基于编译器中间语言的软件运行时可靠性研究

发布时间：2018-02-01 11:09

本文关键词： 可靠性抽象语法树插桩软件错误隔离编程接口建模　出处：《清华大学》2015年博士论文　论文类型：学位论文

【摘要】：随着计算机软件应用在越来越多的关键岗位,软件运行时的不可靠行为会造成严重的危害。在软件运行时对其进行分析和监控,可以诊断和约束软件的不可靠行为,消除软件不可靠性带来的危害。对软件编译时生成的编译器中间语言进行分析和插桩,可以分析和监控软件运行时的行为。因此在软件运行时如何通过编译器中间语言增强软件的可靠性成为一个重要的研究问题。本文针对软件运行时的可靠性进行研究,主要研究完全优化后的插桩无法准确定位语句和表达式的问题、运行时软件错误隔离缺乏可移植性问题和第三方软件库文档缺乏解析的问题。本文的主要贡献如下:(1)抽象语法树级别的插桩方法PAST。由于编译时完全优化的软件调试信息并不准确,因此在软件开发人员对运行时软件进行分析时,存在着无法准确定位带有语义信息的语句和表达式的问题。PAST利用未优化的编译器中间语言和带有语义信息的抽象语法树之间的对应关系,为开发人员提供带有语义信息的插桩方式,从而解决了在运行时无法准确定位语句和表达式的问题。此外,PAST在运行时不仅提供跨平台的分析框架,同时将分析工具在软件运行时的基本开销降低至可以忽略,这使得PAST可以直接应用在产品级软件的分析中。(2)基于中间语言的软件错误隔离方法WebC。在机器指令级别监控运行时软件行为的软件错误隔离方法,需要为不同的体系结构设计不同的运行时行为约束,这导致软件错误隔离方法在不同体系结构之间存在缺乏可移植性的问题。WebC利用编译器中间语言具有操作可移植性的优点,在编译器中间语言级别插入软件运行时的约束,从而解决了软件错误隔离缺乏可移植性的问题。此外,WebC不仅可以直接嵌入在浏览器中,在运行时WebC安全约束对软件性能的影响也并不明显。(3)基于自然语言处理的编程接口文档解析。使用自然语言编写的编程接口文档,无法解析成包含软件运行时行为约束的、形式化的规范。本方法通过识别文档结构、识别自然语言单词,解决了自然语言编程接口文档无法解析成软件运行时规范的问题。
[Abstract]:With the application of computer software in more and more critical positions, the unreliable behavior of software running will cause serious harm. It can diagnose and constrain the unreliable behavior of software, eliminate the harm caused by software unreliability, and analyze and insert the intermediate language of compiler generated when the software is compiled. It is possible to analyze and monitor the behavior of software runtime. Therefore, how to enhance the reliability of software through compiler intermediate language becomes an important research problem. This paper focuses on the reliability of software runtime. Research. This paper mainly studies the problem that the perfectly optimized inserted pile can not accurately locate the statements and expressions. The lack of portability of runtime software error isolation and the lack of parsing of third-party software library documents. The main contributions of this article are as follows: 1). Abstract syntax tree-level piling method pass. Due to the complete optimization of software debugging information at compile time is not accurate. So when software developers analyze runtime software. There is a problem that statements and expressions with semantic information can not be accurately located. Fast uses the unoptimized compiler intermediate language to match the abstract syntax tree with semantic information. The method of inserting posts with semantic information is provided to the developers, which solves the problem that the statements and expressions can not be accurately located at run time. In addition, PAST not only provides a cross-platform analysis framework at run time. At the same time, the basic overhead of the analysis tool is reduced to negligible when the software is running. This allows PAST to be used directly in the analysis of product-level software. An intermediate language-based software error isolation method WebC. a software error isolation method for monitoring runtime software behavior at the machine instruction level. Different runtime behavior constraints need to be designed for different architectures. This leads to the lack of portability of software error isolation methods between different architectures. WebC makes use of compiler intermediate language to have the advantage of operational portability. The constraint of software runtime is inserted at the intermediate language level of compiler, which solves the problem that software error isolation lacks portability. In addition, WebC can not only be embedded directly in browser. At runtime, the impact of WebC security constraints on software performance is not obvious.) programming interface documents based on natural language processing. Programming interface documents written in natural language. This method can not be parsed into a formal specification that contains the behavior constraints of the software runtime. This method recognizes natural language words by recognizing the structure of the document. It solves the problem that the interface document of natural language programming can not be parsed into software runtime specification.
【学位授予单位】：清华大学
【学位级别】：博士
【学位授予年份】：2015
【分类号】：TP311.53

【相似文献】