Research On Methods For Executable Test Sequence Generation Based On EFSM

There exists ambiguity problem of protocol specifications for that they are described bynatural language. The root causes of different protocol implementation is that protocoldevelopers have various understanding of protocol specification. The slightest deviation betweenthe protocol implementation and the protocol specification may cause incalculable consequencesfor computer communication network. The aim of protocol conformance testing is to validate thedegree of conformance between protocol specification and protocol implementation. ExtendedFinite State Machine(EFSM) provides a relative unambiguous and accurate platform forconformance testing and methods based on EFSM to generate test sequences are a hot topic incurrent research area.EFSM expands FSM(Finite State Machine) on variables and predicate conditions whichcause the unfeasible problems of generated test sequences. Studying deeply we can find that theessence is that data conflicts exist between predicate conditions of transitions. Aiming at the keyissues using EFSM to generate executable test sequences, a systematic and in-depth work hasbeen researched in this document. The research work and innovations can be attributed to thefollowing four aspects:(1) During the procedure of test sequence generation, this paper finds the limitations ofexecutable analysis tree method under particular test scenario. Through extracting characteristicsof the related EFSM, this paper analyzes and discusses its nature reasons.(2) The constraint relationship between transitions on feasibility has been deeplyresearched. Based on the relation, predicate dependence graph of the EFSM is constructed todirect the generation of test sequences. PDG extends the forecast from one step to N-step whichmakes the direction of test sequence generation more accurate.(3) The edge of predicate dependence graph has been quantified. A heuristic optimizationfunction is proposed and the three questions about test sequence generation become the factor ofthe heuristic function which will be used to generate test sequences.(4) This paper proposes a method for generating test sequences based on parametricexecutable analysis tree. The new method generates test sequences without input test data. Alongwith the generation of test sequences, constraints about variables will be derived. The values of variables which meet the constraints will always trigger the derived test sequences whichimplements automatic generation of feasible test sequences.