Interactive Patch Filtering via Test Generation
Quanjun Zhang, Xu Zhai, Shicheng Xu, Wanmin Huang, Jingui Zhang, Yaoming Fan · 2022 9th International Conference on Dependable Systems and Their Applications (DSA) · 2022
Automatic program repair (APR), which aims to fix software bugs without human intervention, is getting in-creasing attention from academic and industrial communities. Although promising outcomes regarding correctly-fixed bugs have been achieved recently, existing APR tools still suffer from the low accuracy of generated patches. In fact, it is fundamentally difficult to avoid generating incorrect patches due to the weak available test suite. In this paper, to improve the accuracy of patches generated by APR tools, we propose a novel HUman-machine interactive patch filterinG apprOach (HUGO) to help developers identify correct patches by generating additional test cases. We also implement the approach as an Eclipse plugin and evaluate the effectiveness and usefulness of the implementation. The results on the Defects4J dataset show that the proposed method can filter out 82.61 % of the incorrect patches, and improve the accuracy of patches by 25%.