PAFL - Enhancing Fault Localizers by Leveraging Project-Specific Fault Patterns

Explore a 15-minute conference presentation introducing PAFL (Project-Aware Fault Localization), an innovative technique that enhances existing fault localization methods by leveraging project-specific fault patterns. Learn how researchers from Korea University and Samsung Electronics address the limitation of universal localization strategies that fail to consider recurring fault patterns unique to each software project. Discover the two key innovations behind PAFL: a domain-specific fault pattern-description language called "crossword" that represents various fault patterns and their impact on statement suspiciousness scores, and an algorithm that synthesizes these crosswords from past buggy versions of projects. Understand how PAFL mines fault patterns from historical buggy versions and uses them to update suspiciousness scores computed by baseline fault localizers. Examine the comprehensive evaluation results demonstrating PAFL's effectiveness across seven baseline fault localizers and 12 real-world C/C++ and Python projects, showing robust and efficient improvements in fault localization performance. Access supplementary materials including the full research article, reusable artifacts, and reproduced results that support the findings presented at the OOPSLA 2025 conference.