Accelerating Mutation-Based Fault Localization via Optimized Mutant Execution
Xia Li, Ryan Juza · 2023
As modern software systems become more and more complex, bugs (also called faults) are prevalent and inevitable. Automated fault localization focuses on identifying faulty program elements that cause software failures. Mutation-based fault localization (MBFL) is one popular approach and has been proved to be more effective in real bugs. However, MBFL brings a problem of execution cost since it requires the execution of the whole test suite against a huge number of generated mutants. In this paper, we propose FastMBFL, a novel technique to improve the efficiency of MBFL by considering the importance of the impact information originated from failed test while maintaining the similar accuracy. The experimental results show that in terms of the widely used evaluation metric Mutant-Test Pair (MTP), FastMBFL can reduce the mutation cost by 24%–40% compared with the state-of-the-art technique.