Constructing test behavior models using simulated system answers for the analysis of test behavior anomalies
In the standardization of test specifications, it is common that no actual systems exist against which the tests can be executed. Test specifications are developed abstractly in high level languages such as the Testing and Test Control Notation (TTCN-3), but they can only be executed when a separate adaptation layer is implemented. Static syntactical and semantical analyses as provided by the compiler and proper manual code reviews are the only means to find mistakes in such test specifications at early stages of design. In this paper, we demonstrate that it is possible to execute abstract test specifications when the system does not exist yet. We use the information provided within the test cases to simulate answers of the system by generating inverse messages to expected messages in the abstract test case. By following a specific coverage-criterion strategy, we are able to execute a sufficient amount of test paths to reverse-engineer behavioral models of test cases which can then again be used for the analyses of potential problems.
Full Text: PDF