Let's look at:
A real-world Perl library with inadequate test coverage
The tool we used to measure that test coverage