On path-wise automatic generation of test data for both white-box and black-box testing

Automatic generation of test data for a given path in a program is an elementary problem in software testing, the difficulty of which lies in how to solve the nonlinear constraint. Gupta et al. (1998) proposed a method, which is referred to as the iterative relaxation method, to address the above problem by linearizing the predicate functions. This paper improves the iterative relaxation method by omitting the constructions of predicate slice and input dependency set, and proves the equivalence of systems of constraints generated by both methods. Since it is not necessary for our method to analyze the dependencies between statements on the path in the course of deriving a system of constraints, our method still works when some statements are only object or executable codes rather than source codes on the path. Therefore, our method can also be used for generating test data for black-box testing and regression testing. We have developed a prototype of a path-wise test data generator whose fundamental algorithm is presented in this paper. The initial experiments with this prototype have shown that our method is practical.