Better predicate testing
Gary Kaminski, Paul Ammann, Jeff Offutt · 2011
Mutation testing is widely recognized as being extremely powerful, but is considered difficult to automate enough for practical use. This paper theoretically addresses two pos-sible reasons for this: the generation of redundant mutants and the lack of integration of mutation analysis with other test criteria. By addressing these two issues, this paper brings an important mutation operator, relational-operator-replacement (ROR), closer to practical use. First, we de-velop fault hierarchies for the six relational operators, each of which generates seven mutants per clause. These hierar-chies show that, for any given clause, only three mutants are necessary. This theoretical result can be integrated easily into mutation analysis tools, thereby eliminating generation of 57 % of the ROR mutants. Second, we show how to bring the power of the ROR operator to the widely used Multiple Condition-Decision Coverage (MCDC) test criterion. This theoretical result includes an algorithm to transform any MCDC-adequate test set into a test set that also satisfies RORG, a new version of ROR appropriate for the MCDC context. The transformation does not use traditional muta-tion analysis, so can easily be integrated into existing MCDC tools and processes.