New test perma-red/orange failures first seen Sun Nov 6 2016, 0:59:48: https://treeherder.mozilla.org/#/jobs?repo=comm-central&revision=c140c92d693ecf0231baf10331afabedca07fbbf Those weren't there on the day before. TEST-UNEXPECTED-FAIL | xpcshell-icaljs.ini:calendar/test/unit/test_alarmservice.js | xpcshell return code: 0 TEST-UNEXPECTED-FAIL | xpcshell-icaljs.ini:calendar/test/unit/test_alarmservice.js | test_addItems - [test_addItems : 12] addTestItems Test 6; check fired - false == true TEST-UNEXPECTED-FAIL | xpcshell-icaljs.ini:calendar/test/unit/test_datetime.js | xpcshell return code: 0 TEST-UNEXPECTED-FAIL | xpcshell-icaljs.ini:calendar/test/unit/test_datetime.js | really_run_test - [really_run_test : 74] 1 == 0 I suspect some subtle M-C change, maybe some JS change (just guessing). Regression range M-C: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=73b008257626&tochange=c44c01dfd264 Everything that landed after midday on the 5th of November 2016.

Wild guess: The update to SQLite? 41d172f46ee9 Ryan VanderMeulen — Bug 1315244.

I've recently submited a patch on that file: https://hg.mozilla.org/comm-central/rev/4cb52bea1cc6 but "Test 6" has not been touched, apart from adding a message to identify the tests that fail. It could be the dst change on the 6th of November for America\Los Angeles timezone. "Test 6" sets an alarm for a daily recurring event starting 2 days before and the 6th of November the day lasts 25 hours instead 24.

On the other file (test_datetime.js) the failure seems here: https://dxr.mozilla.org/comm-central/source/calendar/test/unit/test_datetime.js#74 and again it seems a DST change issue (New York timezone this time) that happens with ical.js enabled (see also Bug 1218500).

Tests are passing again. Maybe this was a one time failure because tests were executed during daylight saving time to standard time transition.

What's strange with this test failures that these only occured for ical.js - this smells like an ical.js bug. Regardless of the tests now passing again, I suggest to keep this bug open to understand what happend. And if turns out to be an expected behaviour, we should convert the tests to fail expectedly when in transition from/to DST. There have also been test failures in the todaypane mozmill test, which now also disappeared (n.b. Mozmill tests are also running with ical.js)

Has this issue come back? Is there some way we can reproduce this failure? If not, I'd suggest we close this bug.

It will come back on the next DST change in the US ;-) Or did it get fixed in bug 1352148?

I also think this _will_ come back on Automn DST switch, especially since this is an ical.js only failure. I am tempted to leave this open as it looks like an ical.js problem instead of one of the test code itself, but I haven't dig into it myself.

(In reply to Jorg K (GMT+2) from comment #7) > ... Or did it get fixed in bug 1352148? No, it's a different issue and a different cause. Maybe I'm wrong but I think it might depend, somehow, on the automatic normalization of the date/time.