User Details
- User Since
- Oct 17 2014, 3:40 PM (581 w, 4 d)
- Roles
- Disabled
- LDAP User
- Cmcmahon
- MediaWiki User
- Cmcmahon(WMF) [ Global Accounts ]
Mar 12 2015
Note that I removed the "&" from the string with this commit: https://gerrit.wikimedia.org/r/#/c/195683
Mar 11 2015
Just updating to note that fixing this should turn the browser test builds back to green.
Mar 10 2015
Note: I set the version of Chrome manually in Jenkins for VE and the test went back to green. https://integration.wikimedia.org/ci/job/browsertests-VisualEditor-en.wikipedia.beta.wmflabs.org-linux-chrome-sauce/
Note that we don't do it, but it is possible to specify the version of Selenium for the SauceLabs jobs. As soon as they are supported at Sauce, we should upgrade to Selenium 2.45 and Chrome 41
Mar 9 2015
Mar 5 2015
Feb 26 2015
We could even make pools of users with different sets of privileges.
Feb 25 2015
I think index isn't helpful here because of the risk of a false positive when the test should fail
Feb 24 2015
Feb 23 2015
Lurking also. I've talked a bit with Arthur about the assumptions underlying burndown charts, and I don't think we've exhausted the subject.
Feb 20 2015
Feb 19 2015
Feb 17 2015
Feb 13 2015
Feb 12 2015
Feb 10 2015
now seeing upload fail with no js errors:
after fixing the gadget, seeing:
Feb 9 2015
Feb 6 2015
Handled this by better managing test users. There was a rogue test logging out.
See https://www.mediawiki.org/wiki/Wikimedia_Release_Engineering_Team/Process for current description of RelEng process understanding.
Feb 5 2015
Removed the explicit onfocus event from the Login method in mediawiki_selenium shared code. Update mediawiki_selenium to version 0.4.2 or later if Firefox shows the "a is undefined" error.
Feb 4 2015
I am pretty sure that on or around 16 January VE stopped handling this situation gracefully.
Errors in the console when this happens:
Here is a copy of the generic failure that is shared among all the repos:
I re-opened this, but I removed the VisualEditor project tag, because the problem is known to also affect Flow and MobileFrontend.
Feb 3 2015
I think this happens across repos because the shared step "Given I am logged in" fires an "onfocus" event on the login button before clicking in order to avoid missed clicks. If we can't fix this in JQuery we can remove the onfocus and see if that helps.
For a list of features shared among repositories, see https://www.mediawiki.org/wiki/Quality_Assurance/Browser_testing/Shared_features
This causes failures in five different browser tests.
We had a well-attended session where we demonstrated examples of good refactoring across a number of repositories.
I'd like to have Release Engineering in this conversation if you don't mind. I started a rough description of all the bits that RelEng does in support of Product teams here: https://www.mediawiki.org/wiki/User:Cmcmahon/releng_process
Feb 2 2015
I mentioned this to Chad, who pointed out that media/file storage on beta cluster is still on an NFS filesystem so searching media files always has been and always will be slow.
It's not exactly fixed. Elena discovered that by not firing an explicit javascript 'onfocus' event, the error doesn't show up.
It's not exactly fixed.
Jan 31 2015
Note that the search seems to provide complete results, but VE never recognizes that the results are finished.
Jan 30 2015
this seems likely to be an issue with SauceLabs and/or ChromeDriver, the @browser.window.resize_to command seems to have just stopped working for Chrome on SauceLabs, even though it works locally.
Works for me , resizes window, and passes locally in Chrome and Firefox. Looking at the Sauce builds...
It looks like:
Jan 29 2015
Literally every browser test in every repo that I run locally produces this error in Firefox.
Made highest priority, this is preventing important work.
Jan 27 2015
Jan 23 2015
Here is the stack track from trying to run an Echo test:
Also seen in Chrome as of 23 January
Jan 16 2015
Jeff, yes please update my key. I'm pretty sure the one in place right now is from a machine that has been wiped.
Jan 14 2015
Maybe a joint TPG-QA thing? I recently floated the idea of actually doing Acceptance Test Driven Development somewhere:
Erik Bernhardson fixed it with https://gerrit.wikimedia.org/r/#/c/184737/
Jan 13 2015
Need this merged: https://phabricator.wikimedia.org/T86685
