A Topology-Based Model for Estimating the Diagnostic Efficiency of Statistics-Based Approaches

Abstract

Spectrum-based fault localization (SFL) and dynamic code coverage (DCC) are two statistics-based fault localization techniques used in software fault diagnosis. Due to their nature (statistical analysis of the coverage information), the best technique of the two depends greatly on the system under test code structure and size. We propose a lightweight, topology-based analysis to quickly estimate the project under test coverage matrix when executed, based on the source code structure. This analysis will choose which fault localization technique to use by creating an hierarchical model of the system. To validate our proposed approach, an empirical evaluation was performed, injecting faults in six real-world software projects. We have demonstrated that using the topology-based analysis to choose the best fault localization technique provides a better execution time performance on average (23%) than using DCC (9%), when comparing to SFL.