(A) plugin-based framework for supporting GUI application testing

Abstract

Graphical user interfaces are pervasive in modern software systems, and to ensure their quality it is important to test them. Because manual creation and execution of test cases is expensive, researchers have attempted to create GUI testing techniques that automate the processes of generating and executing test cases. In GUI testing, however, execution results of testing techniques are largely affected by the configuration of the prototype tools such as the granularity of events and input parameters, which make it difficult to compare different classes of techniques fairly. To overcome this situation, we need to be able to compare GUI testing techniques under the same environment and the ability to capture events for equivalent GUI states. In this thesis, we develop a new GUI testing framework, {\sc GUITester}, based on a plug-in architecture, which enables people to extend each component of the framework, and provide a way for users to explore GUIs of an AUT (Application Under Test), configure events, and test the AUT. A plugin-based flexible architecture makes allows researchers to implement and assess various GUI testing techniques easily. To provide evidence that our framework is effective, we exploit the framework to study the relative strengths of two primary classes of automated GUI testing approaches, those based on static models and those based on dynamic event-extraction, showing that our framework can support developing and comparing various GUI testing techniques. Our study reveals several cost and effectiveness tradeoffs between the techniques, with implications for research and practice. Finally, we propose a new GUI testing technique that uses runtime feedback information and software analysis techniques to avoid redundant event executions, providing evidence that our framework can help researchers develop and implement techniques that require complicated processes.