Automated Program Repair

CS 439 tackles the question of whether software can fix its own bugs: given a failing program, how do you automatically locate the defect and synthesize a patch that actually works? You'll move through fault localization, then the major repair paradigms — heuristic, template, constraint-based, and learning-based synthesis — culminating in a project where you build and benchmark a repair tool against real buggy programs. It sits at the intersection of software engineering, program analysis, and increasingly ML, and assumes you're comfortable enough with compilers and testing to reason about programs as data rather than just write them.