Software Engineering · Testing
The percentage of code lines (or branches) executed at least once by the test suite.
In practice: Coverage is a useful floor, not a ceiling — 80% coverage is better than 10%, but 100% coverage doesn't mean the tests are actually meaningful. Tests that don't assert anything can hit 100%.
Debated: Coverage targets are controversial — many argue they incentivize writing tests that inflate numbers rather than tests that catch bugs, and a hard threshold can waste effort on genuinely untestable error paths.