2017 IEEE International Conference on Software Maintenance and Evolution (ICSME) (2017)
Sept. 17, 2017 to Sept. 22, 2017
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/ICSME.2017.37
Spectrum-based fault localization (SFL), the technique producing a rank list of statements in descending order of their suspiciousness values, is nowadays widely used in current automated program repair tools. There are two different algorithms for these tools to choose statements selected for modification to produce candidate patches from the list: one is the rank-first algorithm based on suspiciousness rankings of statements, the other is the suspiciousness-first algorithm based on suspiciousness value of statements. However, to our knowledge there is no research work implementing the two algorithms in the same repair tool or comparing their effectiveness. In this paper, we conduct an empirical research based on the automated repair tool Nopol with the benchmark set of Defects4J to compare these two algorithms. Preliminary results suggest that the suspiciousness-first algorithm is not equivalent to the rank-first algorithm and behaves better in parallel repair and patch diversity.
Maintenance engineering, Tools, Measurement, Computer bugs, Classification algorithms, Benchmark testing, Java
Deheng Yang, Yuhua Qi, Xiaoguang Mao, "An Empirical Study on the Usage of Fault Localization in Automated Program Repair", 2017 IEEE International Conference on Software Maintenance and Evolution (ICSME), vol. 00, no. , pp. 504-508, 2017, doi:10.1109/ICSME.2017.37