Last modified: 2013-07-02 11:10:33 UTC
The number of browser tests and the number of browser test builds have both increased to the point that our original idea of a mid-day build and a late-day build (Pacific time) no longer serves us well. The builds take long enough that they run essentially back-to-back all afternoon every day. I suggest we run all the builds at 11:00 Pacific time to have timely build information available around deployment time during the US working day. Then I suggest that we run all the builds once at 19:00 Pacific time (GMT-8) so that the results will be available for analysis in the morning Croation time (GMT-1)
Actually, upon reflection, I think we can support the weekly releases better, and move toward even faster release cycles perhaps by something like this: Builds for beta labs to run several times per day. Possibly run Chrome/FF builds more often than IE builds for speed. Builds for test2 to match deployments to test2/mw.o depending on features tested e.g. VisualEditor. Builds for production to match deployments to prod. "Sufficiently advanced testing is indistinguishable from monitoring." Builds for mobile TBD.
Fixed in https://gerrit.wikimedia.org/r/#/c/69865/ If there is still something to be done here, reopen the bug.