Auto-tools/Projects/Mozmill/Meeting 2010 11 30
Mozmill Buildbot Bustage
Last weeks Notes, per Henriks request
- Policy for handling test breakage (newgroup thread)
- Disabling tests
- Local test data (default bookmarks not really local)
- We need the "make" command to run Mozmill tests in mozilla-central
- We need documentation for developers how to run Mozmill tests
- Updating Mozmill on buildbot?
- Additionally bug fix release 126.96.36.199?
- Necessary to play nicely with our tests (see bug 613802)
- Mozmill testday this Friday, Nov 26th
- Mozmill 1.5.1 has been approved on AMO
- For disconnect errors we not always kill the browser (X-platform) - Need better stop logic (jhammel noted interests) - bug 615662
Timeline of Bustage
- Feature checked in on Friday that broke a mozmill test
- The only people around on Saturday to deal with broken test were Ehsan and Philor, not knowing what to do, they hid the mozmill column on tinderbox. Note this is not the same as backing out the buildbot code that runs mozmill and is much easier to reverse.
- The primary reason for contention was that if a test harness is on m-c it should be on try. If this had been the case, the dev would never have checked in, he'd have sent us an email and we could have handled this appropriately by fixing the test.
- The bug for adding Mozmill to try was filed by JhFord on Nov. 8, when Mozmill went live on m-c. It was considered RelEng territory and was not touched by the A-team. RelEng got busy and it dropped out of their triage, nothing was done.
- The decision to not do "standard" make targets for mozmill was one that we made during the mozmill+buildbot phase. We thought it would be sufficient to have make targets that worked post-"make test-package" stage. This was quite simply, wrong. It was well argued both ways.
Getting Mozmill turned back On
- Henrik already fixed the test (go Henrik and Jeff!) - not really (other problems)
- Jeff has makefile targets in a patch
- I'm pushing on RelEng this week about try server support
- Jeff, Joel, and I are reviewing including Manifests for 1.5.2
- Where are we with 1.5.2?
- Looking ahead
- Inclusion as the Firefox Driver for Selenium (davehunt)
- Overall 2.0 Python Refactoring (jhammel)