Understanding and generating high quality patches for concurrency bugs

Abstract

Concurrency bugs are time-consuming to fix correctly by developers and a severe threat to software reliability. Although many auto-fixing techniques have been proposed recently for concurrency bugs, there is still a big gap between the quality of automatically generated patches and manually designed ones. This paper first conducts an in-depth study of manual patches for 77 real-world concurrency bugs, which provides both assessments for existing techniques and actionable suggestions for future research. Guided by this study, a new tool HFix is designed. It can automatically generate patches, which have matching quality as manual patches, for many concurrency bugs.