Supporting automated system-level test scenario generation

Abstract

Software plays a significant role in society. It penetrates every field such as telecommunications, public administration, cooperation management, etc. In a software development life cycle (abbr., SDLC), software testing is a key phase. It accounts for a large part of software development costs. This is a consequence of testing being performed late in an improvised and impromptu way under the discretion of project managers. Continual testing as proposed by the extreme programming approach advocates that automated testing be performed in the early phases of the SDLC. Automation helps avoid disorderly and unsystematic progress of testing tasks and assignments. This thesis presents an approach for the generation of test cases from use cases---a form of requirements used in the early phases of the SDLC. We first needed a way to combine related use cases in order to infer system-level test cases spawning over several use cases. We developed an approach to infer use case sequential relations based on a comparison of pre-conditions and post-conditions. This approach offers the benefit of obtaining use case sequential relations without solely relying on the traditional UML use case relationships (i.e., include, extend and generalization). It helps to avoid the functional decomposition of use cases. We then propose an automated approach for die generation of test scenarios, a step toward complete and concrete test cases. Test scenarios are generated using depth-first traversal of control flow-based state machines obtained from use cases. The construction of these control flow-based state machines considers traditional UML use case relationships as well as inferred sequential relations. Depth-first traversal of control flow-based state machines is controlled by a coverage criterion inspired from traditional white-box code coverage.