Mutation-based Test-Case Prioritization in Software Evolution

During software evolution, to assure the software quality, test cases for an early version tend to be reused by its latter versions. As a large number of test cases may aggregate during software evolution, it becomes necessary to schedule the execution order of test cases so that the faults in the latter version may be detected as early as possible, which is test-case prioritization in software evolution. In this paper, we proposed a novel test-case prioritization approach for software evolution, which first uses mutation faults on the difference between the early version and the latter version to simulate real faults occurred in software evolution, and then schedules the execution order of test cases based on their fault-detection capability, which is defined based on mutation faults. In particular, we present two models on calculating fault-detection capability, which are statistics-based model and probability-based model. Moreover, we conducted an experimental study and found that our approach with the statistics-based model outperforms our approach with the probability-based model and the total statement coverage-based approach, and slightly outperforms the additional statement-coverage based approach in many cases. Furthermore, compared with the total or additional statement coverage-based approach, our approach with either the statistics-based model or the probability-based model tends to be stably effective when the difference on the source code between the early version and the latter version is non-trivial.