Research And Implementation Of Bug Fixing Recommendation Technology

Bugs in software systems are inevitable.Bug fixing is time-consuming and labor-intensive With the explosive growth of the scale and complexity of software products,the legacy problems in the software continue to grow up which makes the code more and more numerous and complicated.Therefore,when experienced developers face bugs,they not only spend a lot of time on understanding the code,but also feel difficult to fix the bug.Existing bug fixing usually employ previous similar change operations,as a result,bugs may not be fixed correctly,even makes bug fixing more difficultIn this thesis,we analyze the commonly used fine-grained change operations,and propose fix templates for bug fixing based on these fine-grained change operations.Then,a bug fixing recommendation method is proposed to help developers improve the efficiency of bug fixing.The work mainly includes the following(1)We analyze bug reports and commit reports to explore the correlation of software bugs and fine-grained change operations.We conduct an empirical study to exploit software bugs and fine-grained change operations.The results show that the fine-grained change operations have no obvious relation with the change operations.In addition,most of the bug fixing procedure uses common fine-grained change operations.(2)We analyze the fine-grained change operations and build coarser-grained fix templates through the differences between the buggy version and the fixed version in the source code file We define some templates based on fine-grained change operations to improve the effectiveness of bug fixing,and empirically study the use frequency of the defined templates.The results show that these templates can be applied to fix bugs.(3)We utilize deep learning technology,combining with tree-based convolutional neural network to provide suggestions for bug fixing.We utilize the technology based on information retrieval for bug localization and provide templates and suggestions for bug fixing.The empirical study shows that the accuracy of the templates,moreover,the feedback from five participants indicate that the recommended fix templates are effective for bug fixing.