Software Testing: Concepts and Operations

14Metrics for Software Testing

The discipline of software metrics is concerned with defining quality attributes of software products and sizing them up with quantitative functions; such quantitative functions can then be used to assess the quality attributes of interest and thereby provide a basis for quantitative analysis and decision-making. Metrics have been used to capture such quality attributes as reliability, maintainability, modularity, verifiability, complexity, and so on. In this chapter, we focus our attention on software metrics that are relevant to testing and test-related attributes. We classify the proposed metrics into six categories that espouse the lifecycle of a system failure, namely:

• Fault proneness—the density of faults within the source code.
• Fault detectability—the ease of detecting faults in the source code.
• Error detectability—the ease of detecting errors in the state of a program in execution.
• Error maskability—the likelihood that an error that arises during the execution of a program gets masked before it causes a program failure.
• Failure avoidance—the ease of detecting and avoiding program failure with respect to its intended function.
• Failure tolerance—the likelihood that a program satisfies its specification despite failing to compute its intended function.

We review these metrics in turn, below. But first, we briefly introduce some elements of information theory, which we use in subsequent sections of this chapter. Given a random variable X on ...