学术文献库

Automated test case generation tools help businesses to write tests and increase the safety net provided by high regression test coverage when making code changes. Test generation needs to cover as much as possible of the uncovered code while avoiding generating redundant tests for code that is already covered by an existing test-suite. In this paper we present our work on a tool for the real world application of integrating formal analysis with automatic test case generation. The test case generation is based on coverage analysis using the Java bounded model checker (JBMC). Counterexamples of the model checker can be translated into Java method calls with specific parameters. In order to avoid the generation of redundant tests, it is necessary to measure the coverage in the exact same way as JBMC generates its coverage goals. Each existing coverage measurement tool uses a slightly different instrumentation and thus a different coverage criterion. This makes integration with a test case generator based on formal analysis difficult. Therefore, we developed BlueCov as a specific runtime coverage measurement tool which uses the exact same coverage criteria as JBMC does. This approach also allows for incremental test-case generation, only generating test coverage for previously untested code, e.g., to complete existing test suites.