DroidBugs: An Android Benchmark for Automated Program Repair

TL;DR

DroidBugs introduces a new Android benchmark with 13 real bugs from 360 open-source projects, highlighting the challenges of automated program repair in mobile development.

Contribution

The paper presents DroidBugs, the first benchmark for mobile app repair based on real Android bugs, filling a gap in APR research for mobile platforms.

Findings

APR tools face significant challenges on mobile bugs

Fault localization strategies impact repair success

Mobile bugs are more complex than traditional benchmarks

Abstract

Automated Program Repair (APR) is an emerging research field. Many APR techniques, for different programming language and platforms, have been proposed and evaluated on several Benchmarks. However, for our best knowledge, there not exists a well-defined benchmark based on mobile projects, consequently, there is a gap to leverage APR methods for mobile development. Therefore, regarding the amount of Android Applications around the world, we present DroidBugs, an introductory benchmark based on the analyzes of 360 open projects for Android, each of them with more than 5,000 downloads. From five applications, DroidBugs contains 13 single-bugs classified by the type of test that exposed them. By using an APR tool, called Astor4Android, and two common Fault Localization strategy, it was observed how challenging is to find and fix mobile bugs.