Automated GUI Testing For Smartphone Applications Based On UML Activity Diagram

Along with the increasing popularity of smartphones, more and more people are relying smartphone applications to manage their daily life (e.g., online payment, instant messaging, etc.). The safety and correctness are becoming more and more important for smartphone applications. Therefore, it is required that sufficient testing should be conducted before the releasing of the smartphone application products. However, due to the tense time-to-market pressure in the fierce smartphone application market, it is hard to guarantee such testing adequacy within limited testing time. As a typical reactive system, smartphone applications interact with users via Graphical User Interface (GUI) intensively. Improper GUI designs or GUI errors can easily hurt the user experience, or even cause a catastrophe. Therefore, GUI testing plays an extremely important role in the design of smartphone applications. However, due to the lack of formal models to describe user behaviors, currently most GUI testing work is performed manually or randomly while the adequacy of the GUI testing cannot be evaluated. Moreover, smartphone application venders update their product very frequently. Before the shipping of an updated version, GUI tests need to be regenerated and GUI testing needs to be re-conducted, which is very expensive and time-consuming. Above all, GUI testing is becoming a major bottleneck in the design of smartphone applications.As a de facto standard modeling language, Unified Modeling Language (UML) is widely used in software design. Based on the UML activity diagrams, this thesis proposes an automated GUI testing method for smartphone application. This thesis makes following three major contributions:1) Extending the UML activity diagrams to support user behavior modeling for GUI testing. We propose a user behavior modeling method based on UML activity diagrams. By extending the Petri-Net like semantics of UML activity diagrams, our approach can enable the user behavior modeling of multi-view based smartphone application. Our approach also supports the automated extraction fo aser behavior as well as an objective coverage analysis using various proposed coverage criteria.2) Proposing an automated GUI testing framework for smartphone application based on UML activity diagrams, and developing a tool-chain based on this framework. The testing framework and tool chain support the user behavior modeling, automatic user behavior extraction, automatic generation and execution of test scripts and regression test scripts, automatic test coverage analysis and error diagnosis.3) Validating the efficacy of our approach using various large-scale case studies. To check the effectiveness of our automated GUI test method, we conducted experiments on multiple industrial smartphone applications using our tool-chain. The experiment results show that our automated GUI testing framework and tool-chain can significantly reduce the overall GUI testing time, while ensuring the test adequacy and finding the inconsistency between GUI designs and implementations.