Fix Director for Automated Program Repair

Abstract

As the scale of software increases, automated program repair (APR) is required as an essential technology to reduce bug-fixing manual effort. One of the most effective APR approaches is the template-based APR. Although this data- driven approach has made many improvements in fixing bugs, there is still a lack of information utilization at each stage of the program repair. In this paper, we propose a fix pattern prioritization APR technique. We do not just use the test cases to determine whether the source file is fault or not, but as a material for why the bug occurred. With this information, we select a more appropriate fix action and apply it first to increase the efficiency of the automated program repair.