Towards automatic software fault location through decision-to-decision path analysis

Software development is a complex and error prone process. As a result of this process, much time is spent debugging software. This debugging process actually consists of two activities, fault localization and repair. For most problems, much of the debugging effort is devoted to fault localization. In this paper, current fault localization techniques are surveyed and a new technique called relational path analysis is proposed. Relational path analysis suggests that there exists information associated with stored execution paths of programs that, when analyzed heuristically, can localize faults with statistical significance. This paper presents a set of candidate heuristics for relational path analysis and the results of an experiment utilizing the heuristics. Conclusions regarding the effectiveness and usability of this technique and future research in this area are also discussed.