Applications of Optimization to Logic Testing

Abstract

A tradeoff exists in software logic testing between test set size and fault detection. Testers may want to minimize test set size subject to guaranteeing fault detection or they may want to maximize faults detection subject to a test set size. One way to guarantee fault detection is to use heuristics to produce tests that satisfy logic criteria. Some logic criteria have the property that they are satisfied by a test set if detection of certain faults is guaranteed by that test set. An empirical study is conducted to compare test set size and computation time for heuristics and optimization for various faults and criteria. The results show that optimization is a better choice for applications where each test has significant cost, because for a small difference in computation time, optimization reduces test set size. A second empirical study examined the percentage of faults detected in a best, random, and worst case, first for a test set size of one and then again for a test set size of ten. This study showed that if you have a limited number of tests from which to choose, the exact tests you choose have a large impact on fault detection.