On the Effect of Counters in Guard Conditions When State-Based Multi-objective Testing

During test case generation from an extended finite state machine (EFSM), the counter problem is caused by the presence of guard conditions that refer to counter variables. Because such variables are initialized and updated by transitions in the EFSM, every traversal of the state machine graph is not necessarily feasible, i.e., executable. The problem manifests itself by the fact that a transition, a sequence of transitions, or a more complex behavior in the state machine, has to be repeatedly triggered to eventually trigger a specific behavior (another transition). In this paper we define different manifestations of the counter problem and experiment with a new search based solution for that problem. We also investigate how the counter problem affects a multi-objective genetic algorithm that generates test suites from an EFSM. We evaluate our solution and compare it with an existing one, using three different case studies.