FURI | Spring 2026

Extending PREP: Program Repair Enhancement via Preprocessing

Search-based Automated Program Repair (APR) tools reduce the burden of fixing bugs in software, but mismatches between tool expectations and real code often prevent identification of viable fixes. Program Repair Enhancement via Reprocessing (PREP) addresses this by applying static transformations to better expose program structures expected by APR tools. This project duplicates prior evaluations of the APR tool, Prophet, on the Codeflaws dataset with and without PREP and investigates how transformations impact repair quantity and alignment with human-written fixes. Finally, it proposes a transformation targeting Prophet’s repair performance on defects involving variables, operands, or constants, to ultimately improve APR effectiveness.