Abstraction based automated test generation from formal tabular requirements specifications

  • Authors:

  • Renzo Degiovanni;Pablo Ponzio;Nazareno Aguirre;Marcelo Frias

  • Affiliations:

  • Departamento de Computación, FCEFQyN, Universidad Nacional de Río Cuarto, Córdoba, Argentina;Departamento de Computación, FCEFQyN, Universidad Nacional de Río Cuarto, Córdoba, Argentina;Departamento de Computación, FCEFQyN, Universidad Nacional de Río Cuarto, Córdoba, Argentina;Departamento de Ingeniería Informática, Instituto Tecnológico Buenos Aires and CONICET, Buenos Aires, Argentina

  • Venue:
  • TAP'11 Proceedings of the 5th international conference on Tests and proofs
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

We propose an automated approach for generating tests from formal tabular requirements specifications, such as SCR specifications. The technique is based on counterexample guided abstraction refinement and the use of SMT solving. Moreover, in order to effectively perform automated test generation, we take advantage of particular characteristics of tabular requirements descriptions to aid the abstraction and abstraction refinement processes. The exploited characteristics are, most notably, the organisation of the requirements specification in modes, which is used to build an initial abstraction, and the execution model of tabular specifications, which is directed by changes observed in environment variables and is exploited for modularising the transition relation associated with tables, simplifying the calculation of abstractions. These characteristics enable us to effectively perform automated test generation achieving good levels of coverage for different criteria relevant to this context. We compare our approach with a standard abstraction analysis, showing the benefits that exploiting the mentioned characteristics of tables provide. We also compare the approach with model checking based generation, using several model checking tools. Our experiments show that the presented approach is able to generate test cases from models whose complexity, with respect to the sizes of variables and data domains, cannot be coped with well by the model checkers we used.