| Testing graphical user interfaces (GUIs) is difficult -- they're enormously complex, difficult to verify, and very likely to change. This thesis presents an overview of these challenges, followed by an overview of previous attempts to make GUI testing practical that categorizes each approach based on the challenge it addresses. A new tool for GUI testing, LEET, was created to apply automated rule-based verifications to manual exploratory test sessions that had been recorded. Five main research questions are identified: 1) Can rule-based exploratory testing be used to catch high-level, general bugs? 2) Can rule-based exploratory testing be used to catch low-level, specific bugs? 3) Can rules be reused in tests for different applications? 4) How often is it possible to use keyword-based testing on GUIs? 5) Is rule-based exploratory testing less effort than writing equivalent tests by using a capture/replay tool and inserting verifications manually?;A preliminary evaluation of rule-based exploratory testing with LEET is performed. This pilot study was able to provide some answers to the first four questions. It was found that rules can be used to detect both general and specific bugs, but rules for general bugs were easier to transfer between applications. It was also suggested that, despite the advantages that keyword-based testing presents to human testers, it interferes with the process of creating automated test oracles and the process of transferring rules between applications. The fourth question, however, was unanswerable based on the results of these pilot studies, so instead a set of recommendations for future work on the subject is presented. |
