Enhance Fault Localization Using a 3D Surface Representation

Debugging is a difficult and time-consuming task in software engineering. To locate faults in programs, a statistical fault localization technique makes use of program execution statistics and employs a suspiciousness function to assess the relation between program elements and faults. In this paper, we develop a novel localization technique by using a 3D surface to visualize previous suspiciousness functions and using fault patterns to enhance such a 3D surface. By clustering realistic faults, we determine various fault patterns and use 3D points to represent them. We employ spline method to construct a 3D surface from those 3D points and build our suspiciousness function. Empirical evaluation on a common data set, Siemens suite, shows that the result of our technique is more effective than four existing representative such techniques.