Closed Bug 1382972 Opened 2 years ago Closed 2 years ago
Connection-add Ice Candidate .html | Add valid candidate should never resolve when pc is closed - Peer connection is closed
59 bytes, text/x-review-board-request
Filed by: archaeopteryx [at] coole-files.de https://treeherder.mozilla.org/logviewer.html#?job_id=116283512&repo=mozilla-inbound https://archive.mozilla.org/pub/firefox/tinderbox-builds/mozilla-inbound-win64-pgo/1500611423/mozilla-inbound_win8_64_test_pgo-web-platform-tests-e10s-4-bm110-tests1-windows-build52.txt.gz
Duplicate of this bug: 1382986
:Standard8, can you help find someone to triage this? Its perma-fail on w10 debug-only atm: Interestingly, it only started failing on https://treeherder.mozilla.org/#/jobs?repo=mozilla-central&revision=e99e3e62cfa8bbc7d430f64d4fbdf77ffb81e979&filter-resultStatus=testfailed&filter-resultStatus=busted&filter-resultStatus=exception&filter-resultStatus=retry&filter-resultStatus=usercancel&filter-resultStatus=runnable&filter-resultStatus=success&filter-searchStr=windows+wpt10&filter-tier=1&filter-tier=2&filter-tier=3 where before that it was consistently green
That is https://github.com/soareschen/web-platform-tests/blob/master/webrtc/RTCPeerConnection-addIceCandidate.html but I'm not sure how we injest wpt testing into our CI so I don't know which rev of that we ran, specifically.
if this was fixed upstream, we can ask :jgraham to sync upstream to our source. :jgraham, how can we determine which revision of a wpt test is in our tree?
Whiteboard: [stockwell needswork]
Quite frankly I think these WPT tests are just really poorly written. I highly recommend to disable/skip these failing tests for now and complain upstream.
:jgraham who is ni'd on this bug already would have a bigger picture about the origination of the tests and expectations/requirements of them prior to being in-tree. He might have ideas. If there are a bunch of these that are randomizing us, we either have a flaky browser (as other browser vendors run these same tests), flaky infrastructure, or possibly others are just skipping these tests and we are the only ones running them.
I created https://github.com/w3c/web-platform-tests/issues/6611 with a suggestion of how make these tests more reliable. I think we should skip them until they are fixed/improved to avoid the noise.
Looks like there is now some PR upstream to remove these tests: https://github.com/w3c/web-platform-tests/pull/6616. If someone marks it reviewed I can make sure it's merged and then start a new wpt sync. jmaher: In general [vcs] log testing/web-platform/tests will tell you the last upstream revision that was imported e.g. james@ginny:~/develop/gecko$ git log --grep=revision --oneline -n1 -- testing/web-platform/tests/ 7f18a43b0677 Bug 1381842 - Update web-platform-tests to revision 85ae24e3dc80ee63b6dc2ed78a922cff68c6e819, a=testonly
For now I'm going to disable this test in our repo.
Comment on attachment 8890034 [details] Bug 1382972 - disable wpt test RTCPeerConnection-addIceCandidate.html https://reviewboard.mozilla.org/r/161088/#review166528 This is a rather extreme solution :-) I was more aiming for skipping only the tests which are causing problems (or skip all which are written with the same problematical pattern). I guess this is a way to skip them http://searchfox.org/mozilla-central/source/testing/web-platform/meta/webrtc/RTCPeerConnection-onnegotiationneeded.html.ini#17
Attachment #8890034 - Flags: review?(drno) → review-
:jgraham is this a way to skip a single test in WPT http://searchfox.org/mozilla-central/source/testing/web-platform/meta/webrtc/RTCPeerConnection-onnegotiationneeded.html.ini#17 ? Or is there a better way to skip/not execute a single test within a given file?
Flags: needinfo?(standard8) → needinfo?(james)
[RTCPeerConnection-onnegotiationneeded.html] type: testharness expected: TIMEOUT [task for negotiationneeded event should be enqueued for next tick] expected: TIMEOUT [addTransceiver() should fire negotiationneeded event] expected: NOTRUN [Calling addTransceiver() twice should fire negotiationneeded event once] expected: FAIL [negotiationneeded event should not fire if signaling state is not stable] expected: FAIL [negotiationneeded event should fire only after signaling state go back to stable] disabled: https://bugzilla.mozilla.org/show_bug.cgi?id=1382972 expected: NOTRUN Will cause the result of "negotiationneeded event should fire only after signaling state go back to stable" to be ignored. The test will still run of course.
Comment on attachment 8890034 [details] Bug 1382972 - disable wpt test RTCPeerConnection-addIceCandidate.html https://reviewboard.mozilla.org/r/161088/#review166910 LGTM
Attachment #8890034 - Flags: review?(na-g) → review+
Pushed by firstname.lastname@example.org: https://hg.mozilla.org/integration/autoland/rev/78c932ff7deb disable wpt test RTCPeerConnection-addIceCandidate.html r=ng
Whiteboard: [stockwell needswork] → [stockwell disabled]
You need to log in before you can comment on or make changes to this bug.