SeTGaM: Generalized Technique for Regression Testing Based on UML/OCL Models

In this paper we introduce SeTGaM, a Model-Based Regression Testing (MBRT) approach based on UML/OCL behavioral models. SeTGaM is a test selection and classification approach that also generates new tests to cover new functionalities of a new version of a system. We extract the behavior of the system from guards/transitions of state charts or pre/post conditions in operations of class diagrams to which we apply impact analysis. This makes it possible to apply our approach to models that use state charts and class diagrams or models without state charts (that only consist of class diagrams). This makes the technique applicable to a larger number of industrial systems. We also propose to reduce the number of false positive dependencies by using a constraint solver. We implemented our approach as plug in for IBM Rational Software Architect and evaluated the tool on two case study systems including an industrial system from the smart card domain. The evaluation confirms that the approach is effective in identifying changes and reducing the effort needed to test a new version of the system. The results also show that the approach is efficient with execution times between 2-3 minutes for most cases. SeTGaM was also able to precisely identify all modification revealing tests.