Towards a Systematic Process to Produce High-Quality Software Fixes

软件bug在每个软件应用程序中都很普遍。开发人员花费大量的时间和精力来修复软件错误。自动化程序修复技术的最新进展显示了一个有希望的方向，即提供机器生成的软件修复，以提高手动调试的负担。软件修复，无论是手工的还是机器生成的，都是密集的软件质量保证活动的交付物，在提高软件质量方面扮演着重要的角色。然而，很大一部分软件修复是不正确的：软件修复可能不能完全解决问题（例如，软件故障），或者可能引入新的问题（例如，软件修复本身包含错误）。低质量的软件修复是由修复者（即开发人员或自动程序修复技术）生成的。不幸的是，当前的质量保证研究和技术在检测低质量的软件修复和进一步提供指导以实现更高质量的软件修复方面是无效的。

Software bugs are prevalent in every software application. Developers spend a tremendous amount of time and effort in fixing software bugs. Recent advances in automated program repair techniques show a promising direction in providing machine-generated software fixes to elevate the burden of manual debugging. Software fixes, either manually or machine-generated, which are deliverables of intense software quality-assurance activities, play an important role in improving software quality. However, a significant portion of software fixes are incorrect: software fixes may not resolve the problems (e.g., software failures) completely, or may introduce new problems (e.g., the software fix itself contains bugs). Low-quality software fixes are produced by fixers (i.e., developers or automated program repair techniques). Unfortunately, current quality-assurance research and techniques are ineffective in detecting the low-quality software fixes and further providing guidance to achieve better-quality software fixes. The goal of this proposal is to provide a systematic process for producing high-quality software fixes. The proposal has three research objectives (ROs). RO1 - Obtaining a deep understanding of low-quality software fixes in software applications from different domains. Our work will focus on the problems in software fixes and refinement efforts that developers make to fix such problems. Also, our work will provide a deep understanding of the reasons why state-of-the-art bug detection techniques fail to find the problems in software fixes. RO2 - Helping fixers detect and improve low-quality software fixes. My students and I will design novel static and dynamic bug detection techniques that target the problems in low-quality software fixes. Such work will provide invaluable warnings to developers regarding potential problems in the to-be-committed software fixes. Also, we will build techniques to utilize the RO1 findings to provide refinement suggestions to produce high-quality software fixes. Furthermore, RO3 - Providing a generic feedback-oriented framework that integrates the detection and improvement of low-quality software fixes (RO2) into the current workflow of bug-fixers. The proposed framework will provide an iterative process to guide fixers to produce high-quality software fixes especially when the initial software fixes are insufficiently refined by RO2 (e.g., due to inadequate testing oracles and incomplete refinement efforts). The outcome of this research will provide complete solution to assist developers with producing high-quality software fixes. The proposed techniques will significantly improve software quality and reduce software development costs among Canadian companies. The proposed research will also train five highly qualified personnel (HQP) and allow them to contribute to state-of-the-art Software Engineering research and practice.