Applying Spectrum-Based Fault Localization to Android Applications

The pressing demand for high-quality mobile applications has a major influence on Software Engineering practices, such as testing and debugging. The variety of mobile platforms is permeated with different resources related to communication capabilities, sensors, and user-controlled options. As a result, applications may exhibit unexpected behaviors and resource interactions can introduce failures that manifest themselves in specific resource combinations. These failures can affect the quality of mobile applications and degrade the user experience. To reduce human effort of manual debugging, several techniques have been proposed and developed aiming to partially or fully automate fault localization. Fault localization techniques, such as Spectrum-based Fault Localization (SBFL), identify suspicious faulty program elements related to a software failure. However, we still lack empirical knowledge about the applicability of fault localization techniques in the context of mobile applications, specifically considering resource interaction failures. To address this problem, this paper evaluates the use of SBFL aiming to locate faults in 8 Android applications and verify the sensitivity of SBFL to variations in resource interactions. We rely on mutation testing to simulate faults and on the Ochiai coefficient as an indicator of the suspicious faulty code. Our results indicate that SBFL is able to rank more than 75% of the faulty code in 6 out of 8 applications. We also observed that the ranking of suspicious code varies depending on the combination of enabled resources (e.g., Wi-Fi and Location) in the mobile applications.