Testability analysis in communicating systems

Abstract

Efficiency of communication software depends on the testability of the Implementation Under Test, and consequently testability must be analyzed at the design step. Design For Testability increases in complexity when the concern is on Testing Through an Environment. Early work by the first author demonstrated the benefits and potential of Refusal Graphs as a formalism to analyze testability. The formal model of the system to be implemented possibly contains internal loops that create divergence and negatively influence the analysis's accuracy. In practice, such a problem arises in situations of Testing Through an Environment where (i) the global specification is obtained by composition of elementary behaviors, and (ii) part of the communications cannot be observed. In the paper, enhanced refusal graphs with divergence are introduced to handle the divergence problem for testability analysis including systems whose communications remain partly unobservable. Two notions, Conformance Through an Environment and Degradation of Testability, are formally introduced for Testing Through an Environment. The testability analysis algorithm has been implemented in a prototype written in C.