On the Industrial Applicability of Visual GUI Testing

摘要: There is constant pressure on the software development industry to improve quality and deliver new innovative faster more efficiently. These demands affect all aspects of development, from requirements engineering testing. Today, much testing performed with costly, tedious error-prone manual practices, especially higher levels system abstraction such as in acceptance Increased test automation has been proposed a key solution help alleviate these test-related problems found increasing use. Automated tests typically consist scripts that can be executed give frequent feedback system's quality, ensure previously correct functionality not negatively affected by changes, also relieve resources e.g. human testers. However, evolves, changes require maintenance tests, which costly. Furthermore, most automated techniques approach lower level abstraction, low-level components functions, but their use for questioned. Test have automate interaction graphical user interface (GUI) system. For example, record replay tools used emulating interacting Existing literature pointed out several shortcomings techniques, example they are sensitive GUI layout code access source code, dependent specific platform or operating etc. In this thesis we evaluate if Visual Testing (VGT), novel technique combines image recognition, high-level, The strength VGT lies recognition makes robust instances where previous GUIbased had limitations. Yet, only sparsely applied academic body knowledge limited; there lack empirical evaluation experience reports industrial To fill gap conducted three case studies applying different projects at two Swedish companies. We an survey participants multiple provide evidence successfully projects, detail challenges, limitations need overcome widespread adoption. Future work should focus evaluating long-term costs scripts, since our present initial manageable relation benefits technique.