Structuring Erlang BEAM control flow

Abstract

As source code dependencies are usually stored in some precompiled executable representation like bytecode, static analysis frameworks for high-level languages have to be specifically adapted so they can meaningfully analyse these libraries too. This adaptation is not trivial, since compilation is in general not injective, the semantics of low-level instruction sets are often not specified adequately, and the structure of the high-level sources and the low-level target is considerably different. This is also true for the functional Erlang programming language and its assembly-like BEAM bytecode. In this paper, we present a structuring algorithm capable of recovering the Erlang syntax tree of functional branching expressions compiled to BEAM. The implementation of the presented algorithm is part of the RefactorErl static analyser framework. Therefore, the tool is able to represent the semantics of the BEAM programs with an Erlang syntax tree and perform further semantic analysis on it to discover the source dependencies.