Re: [extremeprogramming] What has changed for you over the past 20 years?
"James Grenning" <[email protected]> Wed, 04 Dec 2019 22:13:36 -0500
| Newsgroups | gmane.comp.programming.extreme-programming |
|---|---|
| Message-ID | <[email protected]> |
I warn people of the potential dysfunction of code coverage as a goal. Like Mike said: take out the checks and you still have 100%! For those tests you might as well buy a test generator; then you can get the 100% test coverage bonus for meeting your goals. I suggest to people in code influenced by TDD and design skill, you get areas of 100% and 0%. Something that is a mix is probably a design problem. This is what happens to my code. I have areas that are suitable for TDD and have 100% coverage (not because that is the goal, that is just what happens). Areas that are not suitable for TDD have 0% coverage and complexity of 1 or 2. They tend to be areas that do not change. Cheers! James -------------------------------------------------------------------------- James Grenning - Author of TDD for Embedded C - wingman-sw.com/tddec wingman-sw.com wingman-sw.com/blog twitter.com/jwgrenning facebook.com/wingman.sw [](http://wingman-sw.com) On 4 Dec 2019, at 17:11, Slava Imeshev via Groups.Io wrote: > Of course we mean good tests. I’m doubtful that it’s possible to > claim a good test without good coverage. > > Slava > >> On Dec 4, 2019, at 2:05 PM, Mike Bowler >> <[email protected]> wrote: >> >> We rapidly get a collection of tests that increase code coverage >> without testing any actual behaviour. I’ve seen whole test setups >> with few or no assertions. This is then worse than having low test >> coverage because we can’t trust any of it. >> >> Then when the team is in the habit of writing garbage tests, that >> starts to leak into their production code and quality starts dropping >> there too. They care less about the quality. I expect this is some >> variant of Social Proof or the Broken Window Effect. >> >> I can’t think of a single case where imposing a high code coverage >> number on checkin, up front, has resulted in the team writing better >> code and I can think of many cases where the reverse is true. >> >> Now, the one case that I have seen work is where the team starts a >> mandatory code coverage based on where they are today and very >> gradually increases the mandatory amount over time to slowly improve. >> Although even that causes problems when new people are brought onto >> the team or new teams take over that code base. >> >> Mike Bowler >> Agile & Technical, Coach and Trainer >> Cell: 905 409-7052 >> <tel:905%20409-7052>www.GargoyleSoftware.com/mike_bowler >> <http://www.gargoylesoftware.com/mike_bowler> >> www.UnconsciousAgile.com <http://www.unconsciousagile.com/> >> >> Ask about our Certified Scrum Developer (CSD) training classes >> <http://www.gargoylesoftware.com/training/csd>. In-house and public >> classes available. >> >>> On Dec 4, 2019, at 4:51 PM, John Maxwell <[email protected]> wrote: >>> >>> On 12/04/2019 04:47:58 PM, Mike Bowler wrote: >>>>> Yep. TDD is a holly grail. Just having a policy not to check in >>>>> any >>>> code without 100% unit test coverage does wonders to quality. >>>> >>>> I can’t tell if you meant that as sarcasm. I’ve seen that >>>> policy put >>>> in place and the “quality” that results. I expect many of us >>>> here have >>>> seen that disaster. >>>> >>> I haven't. What's the failure mechanism? >>> >>> -- >>> John Maxwell KB3VLL [email protected] >>> >>> "Personally I'm fond of using Dedekind cuts for representing the >>> Reals, >>> but it gets cumbersome to use when balancing one's checkbook." >>> -Charles Haynes >>> >>> >>> >> >> > > > -=-=-=-=-=-=-=-=-=-=-=- Groups.io Links: You receive all messages sent to this group. View/Reply Online (#160407): https://groups.io/g/extremeprogramming/message/160407 Mute This Topic: https://groups.io/mt/39418691/2417047 Group Owner: [email protected] Unsubscribe: https://groups.io/g/extremeprogramming/leave/4902963/619838065/xyzzy [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-