Analyzing Model Dependencies for Rule-based Regression Test Selection

Abstract

Unintended side effects during changes of software demand for a precise test case selection to achieve both confidence and minimal effort for testing. Identifying the change related test cases requires an impact analysis across different views, models, and tests. Model-based regression testing aims to provide this analysis earlier in the software development cycle and thus enables an early estimation of test effort. In this paper, we present an approach for model-based regression testing of business processes. Our approach analyzes change types and dependency relations between different models such as Business Process Modeling Notation (BPMN), Unified Modeling Language (UML), and UML Testing Profile (UTP) models. We developed a set of impact rules to forecast the impact of those changes on the test models prior to their implementation. We discuss the implementation of our impact rules inside a prototype tool EMFTrace. The approach has been evaluated in a project for business processes on mobile devices.