Source code generator

It is a common practice to perform compiler testing with a set of hand-written sample programs. However, gathering larger collections of high-quality test cases is a di ficult and time consuming task. This work explores the potential of automatic source code generation techniques to simplify the construction of test suites for compilers and other source code processing tools (e.g. pretty-printers, static code analysis tools, refactoring engines, etc.). The goal is to design a source code generator capable of producing compiletime correct programs in a selected high-level programming language. A set of theoretical methods and algorithms for pseudo-random source code generation is presented. Specifi c methods are designed for handling context-free and context-sensitive syntax rules of the target languages. The applicability of these methods in practice is demonstrated by the implementation of three prototype tools two source code generators and one grammar analysis tool. One of the source code generators produces random sentences according to a given stochastic context-free grammar while the other can be used for generation of ISO/IEC 7185:1990 conformant Pascal programs