From diagnosis to diagnosability: axiomatization, measurement and application

  • Authors:

  • Yves Le Traon;Farid Ouabdesselam;Chantal Robach;Benoit Baudry

  • Affiliations:

  • IRISA, Campus de Beaulieu, Université de Rennes I, 35042 Rennes Cedex, France;LSR-IMAG, BP 72, 38402 St. Martin d'Hères Cedex, France;LCIS-ESISAR, 50 Av. Barthelemy de Laffemas, BP 54, 26902 Valence Cedex 9, France;IRISA, Campus de Beaulieu, Université de Rennes 1, 35042 Rennes Cedex, France

  • Venue:
  • Journal of Systems and Software
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

Classical views on testing and their associated testing models are not dealing with the question of fault repairing but only focus on fault detection. Diagnosis consists of determining the nature of a detected fault, of locating it and hopefully repairing it. Correlatively, the only standardized quality factors implied in the detection/repair aspects of software engineering are testability and maintainability: those quality factors are misleading since they do not pinpoint this question of the location/repairing effort, that can be identified under the concept of diagnosability. This paper is thus concerned with diagnosability, its definition and the axiomatization of its expected behavior. The paper aims at: • introducing and analysing diagnosability as a significant and complementary dimension of software testability, • producing a high-level definition and axiomatization of a diagnosability measurement generic enough to be adapted to various software paradigms: this property-based approach serves as a measurement "specification", independent on the application context and thus reusable, • detailing a diagnosability measure dedicated to data-flow software and especially test strategies impact on diagnosis and testing effort (from measure implementation to case study), • illustrating the reuse of the high-level axiomatization to the specific question of measuring the impact of assertions (or contracts for a designed by contract OO system) on diagnosis effort and preciseness.Throughout the paper, the concepts are illustrated on a case study provided by an industrial partner. At last, the reusability of the axiomatization is illustrated by proposing a measure of the impact of assertions (or contracts in a design by contract approach) on global software diagnosability. Main lessons concern both the diagnosability significance as a quality factor and the interest of an axiomatization-based methodology for building trustable software measurement.