Linking Semantic Models to Support CSP () B Consistency Checking

  • Authors:
  • Neil Evans;Helen Treharne

  • Affiliations:
  • School of Electronics and Computer Science, University of Southampton;Department of Computing, University of Surrey

  • Venue:
  • Electronic Notes in Theoretical Computer Science (ENTCS)
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

Consistency checking in the CSP @? B approach verifies that an individual controller process, defined using a sequential non-divergent subset of CSP, never calls a B operation outside its precondition. Previously this was done by preprocessing the CSP process to perform a weakest precondition semantics proof. An embedding of the CSP traces model already exists in the PVS theorem prover, which makes use of 'uniform properties' to define valid traces. By including a state model we can extend the notion of uniform properties to define consistency. In this paper we give a framework which uses these semantic embeddings to eliminate the need for preprocessing. CSP @? B supports compositional verification, and the added benefit of this framework is that rely/guarantee style decomposition emerges naturally during a proof of consistency.