Mechanical verification of object oriented programs is a central software engineering problem. Any successful solution to the problem should strike a delicate compromise between amenability to automation and several software engineering factors, such as the form and ease of specifications, demands on software developers to provide invariants and hints, development and use of relevant mathematical theories, and software design. The objective of this paper is to illustrate these issues through our experiments towards push-button verification of an imperative object-based code in a modular fashion. 1.

Components that encapsulate maps are among the most fundamental for the development of modern software. The concept of associating keys with values is important for a wide range of applications and it can accommodate a broad variety of implementations with diverse performance profiles. The foundational-yet-sophisticated nature of this concept makes it an ideal benchmark for software verification efforts. A tension between modular reasoning and the usefulness of a map can be observed, in particular, in systems where the keys may be reference types. This paper elaborates on the criteria for modular verification of robust map components and their clients and it presents a survey of existing attempts to verify map components. 1.

Abstract not found

www.vscomp.org Abstract. We, the organizers and participants, report our experiences

Abstract. Anecdotal experience constructing proofs of correctness of code built from reusable software components reveals that they tend to be relatively trivial bookkeeping exercises: they rarely require a substantive mathematical deduction. A careful empirical analysis of hundreds of verification conditions (VCs) for a library of component-client code shows the level of sophistication each proof requires, and suggests how to use the results to characterize a notion of mathematical “obviousness.” 1