Code Coverage

There are 3 criteria that the test input must satisfy in order to ensure we're detecting faults in our code.

Reachability: the test input must allow the test to reach or execute the faulty statement.
Necessity: the test input must make the faulty statement produce an incorrect intermediate result.
Propagation: the test input must propagate the incorrect intermediate result so that the end result is incorrect and observable.
(Dianxiang Xu 330-334)

Code Coverage / Statement Coverage

Code coverage is a necessary condition for revealing faulty statements because it satisfies the reachability criteria. But 100% coverage does not guarantee fault detection because it may not meet the necessity or the propagation criteria.

The local static function MySqrt has a fault in its conditional logic (where it incorrectly checks x <= 0 instead of x < 0). The test input of 1 executes all lines but the negative return path. The test input of -1 executes all lines but the Math.Sqrt calculation path.

[TestCase(1, 1)]
[TestCase(-1, -1)]
public void NonRevealingFault(double input, double expected)
{
    static double MySqrt(double x)
    {
        if (x <= 0) // [FAULT] should be x < 0
        {
            return -1;
        }
        return Math.Sqrt(x);
    }

    var actual = MySqrt(input);
    Assert.AreEqual(expected, actual);
}

Together they achieve 100% coverage and have executed the fault line twice. However, they did not reveal the fault! A test input of 0 would have revealed the fault since it meets all three criteria.

This is because √0 = 0, but MySqrt(0) returns -1 instead of 0.