Codeflaws: a programming competition benchmark for evaluating automated program repair tools

Abstract

Several automated program repair techniques have been proposed to reduce the time and effort spent in bug-fixing. While these repair tools are designed to be generic such that they could address many software faults, different repair tools may fix certain types of faults more effectively than other tools. Therefore, it is important to compare more objectively the effectiveness of different repair tools on various fault types. However, existing benchmarks on automated program repairs do not allow thorough investigation of the relationship between fault types and the effectiveness of repair tools. We present Codeflaws, a set of 3902 defects from 7436 programs automatically classified across 39 defect classes (we refer to different types of fault as defect classes derived from the syntactic differences between a buggy program and a patched program).