Reproducing Context-sensitive Crashes of Mobile Apps using Crowdsourced Monitoring

While the number of mobile apps published by app stores keeps on increasing, the quality of these apps varies widely. Unfortunately, for many apps, end-users continue experiencing bugs and crashes once installed on their mobile device. Crashes are annoying for end-users, but they de fi nitely are for app developers who need to reproduce the crashes as fast as possible before fi nding the root cause of the reported issues. Given the heterogeneity in hardware, mobile platform releases, and types of users, the reproduction step currently is one of the major challenges for app developers. This paper introduces M o T i F, a crowdsourced approach to support app developers in automatically reproducing contextsensitive crashes faced by end-users in the wild. In particular, by analyzing recurrent patterns in crash data, the shortest sequence of events reproducing a crash is derived, and turned into a test suite. We evaluate M o T i F on concrete crashes that were crowdsourced or randomly generated on 5 Android apps, showing that M o T i F can reproduce existing crashes e ff ectively.