This paper explains some software testing logic coverage criteria demand inputs that guarantee detection of a large set of fault types. One powerful such criterion, MUMCUT, is composed of three criteria, where each constituent criterion ensures the detection of specific fault types. In practice, the criteria may overlap in terms of fault types detected, thereby leading to numerous redundant tests, but due to the unfortunate fact that infeasible test requirements don't result in tests, all the constituent criteria are needed. The key insight of this paper is that analysis of the feasibility of the constituent criteria can be used to reduce test set size without sacrificing fault detection. In other words, expensive criteria can be reserved for use only when they are actually necessary. This paper introduces a new logic criterion, Minimal-MUMCUT, based on this insight. Given a predicate in minimal DNF, a determination is made of which constituent criteria are feasible at the level of individual literals and terms. This in turn determines which criteria are necessary, again at the level of individual literals and terms. This paper presents an empirical study using predicates in avionics software. The study found that Minimal-MUMCUT reduces test set size -- without sacrificing fault detection -- to as little as a few percent of the test set size needed if feasibility is not considered.

What is the flaws, strengths, weaknesses, and/or potential for future work of this statement above?