Coverage and mutation

Coverage and mutation show what the tests do not check. Coverage finds the lines no test runs; mutation testing makes small changes to the code and finds the ones no test notices. Loops read them to judge the tests, not the code.

What it sees

Coverage says a line ran during the tests, not that anything checked its result. Mutation testing goes further: a mutant that survives is a change to the code that no test would notice. Coverage of the changed lines matters more than the total, which can hide a new gap behind old covered code. Neither says whether the tests check the right behaviour.

Reading the signal

• Uncovered lines: each changed line that no test ran, as file:line.
• A surviving mutant: the change to the code that survived. Either a test lacks an assertion, or the mutant is equivalent to the original and no test could catch it.
• A score under its threshold: the share of covered lines or killed mutants fell below its floor.

Mutation runs are slow, so they run on the changed files or on a schedule. A mutation run that stopped before it tested its mutants, or a coverage report with no data, has observed nothing.