Chinese Journal of Computers (计算机学报) 2013/36:12 PP.2406-2419
For predicates based statistical debugging, the strength of gathering and utilizing execution information of predicates has a strong influence on the accuracy of software fault localization. This paper mainly focuses on two questions: (1) Whether the accuracy of fault localization can be improved by enriching the predicate execution information? (2) How to evaluate the relationship between the accuracy of fault localization and the augment of predicate execution information? Based on the researches related with the two questions, a new statistical method, called Pesla, is proposed, which introduces the predicate execution-sequences to enrich the predicate execution information. Experimental results clearly demonstrate that more predicate execution information can indeed improve the accuracy of the fault localization. Furthermore, if the predicate execution information of target program meets the algorithm's demand, the higher accuracy of the fault localization will be achieved as the more predicate execution information is used.