PanicKiller

In this paper, we propose an infrastructure aiming at Rust panic bugs. We construct the first programming fixing dataset Panic4R containing 52 real-world panic bugs and patches, as well as 40 synthetic patches. Based on Panic4R, we further perform pattern mining to summarize the general fixing patterns with Rust-specific language features. Finally, we propose an APR tool PanicKiller for fixing the panic bugs in real-world and large-scale Rust programs. The experiment results show that PanicKiller is more effective in both fault localization and patch generation compared with the commercial LLM-based tools. Moreover, PanicKiller has successfully fixed 4 panic bugs, and they all have been confirmed by developers.