Component Data-Flow Analysis for Large-Software Systems

9900988 Barbara ryder大型工业规模的软件系统（100万行代码）缺乏足够的测试和程序理解工具，特别是当这样的系统包含异构组件和/或遗留代码时。需要可伸缩的编译时分析技术作为适当的软件工具的基础，这些工具可以为可预测的成本提供可调整的精度级别。对系统的任意组件（如模块）的分析结果应该是可获得的，即使整个系统的代码不能用于检查。本研究的目标是探索适用于这些大型工业系统的新分析，强调完整系统源代码的可伸缩性和不可用性问题。这将通过设计“组件数据流分析”模式来完成，这是一种可扩展的数据流技术，在给定一些关于不可用程序组件的粗粒度数据流信息后，能够对源代码可用的特定组件应用可接受的精确数据流分析。通过使用PROLANGS分析框架（PAF）构建的原型实现（一个用于编译时程序分析的公开软件平台），组件分析的效用将通过其在测试和程序理解中有用的几个数据流问题中的应用来演示

9900988 Barbara RyderLarge industrial-sized software systems (of 1 million lines of code) suffer from a lack of adequate testing and program understanding tools, especially when such systems contain heterogeneous components and/or legacy codes. Scalable compile-time analysis techniques are needed to serve as a basis for appropriate software tools which can provide adjustable levels of precision for predictable cost. Analysis results should be obtainable for an arbitrary component (e.g., module) of a system, even if the code for the entire system is not available for inspection. The goal of this research is to explore new analyses applicable to these large industrial systems, emphasizing the issues of scalability and unavailability of full system source code. This will be accomplished by design of a schema for "component data-flow analysis", a scalable data-flow technique which, given some coarse-grained data-flow information about unavailable program components, is able to apply an acceptably precise data-flow analysis on a particular component whose source code is available. The utility of component analysis will be demonstrated by its application to several data-flow problems useful for testing and program understanding, using a prototype implementation built with the PROLANGS Analysis Framework (PAF), a publically available software platform for compile-time program analysis