Redefinition of Fault Classes in Logic Expressions

Abstract

Fault-based testing selects test cases to detect hypothesized faults. In logic expression testing, many fault classes have been defined by researchers based on the syntax of the expressions. Due to the syntactic nature of the logic expressions, some fault classes may exist in one form (say, disjunctive normal form - DNF) of the logic expressions but not in other forms (say, general form). As a result, different fault-based testing techniques have been developed for different types of logic expressions and these techniques have different fault detecting capabilities. For example, some have high detecting power in DNF but low detecting power in the general form. Another complication arises when software developers decide which forms of logic expressions should be used in the first place. Should software developers use the general form for flexibility but compromise that with fewer fault classes and less fault detection? Or should they use DNF for more fault classes and, hence, better fault detection (because software developers have more ''hypothesized faulty'' scenarios to test) but sacrificing the generality of the expressions. In this paper, we propose a set of uniform definitions of fault classes such that they can be applied irrespective of the syntactic nature of the logic expressions, to produce consistent fault-based testing techniques and fault detection capabilities.