Starting sometime around 20/7-25/7 web-platform-tests started experiencing a large number of new low-frequency intermittent timeouts. There are no obvious wpt changes in that time period , but I do notice that it seems to correspond to dustin's configuration changes for taskcluster, although not every new intermittent occurs on TC, and I can't obviously see any relevant configuration changes. The only plausible-looking marionette change I can see that might affect things is , but it's hard to see how because I believe that code is only called from the python navigate() method, which we only use to load the initial harness page, which isn't being reloaded before the failing tests. So I am back to wondering if something could have changed in the test environment that might affect the worst-case performance of the tests. Otherwise I suppose a gecko change might have had some effect…  http://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?startdate=2016-07-19&enddate=2016-07-25  http://hg.mozilla.org/integration/mozilla-inbound/rev/b2e910bb1d72
The configuration changes that I made were tested to ensure that they produced the same task description (modulo things like treeherder attributes which wouldn't affect pass/fail status). That the failures are occurring on Buildbot as well is also strong evidence these changes are unrelated. The TC test environment is also defined in-tree in testing/docker/desktop-test, so you can look for changes there. But again, Buildbot is completely different so given that the failures span TC and BB, I doubt the test environment is related.
This should hopefully be fixed by bug 1288579.
Closing given bug 1295840 comment 6.
Status: NEW → RESOLVED
Closed: 3 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.