Miscompilation bugs in compilers are extremely hard to find and costly to fix. While a full formal correctness proof remain out of reach for most production compilers, the tooling and techniques for finding miscompilation bugs has become much easier to use. In this talk I’ll describe the approaches the PyPy project has used to find and fix miscompilation bugs in its JIT compiler, and draw out practical lessons applicable to other compiler projects. I’ll discuss how testing, fuzzing, translation validation, and SMT-based approaches can be combined to catch bugs in different parts of a compiler.