Change-aware model checking for evolving concurrent programs based on Program Dependence Net

Abstract

Concurrent software needs to be maintained over time, and the differences between continuous versions tend to be localized. The expense that simply reapplying standard model checking techniques to the new version as they evolve may be infeasible. The existing methods reuse partial state-space to reduce the scope. However, it is obviously costly to analyze on the explosive interleaving space of the evolving concurrent programs. The conservative change-impact analysis methods without considering the specific property and leveraging the verified result from the prior version often results in exploring redundant state-space irrelevant to this property. Moreover, the impact of the deleted elements needs to be analyzed on old version, and their impact needs to be mapped to new version, bringing some dispensable costs. In this paper, we propose a change-aware model checking method based on program dependence net (PDNet) for linear temporal logic (LTL). We first propose an incremental modeling method to construct a PDNet of new version by modification rules. Then, we propose a reuse checking algorithm to judge whether the verified result can be reused based on the PDNet slice. Finally, we implement change-aware model checking tool (DAMER) and validate the advantages of our methods.