PathART: path-sensitive adaptive random testing

Abstract

As test data widely spreading on the input domain may not thoroughly test the program's logic, in this paper, we propose an approach to generating test data widely spreading on a program's execution paths. In particular, we analyze execution paths of the program, distill constraints for executing the paths, calculate the path distance between test data according to their satisfaction for paths' constraints, and then generate test data far away from each other based on their path distance. The experimental results show that our approach significantly reduces the number of test data generated before the first fault is found.