For the record this the test that alerted us to this issue was the following: https://wmf.ci.cloudbees.com/job/MobileFrontend-en.m.wikipedia.beta.wmflabs....
2 problems here - tests run only for the extension the code touches and currently browser tests only run after as these are slow and people would be annoyed if code took 20 minutes to merge. We've had issues in the past where changes in VisualEditor have broken things in MobileFrontend that trigger failed tests. Maybe core should run all browser tests for all deployed extensions as part of the merge process to avoid this?
"If MobileFrontend is so tightly coupled with the desktop login form, that is a problem with MobileFrontend." I don't really understand this. I get people moaning at me all the time that mobile has its own version of Watchlist compared to core, has its own skin etc etc and I get told the opposite that we should be closer to core. This is what we are striving for but we are not quite there yet.
"It seems that commenters here believe that the patch made it impossible to create an account if JavaScript was disabled, or via MobileFrontend - this is obviously not true, it just required an additional confirmation" I've not sure anyone has said it was impossible to create an account but user experience was badly effected as you point out. The statement I made was "Since most mobile device input fields default to lowercase ... pretty much anyone who now tries to register an account on mobile will see a warning that their username has been capitalized and will have to fill in the registration form again" [1]
[1] http://lists.wikimedia.org/pipermail/mobile-l/2014-March/006557.html
On Thu, Mar 6, 2014 at 5:13 PM, Greg Grossmeier greg@wikimedia.org wrote:
<quote name="Chris McMahon" date="2014-03-06" time="17:55:48 -0700"> > On Thu, Mar 6, 2014 at 5:49 PM, OQ <overlordq@gmail.com> wrote: > > > So I'm confused on the timeline here. > > > > Did the commit get merged before the testsuite found the breakage, or did > > the commit get merged despite the testsuite failing? > > > The commit was merged late Wednesday. The automated tests that > demonstrated the problem failed over Wednesday night and we analyzed the > failures early Thursday morning, which is routine.
And to clarify more: Not all tests run right away, some are more expensive and run on a schedule. These are part of that.
Should some of these tests be moved up to run immediately? Yeah, but we'd need to define the set of 'smoke tests' (tests that just test basic functionality, quickly) because we can't run all the tests all the time.
Greg
-- | Greg Grossmeier GPG: B2FA 27B1 F7EB D327 6B8E | | identi.ca: @greg A18D 1138 8E47 FAC8 1C7D |
Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l