Verifying annotated program families using symbolic game semantics

Abstract

Many software systems are today built as program families. They permit users to derive a custom program (variant) by selecting suitable configuration options at compile time according to their requirements. Many such program families are safety critical. However, most existing verification techniques are designed to work on the level of single programs. Their application to program families would require to verify each variant in isolation, in a brute force fashion. This approach does not scale in practice due to the (potentially) huge number of possible variants.In this paper, we propose an efficient game semantics based approach for verification of open program families, i.e. program families with undefined components (identifiers). We use symbolic representation of algorithmic game semantics, where symbolic values for inputs are used instead of concrete ones. In this way, we can compactly represent program families with infinite integers as so-called (finite state) featured symbolic automata. Specifically designed model checking algorithms are then employed to uniformly verify safety of all programs (variants) from a family at once using a single compact model and to pinpoint those programs that are unsafe (respectively, safe). We present a prototype tool implementing this approach, and we illustrate its practicality with several examples.