Fault Localization Technology And Its Practice Based On KBC And Noise Reduction

Software is buggy. The first factor for fixing bug is how to locate errors quickly. which often takes the most human and time resources in the software development cycle. Thus any technology that can reduce the cost is very attractive.Existing fault localization technologies in the field of locating single fault have been carried out, but its accuracy is not enough; and it becomes even worse when the single is extended to the multiple.This thesis improves two key factors of fault localization technologies and it also proposes a new fault indicator KBC and a law of noise reduction with similarity coefficients. KBC is the key fragment dug out of a program's control flow graph, on which execution sequence has richer context information. And the formal law of noise reduction is able to reduce noise for most similarity coefficients. Through these two key technologies, a new framework of fault localization, Minus-FKBC is proposed. The experiments show that, under this framework, the efficiency of single-fault localization can be increased by 10%. With the improvement of input-output optimization (that is, division of test cases and the sequence of ranking list), this framework can be applied to localize multi-fault effectively, thus its localizing efficiency can be upgraded greatly.In order to explore in the research domain about fault localization and its application in production, this thesis describes two applications, one online platform FLOJ. and for the Java project development. Eclipse plug-ins Lunuo. FLOJ can create a new technology of fault localization and assess its performance in hours, which normally cost one research team half a year or more time. Lunuo is the first tool set used in an IDE in production about fault localization. The application of Lunuo in school shows that the use of these tools will be very helpful in teaching.