This doctoral research proposal focuses on a critical yet under-explored problem: how to automatically diagnose and test game compatibility layers (GCLs). GCLs are software systems composed of multiple subcomponents that allow unmodified games to run across different platforms by translating system calls and APIs in real time. In 2025, the number of Linux gaming users increased by 57%, reaching over four million monthly active users. As most commercial games are developed primarily for Windows, Linux users increasingly rely on GCLs to run Windows games on Linux, indicating a rapidly expanding user base for GCLs. However, localizing GCLs' bugs heavily relies on unexpected behavior reported by the gamers (who usually have limited background in programming), thus taking maintainers' extra effort to diagnose and fix bugs. It motivates this doctoral research, aiming to analyze the symptoms, and propose methods to identify bug-related subcomponents within GCLs and further test GCLs. Specifically, I design the following three work packages. 1) An empirical study on GCL's issue reports to build a taxonomy of the unique symptoms in GCLs and uncover relationships between symptoms and GCL's subcomponents; 2) A language model-based framework to help maintainers to accurately identify related subcomponents given user-reported symptoms; and 3) A symptom-guided testing method to uncover more bugs in GCLs and improve the system quality.

@inproceedings{Dai26,
  title = {Automated Diagnosis and Testing of Game Compatibility Layers},
  author = {Hanzheng Dai},
  booktitle = {FSE},
  year = {2026},
  doi = {10.1145/3803437.3804876}
}