Using App Attributes to Improve Mobile Device Selection for Compatibility Testing

Mobile apps have been developed with the aim of attracting a large and diverse number of users. An impediment factor, especially for the Android platform, is a large number of hardware and software configurations available in the market, so app developers face the challenge of producing a highly compatible app. For compatibility testing, an app can be tested in a subset of devices that sufficiently covers the characteristics of devices adopted by users. Moreover, this subset needs to be extracted from up-to-date sources since new devices and APIs are frequently introduced in the market, while others are deprecated. This paper presents an always-updated, app-specific approach, called DeSeCT, for mobile Device Selection for Compatibility Testing. Our approach is divided into the following steps: first, it crawls the Web to obtain an updated list of devices; second, attributes of the app under test are used to filter unwanted mobile devices; and a multiobjective genetic algorithm is employed to tailor a feature-coverage optimized set of mobile devices. The results provide evidence that DeSeCT can produce a better selection of mobile devices for compatibility testing in comparison with the state-of-the-art. DeSeCT avoids the selection of approximately 15% to 18% of unwanted devices.