Leveraging Method Call Anomalies to Improve the Effectiveness of Spectrum-Based Fault Localization Techniques for Object-Oriented Programs

Spectrum-based fault localization (SFL) is a lightweight automated diagnosis technique. However, when applied to object-oriented programs, its diagnostic accuracy is limited as suspicious statements are distributed in different classes. In this paper, we propose an approach to leveraging method call anomalies to improve the effectiveness of SFL techniques for locating faulty statements in an object oriented program. First, we compute the suspiciousness for each class based on the difference in its method call sequences between passed and failed runs. Then, we use the suspiciousness information of classes to refine SFL ranks in order to enhance their fault localization effectiveness of object-oriented software. The empirical results show that the proposed approach is able to improve the effectiveness of SFL techniques.