A combined representation for the maintenance of C programs

Abstract

A programmer wishing to make a change to a piece of code must first gain a full understanding of the behaviours and functionality involved. This process of program comprehension is difficult and time consuming, and often hindered by the absence of useful program documentation. Where documentation is absent, static analysis techniques are often employed to gather programming level information in the form of data and control flow relationships, directly from the source code itself. Software maintenance environments are created by grouping together a number of different static analysis tools such as program sheers, call graph builders and data flow analysis tools, providing a maintainer with a selection of 'views' of the subject code. However, each analysis tool often requires its own intermediate program representation (IPR). For example, an environment comprising five tools may require five different IPRs, giving repetition of information and inefficient use of storage space. A solution to this problem is to develop a single combined representation which contains all the program relationships required to present a maintainer with each required code view. The research presented in this thesis describes the Combined C Graph (CCG), a dependence-based representation for C programs from which a maintainer is able to construct data and control dependence views, interprocedural control flow views, program slices and ripple analyses. The CCG extends earlier dependence-based program representations, introducing language features such as expressions with embedded side effects and control flows, value returning functions, pointer variables, pointer parameters, array variables and structure variables. Algorithms for the construction of the CCG are described and the feasibility of the CCG demonstrated by means of a C/Prolog based prototype implementation.