Research On Fault Localization And Fault Comprehension Method Based On Weighted Software Behavior Graph Mining

Currently, computer software has been widely used in many fields, which arebecoming increasingly large and complex scale, people invest a lot of effort to ensurethe quality of the software, and proposed a lot of automation software fault localizationapproach to help developers quickly locate to the software failure.However, the result of the most of the fault localization approaches are suspicioussequence of statement, which in descending order. However, examining a suspiciousstatement in isolation is hard for developers to understand the fault, often they needmore context information associated with the error.To solve this problem, this paper completed the following tasks:First, based on an abstract syntax tree, instruments the program in of two differentparticle sizes which is statement level and logical expression levels, and then get thecorresponding program execution trace. Due to the current lack of a simple-to-use Clanguage syntax analysis tools, so this article using the C#language to implement a Clanguage syntax analysis tools named CParser.Secondly, to improve the accuracy of fault localization, this paper proposedweighted software behavior graph, using the execution frequency as weights, can betterhandle loops, recursive and other structures. According to the passed execution andfailed execution, were constructed corresponding weighted software behavior graph.Then, using the weighted software behavior graph mining algorithm based onbranch-and-bound search, to identify differences between passed and failed weightedsoftware behavior graph as a bug signature, the resulting Top-K bug signature to supportfault localization and understanding, and gives examples of the bug signature for faultunderstanding.Finally, the approach this paper presented was tested by Siemens Suit, analyze thefault localization accuracy, and for different types of fault, as well as the accuracyimprovement using logical expression level instrument. The results showed that theproposed approach has the higher fault localization accuracy, and better of redundantcode, missing codes and variable substitution errors, and errors will directly change theexecution path, and the logical expression instrument can improve the fault localizationaccuracy.