As model-driven development draws more and more attention,Testing technology of EFSM models has been in-depth studied. However,test data generation of EFSM, which has become a new research area, isstill an open issue. Test data generation, which is derived from testsequence generation, can test EFSM models more accurately.Currently, test data generation of EFSM models are mainly based onheuristic search or symbolic execution. Generally, researches about testdata generation of EFSM models focus on one single model and does notprocess procedure calls. That limits the scope of EFSM models. Toaddress this issue, EFSM model with procedure and corresponding testdata generation strategy are proposed in this paper. First, a sub-procedureis described as an EFSM model. Then the EFSM model representing themain procedure and the EFSM models representing the sub-procedure are merged into a new EFSM model. Genetic algorithm is employed togenerate test data. Finally, the experimental results show that theproposed approach is promising and effective in test data generation forEFSM models in which procedure calls is embedded.As is an initial solution, the proposed approach comes with somedefects. At the end of this paper, analysis about the defects andcorresponding improvement strategies are presented. |