Demo repo to demonstrate problems in code coverage reports with switch cases.

In the test file, we run tests that test the switch case for case 1, case 4 and the default case. case 2 and case 3 are not tested. However, the code coverage, as shown above, shows that we've handled all branches except case 3, thus including case 2 which is not actually covered.

The reason, I suspect, is because the program is transformed to be something like:

module.exports = function (x) {
	switch (x) {
		case 1:
		  incrementBranch_1();
		case 2:
			incrementBranch_2();
			incrementStatement_1();
			return 'ok';
		case 3:
			incrementBranch_3();
		case 4:
			incrementBranch_4();
			incrementStatement_2();
			return 'also ok';
		default:
			incrementBranch_default();
			incrementStatement_3();
			return 'why not';
	}
};

but because of the fallthrough behaviour, incrementBranch_2() also gets called when case 1 is entered.

This falsifies the code coverage, and makes you think you have 100% code coverage when you actually don't. In one of my packages, I duplicated some switch-case heavy code in order to have a reliable code coverage.