Extending traditional static analysis techniques to support development, testing and maintenance of component-based solutions

Abstract

Traditional static code analysis encompasses a mature set of techniques for helping understand and optimize programs, such as dead code elimination, program slicing, and partial evaluation (code specialization). It is well understood that compared to other program analysis techniques (e.g., dynamic analysis), static analysis techniques do a reasonable job for the cost associated with implementing them. Industry and government are moving away from more ‘traditional’ development approaches towards component-based approaches as ‘the norm.’ Component-based applications most often comprise a collection of distributed object-oriented components such as forms, code snippets, reports, modules, databases, objects, containers, and the like. These components are glued together by code typically written in a visual language. Some industrial experience shows that component-based development and the subsequent use of visual development environments, while reducing an application's total development time, actually increase certain maintenance problems. This provides a motivation for using automated analysis techniques on such systems. The results of this research show that traditional static analysis techniques may not be sufficient for analyzing component-based systems. We examine closely the characteristics of a component-based system and document many of the issues that we feel make the development, analysis, testing and maintenance of such systems more difficult. By analyzing additional summary information for the components as well as any available source code for an application, we show ways in which traditional static analysis techniques may be augmented, thereby increasing the accuracy of static analysis results and ultimately making the maintenance of component-based systems a manageable task. We develop a technique to use semantic information about component properties obtained from type library and interface definition language files, and demonstrate this technique by extending a traditional unreachable code algorithm. To support more complex analysis, we then develop a technique for component developers to provide summary information about a component. This information can be integrated with several traditional static analysis techniques to analyze component-based systems more precisely. We then demonstrate the effectiveness of these techniques on several real Department of Defense (DoD) COTS component-based systems.