Today, me, Ben Hall (Cambridge) and Samin Ishtiaq (Microsoft Research) submitted a paper to CAV 2015, the 27th International Conference on Computer Aided Verification, to be held in San Francisco in July. CAV is dedicated to the advancement of the theory and practice of computer-aided formal analysis methods for hardware and software systems; the conference covers the spectrum from theoretical results to concrete applications, with an emphasis on practical verification tools and the algorithms and techniques that are needed for their implementation.
In this paper we build upon our recent work, highlighting a number of key issues relating to reproducibility and how they impact on the CAV (and wider computer science) research community, proposing a new model and workflow to encourage, enable and enforce reproducibility in future instances of CAV. We applaud the CAV Artifact Evaluation process, but we need to do more. You can download our arXiv pre-print; the abstract is as follows:
How many times have you tried to re-implement a past CAV tool paper, and failed?
Reliably reproducing published scientific discoveries has been acknowledged as a barrier to scientific progress for some time but there remains only a small subset of software available to support the specific needs of the research community (i.e. beyond generic tools such as source code repositories). In this paper we propose an infrastructure for enabling reproducibility in our community, by automating the build, unit testing and benchmarking of research software.
(also see: GitHub repo)