Incremental spectrum cloning algorithm for optimization of spectrum-based fault localization

Spectrum-based Fault Localization has emerged as a cost effective method to locate faulty code in software during the debugging process. Recent studies have shown that spectra (execution profiles) cloning for fail test cases can effectively improve the performance of certain spectrum-based Fault Localization ranking metrics. However, the amount of cloning required to optimize the performance varies from one program to another. This is because the structure and fault content of each program is unique. Furthermore, both insufficient cloning and excessive cloning may result in performance deterioration. In this paper, we propose an incremental spectra cloning algorithm that will clone the spectra of fail test cases up to an amount that optimizes the fault localization performance of the ranking metrics used for the software debugged. Experiments conducted on faulty versions of real life software have shown that the proposed algorithm could optimize the cost effectiveness of spectrum-based fault localization by reducing the number lines examined to locate the faulty code.