Achieving High Code Coverage in Android UI Testing via Automated Widget Exercising

Abstract

Automated functional user interface (UI) testing of mobile apps is a challenging task due to the highly interactive nature of the app UIs, and thus it commonly fails to provide high code coverage. In this paper, we present DroidDEV, an automated UI exerciser which aims to achieve high code coverage for real-world Android apps. DroidDEV dynamically builds a finite UI flow graph, generates and injects UI-context-aware inputs to exercise all the widgets on all the UI screens through a best-first search path finding algorithm. We have evaluated DroidDEV on 20 real-world open-source Android apps and compared it with manual UI testing and two other prevalent automated approaches including MobiGUITAR and Monkey. On these subjects, DroidDEV reports code coverage similar to manual UI testing, and it outperforms MobiGUITAR and Monkey in terms of code coverage and exercising time.