Automatic detection of infeasible paths in software testing

Abstract

A challenging problem in path-oriented test data generation is the presence of infeasible paths. Timely detecting these infeasible paths cannot only save test resources but also improve test efficiency. A popular method of detecting infeasible paths is to determine branch correlations, which is a difficult task and usually cannot be done timely and exactly. In this study, the authors propose a method of automatically determining the branch correlations of different conditional statements, therefore detecting infeasible paths. First, some theorems are given to determine branch correlations based on the probabilities of the conditional distribution corresponding to different branches' outcome (i.e. true or false); then, the maximum likelihood estimation is employed to obtain the values of these probabilities; finally, infeasible paths are detected according to branch correlations. The authors apply the proposed method in some typical programs, and the results show that the proposed method can accurately detect infeasible paths. The achievement provides an effective and automatic method of detecting infeasible paths, which is significant in improving the efficiency of software testing.