CRII: SHF: Automatic Extraction of Error-Handling Specifications in Systems Software

软件错误处理是在程序执行过程中检测和响应错误发生的过程。理想情况下，无论何时发生运行时错误，软件系统都应该根据程序员的要求做出响应。年代的意图。不幸的是，情况往往并非如此。错误处理代码很难编写，而且往往难以理解、文档编制和测试。不出所料，错误处理代码经常有bug。软件错误处理程序中的错误是最普遍、最危险、最难以检测的错误。不正确的错误处理在系统软件（如操作系统）中尤其值得警惕，因为用户应用程序依赖于系统软件的可靠性。本研究旨在透过对错误处理规范的自动推理，更好地了解系统软件中的错误处理。错误处理规范描述了系统如何检测错误并从错误中恢复。本研究应用静态程序分析技术来自动推断系统软件中的错误处理规范。这项任务特别具有挑战性，因为系统软件实现了许多故障策略，并且错误处理代码经常在整个系统中扩散。此外，对如此大型代码库的分析经常面临可伸缩性问题。了解现有的错误处理策略是确保系统软件可靠的第一步。这种理解也有可能导致开发新的错误处理机制，以及对错误处理的新编程语言支持。这两种情况都可能对系统软件以外的软件可靠性产生重大影响。

Software error handling is the process of detecting and responding tothe occurrence of errors during the execution of a program. Ideally,whenever a runtime error occurs, software systems should respondaccording to the programmer?s intent. Unfortunately, that is often notthe case. Error-handling code is difficult to write, and tends to bepoorly understood, poorly documented, and poorly tested.Unsurprisingly, error-handling code is often buggy. Bugs in softwareerror handlers are some of the most pervasive, dangerous, anddifficult to detect bugs. Incorrect error-handling is particularlyalarming in systems software (e.g., the operating system) because userapplications depend on the reliability of systems software. Thisresearch aims to gather a better understanding of error handling insystems software through automatic inference of error-handlingspecifications.Error-handling specifications describe how the system detects andrecovers from errors. This research applies static program analysistechniques to automatically infer error-handling specifications insystems software. This task is particularly challenging becausesystems software implements numerous failure policies, anderror-handling code is often diffused through the system. Furthermore,the analysis of such large code bases often faces scalabilityproblems. Understanding existing error-handling strategies is thefirst step to ensure systems software is reliable. This understandingalso has the potential to lead to the development of newerror-handling mechanisms, and new programming language support forerror handling. Both of these could have a significant impact onsoftware reliability beyond systems software.