TIGer: A Test Instance Generator for Large Programs

Testing accounts for a large and growing share of software development costs. Developing tests is labor-intensive; and significant savings would result from tools that could automatically generate tests meeting some given criterion of completeness. ATC-NY proposes to build a set of modules, Tiger, that can be incorporated into existing test generators and will allow the generation of test cases for large systems. Recent work has explored ways to generate tests by model-checking. The limiting factor in model-checking is the size of the state space to be explored and much research has been devoted to strategies for reducing the state spaces to be checked. Nearly all of it addresses the problem of program verification and, as a result, is not directly applicable to generating tests. Nonetheless, this work contains a rich body of ideas, which we will adapt to define and implement Tiger. COMMERCIAL APPLICATIONS: Testing is ubiquitous, and the government and private sector markets for automatic test generation is potentially huge. In addition, a standards body such as NIST could use automatic test generation to develop acceptance tests for implementation of its standards. ATC-NY will address the DoD market by going through the prime contractors with which ATC has collaborative relationships.