Mining Test Repositories for Automatic Detection of UI Performance Regressions in Android Apps

Abstract

The reputation of a mobile app vendor is crucial to survive amongst the ever increasing competition. However this reputation largely depends on the quality of the apps, both functional and non-functional. One major non-functional requirement of mobile apps is to guarantee smooth UI interactions, since choppy scrolling or navigation caused by performance problems on a mobile device’s limited hardware resources, is highly annoying for end-users. The main research challenge of automatically identifying UI performance problems on mobile devices is that the performance of an app highly varies depending on its context—i.e., the hardware and software configurations on which it runs. This paper presents DUNE, an approach to automatically detect UI performance degradations in Android apps while taking into account context differences. First, DUNE builds an ensemble model of the UI performance metrics of an app from a repository of historical test runs that are known to be acceptable, for different configurations of context. Then, DUNE uses this model to flag UI performance deviations (regressions and optimizations) in new test runs. We empirically evaluate DUNE on real UI performance defects reported in two Android apps, and one manually injected defect in a third app. We demonstrate that this toolset can be successfully used to spot UI performance regressions at a fine granularity.