The Research On Test Case Generation Methods For Testing Syntax Analysis Of Compiler Implementation

The compiler is one of the fundamental support tools for software development, which has direct impacts on the application software’s quality. Therefore, the quality of the compiler is essential. As with other software, software testing is the most important method for finding faults within a compiler and assuring its quality. As an implementation of a programming language standard, the compiler has complex input space and software behaviors. It is impossible to achieve full test coverage in compiler testing. So the core problem for compiler testing is how to generate a set of test cases with high testing effectiveness.With complex structures, the test case in compiler testing is a piece of source code written in one type of high level language. Manual test case generation is not only a heavy workload, but also difficult to guarantee adequate coverage. In order to overcome the disadvantages of manual generation, compiler testing requires automatic test case generation methods. The generated test suite should have good abilities for covering the syntactic structure and semantic properties of the programming language. Most of the existing test case generation methods for compiler testing are based on grammar coverage. Since the test coverage rules based on grammar have limited abilities to reveal the syntactic structure of the language, this paper tries to make a description of the language syntax with binary trees and generate test data by combining the syntactic units with each other. On the introduction of a parameter called "the computation path length of the syntactic unit", the whole generation process is under proper control and so is the size of the output test suite. By adjusting values of the computation path length, the devised test case generation method can generate a test suite that subsumes the counterpart from methods based on grammar coverage. This implies that the given generation method in this paper has stronger abilities for covering the syntactic structure.