MURE: Making Use of MUtations to REfine Spectrum-Based Fault Localization

Locating faults in programs is never an easy task. Spectrum-based fault localization (SBFL) techniques estimate suspicious statements by contrasting the coverage spectra collected from passed and failed program runs. Mutation-based such techniques locate faults by trying different mutates with the aim of finding one that involves less turbulence to program behavior. The latter is empirically known more accurate, but with massive increases in time complexity. In this paper, we propose a new approach, MURE, which uses methodology of the latter to refine results of the former. MURE first drives a stateor-the-art SBFL technique Naish2 to output a list of suspicious statements. It then picks out suspicious statement as candidates, generates mutates for them, and estimates their likelihood of relating to faults. Finally, it refines the resultant list by adjusting part of its ordering. An experiment validates its effectiveness by showing a 30% accuracy improvement over Naish2.