I Can See Clearly Now: Clairvoyant Assertions for Deadlock Checking

Abstract

Static analysers are traditionally used to check various correctness properties of software. In the face of refactorings that can have adverse effects on correctness, developers need to analyse the code after refactoring and possibly revert their changes. Here, we take a different approach: we capture the effect of the Hide Delegate refactoring on programs in the ABS modelling language in terms of the base program, which allows us to predict the correctness of the refactored program. In particular, we focus on deadlock-detection. The actual check is encoded with the help of an additional data structure and assertions. Developers can then attempt to discharge assertions as vacuous with the help of a theorem prover such as KeY. On the one hand, this means that we do not require a specific static analyser nor theorem prover, but rather profit from the strength and advances of modern tool support. On the other hand, developers can choose to rely on existing tests to confirm that no assertion is triggered before executing the actual refactoring. Finally, we argue the correctness of our over-approximation.