Description
Details
Related Objects
Event Timeline
Another list @ https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/547/testReport/ – contains some tests that weren’t failing in the other list and vice versa, so apparently this whole business is flaky as well, great :(
I guess I’ll just be very liberal in commenting out tests.
Change 504622 had a related patch set uploaded (by Lucas Werkmeister (WMDE); owner: Lucas Werkmeister (WMDE)):
[mediawiki/extensions/Wikibase@master] Comment out lots of failing Ruby browser tests
Change 504622 merged by jenkins-bot:
[mediawiki/extensions/Wikibase@master] Comment out lots of failing Ruby browser tests
Okay, current status:
- 15 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/550/
- 13 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/551/
- 14 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/552/
- 14 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/553/
- 11 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/554/
- 12 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/555/
I’ll upload another patch.
Change 505757 had a related patch set uploaded (by Lucas Werkmeister (WMDE); owner: Lucas Werkmeister (WMDE)):
[mediawiki/extensions/Wikibase@master] Comment out more failing Ruby browser tests
To be honest, I’m starting to doubt the usefulness of this commenting-out spree. There are now more commented-out scenarios than ones left active, and the way that different test failures appear and vanish between test runs makes me think that most of the failures aren’t related to any individual test.
But let’s merge this and see what happens to the daily CI runs then, I suppose.
Change 505757 merged by jenkins-bot:
[mediawiki/extensions/Wikibase@master] Comment out more failing Ruby browser tests
- 7 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/557/
- 8 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/558/
- 10 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/559/
- 10 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/560/
- 5 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/561/
- 8 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/562/
- 10 failures: https://integration.wikimedia.org/ci/job/selenium-Wikibase-chrome/BROWSER=chrome,MEDIAWIKI_ENVIRONMENT=beta,PLATFORM=Linux,label=DebianJessie%20&&%20contintLabsSlave/563/
I propose we decline this task. It seems clear to me that it’s not the individual tests that are broken, and disabling more and more of them is just a prolonged game of whack-a-mole until we’ve reached the stable end state of having disabled every last ruby browser test.
We should rewrite the tests in Selenium (making sure that they actually pass) and ignore the ruby ones until we can completely remove them.
That's worse than we imagined .. I'm wondering where the flakiness is coming from, just to make sure we do not carry it over with us into the re-written wdio tests. Did you get to figure out a little more why they are flaky already?
I know browser tests can waste great amount of effort in getting them working and useful, and at the end of the tunnel one would wonder whether the benefit justifies the costs.
So let's do this:
- timebox this (4 hours?) to figure out where the flakiness coming from. Is it selenium (we use selenium for wdio too)? is it race-conditions between client and server state? or is it just cucumber/capybara?
- shift focus entirely to re-writing those tests in javascript using wdio, taking with us any useful learnings from 1.
@Lucas_Werkmeister_WMDE since you've put enough effort into this already, please make the decision on the above as you see fit
I didn’t put any time into investigating the failures because as I recall, we decided against that and instead wanted to just disable “the failing tests” (it now turns out there’s no such distinction, any test can apparently fail).
We can still do the timeboxed investigation, but since I don’t think any of the work I made so far will be useful there (it was pretty mechanical work), and I’m no longer on the campsite, I think someone else should do that.
Per discussion above, I'd call it declined as it does not seem to be sensible thing to do.