Mon 20 - Fri 24 October 2014 Portland, Oregon, United States

Mobile app markets have lowered the barrier to market entry for software producers. As a consequence, an increasing number of independent app developers offer their products, and recent platforms such as the MIT App Inventor and Microsoft’s TouchDevelop enable even lay programmers to develop apps and distribute them in app markets.

A major challenge in this distribution model is to ensure the quality of apps. Besides the usual sources of software errors, mobile apps are susceptible to errors caused by the non-determinism of an event-based execution model, a volatile environment, diverse hardware, and others. Many of these errors are difficult to detect during testing, especially for independent app developers, who are not supported by test teams and elaborate test infrastructures.

To address this problem, we propose a static program analysis that captures the specifics of mobile apps and is efficient enough to provide feedback during the development process. We present our analysis for TouchDevelop, but most aspects apply also to other platforms. Experiments involving 51,456 published TouchDevelop scripts show that our analysis analyzes 98% of the scripts in under a minute, and five seconds on average. Manual inspection of the analysis results for a selection of all scripts shows that most of the alarms are real errors.