Open Access

Reinhard Pröll

• The constant improvement of the available computing power nowadays enables the accomplishment of more and more complex tasks. The resulting implicit increase in the complexity of hardware and software solutions for realizing the desired functionality requires a constant improvement of the development methods used. On the one hand over the last decades the percentage of agile development practices, as well as testdriven development increases. On the other hand, this trend results in the need to reduce the complexity with suitable methods. At this point, the concept of abstraction comes into play, which manifests itself in model-based approaches such as MDSD or MBT. The thesis is motivated by the fact that the earliest possible detection and elimination of faults has a significant influence on product costs. Therefore, a holistic approach is developed in the context of model-driven development, which allows applying testing already in early phases and especially on the model artifacts,The constant improvement of the available computing power nowadays enables the accomplishment of more and more complex tasks. The resulting implicit increase in the complexity of hardware and software solutions for realizing the desired functionality requires a constant improvement of the development methods used. On the one hand over the last decades the percentage of agile development practices, as well as testdriven development increases. On the other hand, this trend results in the need to reduce the complexity with suitable methods. At this point, the concept of abstraction comes into play, which manifests itself in model-based approaches such as MDSD or MBT. The thesis is motivated by the fact that the earliest possible detection and elimination of faults has a significant influence on product costs. Therefore, a holistic approach is developed in the context of model-driven development, which allows applying testing already in early phases and especially on the model artifacts, i.e. it provides a shift left of the testing activities. To comprehensively address the complexity problem, a modelcentric software testing life cycle is developed that maps the process steps and artifacts of classical testing to the model-level. Therefore, the conceptual basis is first created by putting the available model artifacts of all domains into context. In particular, structural mappings are specified across the included domain-specific model artifacts to establish a sufficient basis for all the process steps of the life cycle. Besides, a flexible metamodel including operational semantics is developed, which enables experts to carry out an abstract test execution on the modellevel. Based on this, approaches for test case management, automated test case generation, evaluation of test cases, and quality verification of test cases are developed. In the context of test case management, a mechanism is realized that enables the selection, prioritization, and reduction of Test Model artifacts usable for test case generation. I.e. a targeted set of test cases is generated satisfying quality criteria like coverage at the model-level. These quality requirements are accomplished by using a mutation-based analysis of the identified test cases, which builds on the model basis. As the last step of the model-centered software testing life cycle two approaches are presented, allowing an abstract execution of the test cases in the model context through structural analysis and a form of model interpretation concerning data flow information. All the approaches for accomplishing the problem are placed in the context of related work, as well as examined for their feasibility by of a prototypical implementation within the Architecture And Analysis Framework. Subsequently, the described approaches and their concepts are evaluated by qualitative as well as quantitative evaluation. Moreover, case studies show the practical applicability of the approach.…