Debug Support for Model-Based GUI Testing

The fact that model-based testing has not yet attained a high rate of adoption in industry can in part be attributed to the perceived difficulty of debugging long error traces often produced by the online version of this technology. Given the extensive manual labor commonly involved in the debugging phase, automating parts of this process could yield considerable productivity benefits. This paper presents viable debugging strategies applicable in model-based graphical user interface testing, from which two methods were refined and experimented with. The first is based on superimposing log-derived, synchronized subtitles on recorded test run footage, while the second addresses error trace shortening. The results obtained from applying these methods in real-life case studies demonstrate the practical utility of these methods.

1. Introduction

It is commonly recognized that testing and debugging consume a considerable amount of time and resources in software projects due to the great manual work needed in these tasks. Especially black-box testing and the fault analysis and localization related to this type of testing are highly context-sensitive activities, which accounts for the limited number of approaches working across different contexts [1]. However, if we could develop techniques to automate parts of the laborious fault analysis and localization process at least in some widely used context, such as graphical user interface (GUI) testing, this could potentially have significant benefits in the productivity of software production at least in that specific context.