SHF: Small: BugX: In-house Debugging of Field Failures to Improve Software Quality

最近在Apache、Eclipse和Mozilla项目的开发人员中进行的一项调查表明，在调查bug报告时，重新创建现场故障（在用户机器上部署后发生的软件故障）的能力被认为是至关重要的。不幸的是，错误报告中通常包含的信息，如内存转储或调用堆栈，通常不足以重现问题。收集现场数据和帮助内部调试的更先进的方法倾向于向开发人员提供太少的信息，因此是无效的。该项目的总体目标是通过允许、支持和部分自动化现场故障的实际内部调试来提高技术水平。具体来说，这项研究将开发新的技术和工具，让开发人员重现，分析和理解，在内部，在现场观察到的故障。给定一个现场故障，开发的技术将（1）收集一组关于用户机器上故障的适当数据，（2）生成一个或多个可以针对故障应用程序执行的输入，并导致类似于观察到的故障，以及（3）提供故障的根本原因和可能的修复这些原因的提示。为了实现这一目标，研究将结合联合收割机静态和动态程序分析技术，并利用和扩展技术，用于测试部署的软件，输入生成和匿名化，和软件调试。如果成功，这项研究将为开发人员提供前所未有的优势，允许他们以与调试内部故障相同的方式调试现场故障，这将提高软件质量，并使依赖软件的社会各界受益。此外，该项目还将编制并向更广泛的科学界提供教育材料，其中包括研究成果、实施项目内开发的技术的工具以及经验评估中使用的软件基准样本。课程材料、工具、基础设施和基准的可用性将促进知识的发展，使该领域的更多研究成为可能，并最终进一步造福社会。

A recent survey conducted among developers of the Apache, Eclipse, andMozilla projects showed that the ability to recreate fieldfailures--failures of the software that occur after deployment, onuser machines--is considered of fundamental importance wheninvestigating bug reports. Unfortunately, the information typicallycontained in a bug report, such as memory dumps or call stacks, isusually insufficient for recreating the problem. Even more advancedapproaches for gathering field data and help in-house debugging tendto provide too little information to developers and to be thereforeineffective.The overall goal of this project is to improve the state of the art byallowing, supporting, and partially automating, actual in-housedebugging of field failures. Specifically, this research will developnovel techniques and tools that let developers reproduce, analyze, andunderstand, in-house, failures observed in the field. Given a fieldfailure, the developed techniques will (1) collect a suitable set ofdata about the failure on the user machine, (2) generate one or moreinputs that can be executed against the failing application and resultin a failure analogous to the one observed, and (3) provide hints onthe root causes of the failure and possible fixes for these causes. Toachieve this goal, the research will combine static and dynamicprogram analysis techniques and leverage and extend techniques fortesting deployed software, input generation and anonymization, andsoftware debugging. If successful, this research will provideunprecedented advantages to developers by allowing them to debug fieldfailures in the same way in which they debug in-house ones, which willimprove software quality and benefit all segments of society thatdepend on software. Furthermore, the project will develop and makeavailable to the broader scientific community educational materialsthat incorporate research findings, tools that implement thetechniques developed within the project, and samples of the softwarebenchmarks used in empirical evaluations. The availability ofcurriculum materials, tools, infrastructure, and benchmarks willadvance knowledge, enable additional research in the area, andultimately further benefit society.