Closed Bug 1314550 Opened 3 years ago Closed 3 years ago
TEST-UNEXPECTED-ERROR | test
_direct _update .py Test Direct Update .test _update | Attribute Error: 'dict' object has no attribute 'type' for updates to watershed releases
58 bytes, text/x-review-board-request
Bug 1247596 doesn't seem to have fixed this problem completely. So a follow-up seems to be necessary to hardening our update tests. As mentioned on bug 1314456 comment 1 we seem to hit an unexpected watershed release by testing the 47.0.2 release candidate builds. Reason could be that release-localtest also has the offer open for the 50.0rc builds. So when we are updating a 47.0.1 build on Windows it gets successfully updated to 47.0.2 but then it gets an additional update to 50.0rc. This bug will be used to remove this meaningless message and get a better one in place. https://dxr.mozilla.org/mozilla-central/source/testing/firefox-ui/harness/firefox_ui_harness/testcases.py#202 As you can see we trigger the download of the next update but do not wait for its completion. I don't know when the active update is being set but in case the download starts later we could have a race here. So I don't think that checking for the patch type makes any sense today. As what I got its always minor and major has never been used again since the Firefox 4.0 release. Instead we should simply fail with something like "Didn't expect another update offer".
We should get this fixed across all current channels down to release which will become our new 50.0 release.
Comment on attachment 8806666 [details] Bug 1314550 - Better failure message for update tests in case of watershed releases. https://reviewboard.mozilla.org/r/90042/#review89646
Attachment #8806666 - Flags: review?(rail) → review+
Pushed by firstname.lastname@example.org: https://hg.mozilla.org/integration/autoland/rev/fa64968dc3a1 Better failure message for update tests in case of watershed releases. r=rail
Bit late for 50 given that we're already spinning release candidate builds, but I guess landing it as a ride-along with any late approvals wouldn't be awful.
(In reply to Ryan VanderMeulen [:RyanVM] from comment #6) > Bit late for 50 given that we're already spinning release candidate builds, > but I guess landing it as a ride-along with any late approvals wouldn't be > awful. Well, don't we still sync mozilla-beta with mozilla-release? Also in case of any possible security release for Firefox 50 this fix for the tests would be more than helpful. So I really would like to see this finally landed on the release branch.
(In reply to Henrik Skupin (:whimboo) from comment #8) > Well, don't we still sync mozilla-beta with mozilla-release? Only if there are patches still landing there. Which isn't a given once we hit the release candidate stage. Anyway, ni? Wes to get this too if/when Ritu approves any other patches for 50RC2.
Flags: needinfo?(ryanvm) → needinfo?(wkocher)
Thanks a lot!
You need to log in before you can comment on or make changes to this bug.