Abstract

Abstract

Model Based Test Generation and Optimization

Mohamed Mussa A. Mussa, Ph.D.
Concordia University, 2015

Software testing is an essential activity in the software engineering process. It is used to enhance the quality of the software products throughout the software development process. It inspects different aspects of the software quality such as correctness, performance and usability. Furthermore, software testing consumes about 50% of the software development efforts. Software products go through several testing levels. The main ones are unit-level testing, component-level testing, integration-level testing, system-level testing and acceptance-level testing. Each testing level involves a sequence of tasks such as planning, modeling, execution and evaluation.
Plenty of systematic test generation approaches have been developed using different languages and notations. The majority of these approaches target a specific testing-level. However, only little effort has been directed toward systematic transition among testing-levels. Considering the incompatibility between these approaches, tailored compatibility-tools are required between the testing levels. Furthermore, several test models are usually generated to evaluate the implementation at each testing level. Unfortunately, there is redundancy among these models. Efficient reuse of these test models represents a significant challenge. On the other hand, the growing attention to the model driven methodologies bonds the development and the testing activities. However, research is still required to link the testing levels.
In this PhD thesis, we propose a model based testing framework that enables reusability and collaboration across the testing levels. In this framework, we propose test generation and test optimization approaches that at each level consider artifacts generated in preceding testing levels. More precisely, we propose an approach for the generation of integration test models starting from component test models, and another approach for the optimization of the acceptance test model using the integration test models. To conduct our research in rigorous settings, we base our framework on standard notations that are widely adopted for software development and testing, namely Unified Modeling Language (UML). In our first approach, component test cases are examined to locate and select the ones that include an interaction among the integrated components. The selected test cases are merged to generate integration test cases, which tackles the theoretical research issue of merging test cases. Furthermore, the generated test cases are mapped against each other to remove potential redundancies. For the second approach, acceptance test optimization, integration test models are compared to the acceptance test model in order to remove test cases that have already been exercised during the integration-level testing. However, not all integration test cases are suitable for the comparison. Integration test cases have to be examined to ensure that they do not include test stubs for system components.
We have developed two approaches and implemented the corresponding prototypes in order to demonstrate the effectiveness of our work. The first prototype implements the integration test generation approach. It accepts component test models and generates integration test models. The second prototype implements the acceptance test optimization approach. It accepts integration test models along with the acceptance test model and generates an optimized acceptance test model.