Good article and it's saying something that I've also been saying for years. tl;dr: A code coverage measurement only says where you did -not- check for bugs (s

Good article and it's saying something that I've also been saying for years. tl;dr: A code coverage measurement only says where you did -not- check for bugs (since code that was not exercised can't possibly have been checked for bugs); moreover it also does -not- say whether the code that was covered (i.e. exercised) was tested to any meaningful degree (it's perfectly possible to get a code coverage number without checking any results at all). I do disagree with one statement in the article: "In fact I would argue, it’s hard to imagine a code coverage value of over 70% to be of any value whatsoever.". Having seen it done for a mission-critical system, if one uses extremely strict TDD (i.e. no line of code is written without a corresponding unit test that verifies its effects) then the high code coverage number does provide useful direct guidance as to what areas of the code haven't been fully scrutinized. In our case, the efforts paid off: there were no significant bugs found after release.