Bridging the Gap Between Debugging Research and Practice
Date:
Abstract: In this talk, I will provide an overview of our work on Automated Debugging over the last (seven) years.
Researchers have proposed hundreds of Automated Fault Localization (AFL) techniques. However, these AFL tools are hardly adopted in practice due to the gap between debugging research and software practice. To bridge this gap, our work examines the challenges of AFL adoption in practice. This talk will focus on our empirical studies investigating the practicality and effectiveness of AFL in software practice from a human-centric and utility perspective. I will outline our work studying human factors and other threats to the validity of debugging research.
Firstly, I will present, DBGBench, our work examining how professional software engineers debug computer programs. Then, I will present our study evaluating the effectiveness of state-of-the-art AFL technique on real-world programs and bugs (EMSE21). Finally, I will conclude with our recent work investigating the impact of experimental assumptions (made by the most popular AFL techniques in the literature) in debugging practice. Our experiments include a user study examining the soundness, severity and usefulness of three major experimental assumptions in practice (ICSE23).