Towards an Isabelle/HOL formalisation of core Erlang

Abstract

As part of broader work to improve the safety of Erlang systems, we are attempting to detect (and prevent) messages which remain forever unreceived in process' mailboxes using a mix of static and runtime analysis. We have formalised the communicating portion of Core Erlang using Isabelle/HOL, an interactive theorem prover. We can use the Isabelle toolchain to prove properties of our model, automatically prepare documents, and generate verified executable code in a variety of functional programming languages. We formally model a communicating fragment of Core Erlang in a language we call CoErl. After defining the evaluation of expressions, we model the process-local and concurrent semantics of the language using a labelled transition system. We also introduce the notion of mailbox traces which capture communication events during process execution. This is followed by some illustrative examples of the concurrent semantics. Although our CoErl model is a solid foundation for a full formalisation of Core Erlang, it currently lacks higher-order and recursive behaviour. Isabelle/HOL has proved practical for formalising and verifying several properties of CoErl and its trace system, while ongoing and future work focuses on bringing the language to feature parity with Core Erlang and Erlang/OTP.