A Non-Parametric Statistical Debugging Technique with the Aid of Program Slicing (NPSS)

Abstract

A method is introduced in this paper, wh ich promotes automated bug localization. It is based on the combination of two bug localization techniques, Non-Paramet ric Statistical Debugging and Backward Slicing. The proposed method, co mputes some vectors (called execution vectors) based on the status of each basic-block's execution in running of test-cases. According to the behavior of each basic-block in failed test-cases and passed ones, two likelihoods are computed and regards to them, basic-b locks become prioritized. At last static slice of p rogram and dynamic backward slice for one failed test-case are computed. While seeking for faulty statement in ran ked basic-blocks, the method either returns the basic-block's statements in the static backward slice or the part of it presented in the co mputed dynamic backward slice. NPSS has been applied on the Siemens test suite, space, grep and gzip. Our experimental study shows the accuracy and effectiveness of the method in accurate bug localization.