Research And Implementation Of Automatic Warning Validation Based On Triggering Path Constraint Solving

As an important software testing technology,source code defect detection technology has the characteristics of high defect detection efficiency,accurate fault location and high degree of automation.However,in the process of practical application of the technology,it is found that the detection results have problems such as large amount of warnings,high repetition rate and false positives,and it requires professional technicians to invest a lot of time to confirm the defects,which also leads to low efficiency.To solve this problem,the thesis proposes an automatic warning validation method based on the solution of trigger path constraint in the analysis of defect pattern classification.In the aspect of automatic warning validation,the specific research of this thesis is as follows:(1)There are multiple paths passing through the warning in the program,so choosing a path that can trigger the warning as soon as possible can greatly save the time cost of subsequent verification,therefore,the thesis proposes a path sorting method based on the value ranges of variables.When the value ranges of variables on the path is smaller,the test case generation method based on interval computation and is more likely to fail.The method uses the value ranges of variables as the weight to sort the paths and reduce the time cost caused by the failure of interval computation.(2)After solving the constraint set on the test path,the test case will cover the warnings in the actual execution process,but it can not be sure of triggering the warnings.The thesis proposes a test case generation method based on the trigger constraints.This method is based on the defect features of C language code,and combined with the context information related to the warning in the test path,and modifies the test path constraint set,so that the generated test cases can trigger the warning.(3)In order to verify whether the test cases generated by the above method can trigger the warning,and achieve the purpose of judging the authenticity of the warning,the thesis proposes a judgment method based on program instrumentation and debugging.Based on the characteristic information of source code defects,the method monitors the dynamic execution state of test cases and determines whether the test cases trigger warnings through program instrumentation technology and program debugging tools.This thesis analyzes the automatic validation results of 31 kinds of defect codes,and the results show that the above methods improve the automatic level of warning validation to a certain extent.