The State Problem for Evolutionary Testing

This paper shows how the presence of states in test objects can hinder or render impossible the search for test data using evolutionary testing. Additional guidance is required to find sequences of inputs that put the test object into some necessary state for certain test goals to become feasible. It is shown that data dependency analysis can be used to identify program statements responsible for state transitions, and then argued that an additional search is needed to find required transition sequences. In order to be able to deal with complex examples, the use of ant colony optimization is proposed. The results of a simple initial experiment are reported.