Regression Verification Using Impact Summaries

Abstract

Regression verification techniques are used to prove equivalence of closely related program versions. Existing regression verification techniques leverage the similarities between program versions to help improve analysis scalability by using abstraction and decomposition techniques. These techniques are sound but not complete. In this work, we propose an alternative technique to improve scalability of regression verification that leverages change impact information to partition program execution behaviors. Program behaviors in each version are partitioned into (a) behaviors impacted by the changes and (b) behaviors not impacted (unimpacted) by the changes. Our approach uses a combination of static analysis and symbolic execution to generate summaries of program behaviors impacted by the differences. We show in this work that checking equivalence of behaviors in two program versions reduces to checking equivalence of just the impacted behaviors. We prove that our approach is both sound and complete for sequential programs, with respect to the depth bound of symbolic execution; furthermore, our approach can be used with existing approaches to better leverage the similarities between program versions and improve analysis scalability. We evaluate our technique on a set of sequential C artifacts and present preliminary results.