Automatic Fault Localization Using Statistics and Visualization: An Empirical Research Program

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

CCF-0541049 PI: Mary Jean Harrold GA Tech TITLE: Automatic Fault Localization Using Statistics and Visualization: An Empirical Research Program Software errors significantly impact software productivity and quality. One of the most expensive tasks required to reduce the number of faults is debugging, and locating faults is the most difficult and time-consuming component of debugging. Existing automated debugging techniques have limitations that prevent them from scaling to industrial systems, and there are few reports comparing existing techniques. This project will investigate fault-localization techniques that can be used in practice through several activities. First, the research will develop improved techniques that can more quickly guide the user to the faulty regions of the system and exploit programmer knowledge and guidance. Second, the research will design and conduct controlled experiments and case studies to evaluate cost-benefits of new and improved techniques, and to understand the factors contributing to those costs-benefits. Third, the research will assemble an infrastructure of programs, faulty versions, and test suites for future fault-localization approaches. Broader Impacts. The ubiquity of software and the high cost of faults, makes improvements in debugging techniques extremely important. This work will provide practical fault-localization techniques, along with empirical data on those techniques and empirical methodologies that will be made available to other researchers for use in future fault-localization research.