| Description: |
Thesis (PhD)--Stellenbosch University, 2026. ; Callaghan, D. S. 2026. Spectrum-Based Fault Localization for Multiple Faults. Unpublished doctoral dissertation. Stellenbosch: Stellenbosch University [online]. Available: https://scholar.sun.ac.za/items/dafeac1b-9416-49f7-a6ab-3c10081a2b64 ; Debugging is the process of identifying and removing faults from a software system. The task is traditionally performed largely manually and is thus a costly part of the software development life-cycle. Many automated approaches have been proposed to alleviate the cost of debugging, particularly for the task of fault localization, with Spectrum-Based Fault Localization (SBFL) being the most widely used of these techniques for its lightweight and scalable design. However, while SBFL has been shown to perform well for programs containing only a single fault, its accuracy decays for an increasing number of faults. We therefore develop FLITSR (Fault Localization by Iterative Test Suite Reduction), a novel, purely spectrum-based approach that improves the localization of a given SBFL base metric specifically in the presence of multiple faults. FLITSR returns a set of highly suspicious program elements (called a basis), where the execution of each failing test involves at least one basis element, considered as the cause of the failure. We further develop multiple extensions to the standard FLITSR algorithm, allowing FLITSR to return multiple, ordered bases through the FLITSR* multi-round and internal basis ranking extensions. We lastly improve localization effectiveness through increasing the quality of the underlying test suite in a third FLITSR extension, in which large test cases exposing multiple faults are split up in a process of test case purification. Our evaluations show that FLITSR, along with its extensions, consistently and significantly outperform existing localization metrics and methods, including those designed to handle multiple faults. FLITSR not only improves the localization of the first fault in the ranking, ... |