Error handling bugs are widespread in software, compromising its reliability. In C/C++ environments, where errors are often propagated to multiple functions through error codes. Addressing propagated error handling (PEH) bugs across all affected functions is challenging. Our analysis of the Linux Kernel identified 80 historical PEH bugs, with an average repair time of 67 days.

Existing approaches fall into two categories: those based on manually crafted templates and those leveraging the large language models (LLM). However, the performance of these methods is limited by the unique challenges posed by the PEH bug. First, PEH bug impacts numerous functions, making it difficult to locate and repair. Second, error-handling strategies is influenced by the severity and context of errors, and varies across projects, further challenging the selection of an appropriate handling.

This paper introduces EH-Fixer, a conversation-based, automated method for fixing PEH bugs. EH-Fixer employs LLM in a conversation style, utilizing information retrieval to address PEH bugs and validate the correctness of fixes. To assess the performance of EH-Fixer, we constructed a dataset containing 30 PEH bugs and evaluated EH-Fixer against two state-of-the-art approaches. Preliminary results indicate that EH-Fixer successfully fixed 19 more PEH bugs than existing approaches, proving its promising performance in PEH bug repair.