Abstract: Evolutionary testing is a highly effective technique for automatically generating high quality test data, which is used for structural testing. However,under the criterion of Nodeoriented,testing of unstructured programs is inefficicnt and leading the technique degenerates to random testing. In this paper, with regard to the unstructured programs that contain arbitrary jump statement inner a loop body, a method of fitness calculation based on traditional approach was proposed, in which the impact of the number of iteration for evolutionary search was adequately considered. The experiments were then presented and the results show that the fitness function could effectively guide evolutionary search to find recauired test data at low cost.

Key words: Evolutionary testing, Unstructured control flow, Loop, Fitness, Evolutionary search