT logic criterion and prime path coverage. A theoretical comparison of the two criteria is performed in terms of (1) how well tests satisfying one criterion satisfy the other and (2) fault detection. We then compare the criteria experimentally. For 22 programs, we develop tests to satisfy Minimal-MUMCUT and prime path coverage. We use these tests in two separate experiments. First we measure the effectiveness of the tests developed for one criterion in terms of the other. Next we investigate the ability of the test sets to find actual faults. Faults are seeded via a mutation tool and then supplemented with mutants created by DNF logic mutation operators. We then measure the number of non-equivalent mutants killed by each test set. Results indicate that while prime path-adequate test sets are closer to satisfying Minimal-MUMCUT than vice versa, the criteria had similar fault detection and Minimal- MUMCUT required fewer tests.



Back to my home page.