Generation strategies for test-suites of grammar-based software

Abstract

The use of statement coverage has proved to be a useful met- ric when testing code with a test-suite. Similarly, the cover- age of a grammar’s rules is an effective metric when testing a parser. However when testing a whole parser front-end, it is not immediately obvious whether there is a correlation between rule coverage and underlying code coverage. We use a number of generation strategies to generate a series of test-suites. We apply these test-suites to keystone, a parser front-end for ISO C++ and offer empirical evidence to sug- gest which generation strategy offers the best coverage whilst using the least amount of test-cases