A Precise Framework for Source-Level Control-Flow Analysis

Abstract

This paper presents INTRACFG, a declarative and language-independent framework for constructing precise intraprocedural control-flow graphs (CFGs) based on the reference attribute grammar system JastAdd. Unlike most other frameworks, which build CFGs on an Intermediate Representation level, e.g., bytecode, our approach superimposes the CFGs on the Abstract Syntax Tree, enabling accurate client analysis. Moreover, INTRACFG overcomes expressivity limitations of an earlier RAG-based framework, allowing the construction of AST-Unrestricted CFGs: CFGs whose shape is not confined to the AST structure. We evaluate the expressivity of INTRACFG with INTRAJ, an application of INTRACFG to Java 7, by comparing two data flow analyses built on top of INTRAJ against tools from academia and from the industry. The results demonstrate that INTRAJ is effective at building precise and efficient CFGs and enables analyses with competitive performance.