Automatic test generation is a recurring theme in computer science research [1]–[11]. Also called test-input generation, it has many uses in software engineering, ranging from augmenting manual testing efforts to automatically generating test suites. Determining inputs that cause a program to execute a particular code path is generally undecidable [12]. Nevertheless, researchers have proposed many heuristics and approximations to produce tests for realistic source code. These heuristics and approximations are often based on the intuition (sometimes experimentally supported) about the types of programs that the analysis is likely to encounter. The unfortunate side effect is that the accuracy or the performance of the analysis algorithms may be adversely affected by the structure of the input program.