Basing test coverage on a formalization of test hypotheses

Abstract

This paper defines a characterization of protocol confonnance test coverage based on test hypotheses. All test selection methods and coverage computation make use of test hypotheses in one way or another. Test hypotheses are assumptions made on the implementation, which justify the verdict of confonnity provided by testing; thus they are an important part of the coverage. We propose a model of these hypotheses based on functions on automata, enabling a definition of coverage based on test hypotheses, which we call TH-based coverage.

Test Selection, Trace Distance and Heuristics

“…In particular, these are the hypothesis theory developed by [6] and the trace distance theory of [1,2]. The hypothesis theory embodies the trace distance theory (see [6]), but the nice thing about trace distance theory is that it gives a measure for the degree to which a reduced set of traces approximates the original one. So we chose an approach which combines these two theories.…”

“…So we chose an approach which combines these two theories. In our view, first the heuristics (test hypotheses in the theory of [6]) are to be defined. After that, based on these heuristics a trace distance is built.…”

Protocol-Inspired Hardware Testing

“…To progress further will involve defining a suitable coverage function or exploiting some existing selection methodologies (e.g. [1,3]). Applying the method presented in this paper to higher-level specification of digital circuits will also be interesting.…”

Exploiting Symmetry in Protocol Testing

“…the uni formity hypothesis in [15,6]), it is perhaps more appropriate to speak of gray box testing. For the specification FSM it will generally be possible to verify that a particular relation is a symmetry on the system, but for the black box implementation one has to assume that this is the case.…”