Mobile apps can be executed with an extremely large set of partially unpredictable configurations. Indeed, they can be executed on an unbounded combination of devices, operating systems, settings, and user preferences since apps may also interact with other apps or devices that were not even available when they were released. This results in a virtually infinite set of configurations that might be responsible for unexpected behaviors which can be validated in-house only to a negligible extent.
To address this challenge, this paper discusses the application of in-vivo testing to mobile apps. The main idea is to run test cases in the field, where we exploit the intrinsic heterogeneity and variety of the end-user environment to dramatically increase the range of validated configurations. Actually, the many devices available in-the-field generate a naturally distributed and highly scalable environment that can be exploited to timely validate many configurations as soon as they are observed.
PDF version of the paper.