Toward a methodology to expose partially fixed concurrency bugs in modified multithreaded programs

Abstract

Many multithreaded programs incur concurrency bugs. A modified version of such a program, which the exposed concurrency bug is deemed fixed, should be subject to further testing to validate whether the concurrency bug may only be partially fixed. In this paper, we present a similarity-based regression testing methodology to address this problem. It is based on the notions of similar execution contexts of events and bug signatures. To the best of our knowledge, it also presents the first regression testing technique that manipulates thread schedules using a similarity-based active testing strategy.