HMER: A Hybrid Mutation Execution Reduction approach for Mutation-based Fault Localization

Abstract

Identifying the location of faults in programs has been recognized as one of the most manually and time cost activities during software debugging process. Fault localization techniques, which seek to identify faulty program statements as quickly as possible, can assist developers in alleviating the time and manual cost of software debugging. Mutation-based fault localization(MBFL) has a promising fault localization accuracy, but suffered from huge mutation execution cost. To reduce the cost of MBFL, we propose a Hybrid Mutation Execution Reduction(HMER) approach in this paper. HMER consists of two steps: Weighted Statement-Oriented Mutant Sampling(WSOME) and Dynamic Mutation Execution Strategy(DMES). In the first step, we employ Spectrum-Based Fault Localization(SBFL) techniques to calculate the suspiciousness value of statements, and guarantee that the mutants generated from statements with higher suspiciousness value will have more chance to be remained in the sampling process. Next, a dynamic mutation execution strategy is used to execute the reduced mutant set on test suite to avoid worthless execution. Empirical results on 130 versions from 9 subject programs show that HMER can reduce 74.5%-93.4% mutation execution cost while keeping almost the same fault localization accuracy with the original MBFL. A further Wilcoxonsigned−ranktest indicates that when employing HMER strategy in MBFL, the fault localization accuracy has no statistically significant difference in most cases compared with the original MBFL without any reduction techniques.