Filed by: aiakab [at] mozilla.com https://treeherder.mozilla.org/logviewer.html#?job_id=172887790&repo=mozilla-inbound https://queue.taskcluster.net/v1/task/UrTiIaj1SP2BKzTCvgarxA/runs/0/artifacts/public/logs/live_backing.log

Link to a recent log https://treeherder.mozilla.org/logviewer.html#?job_id=172887790&repo=mozilla-inbound&lineNumber=5415 Part of that log: 15:56:39 INFO - 565 INFO TEST-OK | dom/media/test/test_mediarecorder_principals.html | took 3457ms 15:56:39 INFO - GECKO(3060) | --DOMWINDOW == 19 (0000016FD2B4F000) [pid = 7660] [serial = 466] [outer = 0000000000000000] [url = about:blank] 15:56:39 INFO - GECKO(3060) | --DOMWINDOW == 18 (0000016FD264F400) [pid = 7660] [serial = 462] [outer = 0000000000000000] [url = about:blank] 15:56:39 INFO - GECKO(3060) | ++DOMWINDOW == 19 (0000016FD3229C00) [pid = 7660] [serial = 478] [outer = 0000016FD2B4E800] 15:57:04 INFO - GECKO(3060) | #01: mozilla::detail::RunnableMethodImpl<RefPtr<mozilla::VideoTrackEncoder>,void ( mozilla::VideoTrackEncoder::*)(__int64),1,0,__int64>::Run() [xpcom/threads/nsThreadUtils.h:1218] 15:57:04 INFO - GECKO(3060) | #02: mozilla::TaskQueue::Runner::Run() [xpcom/threads/TaskQueue.cpp:244] 15:57:04 INFO - GECKO(3060) | #03: nsThreadPool::Run() [xpcom/threads/nsThreadPool.cpp:230] 15:57:04 INFO - GECKO(3060) | #04: nsThread::ProcessNextEvent(bool,bool *) [xpcom/threads/nsThread.cpp:1100] 15:57:04 INFO - GECKO(3060) | #05: NS_ProcessNextEvent(nsIThread *,bool) [xpcom/threads/nsThreadUtils.cpp:519] 15:57:04 INFO - GECKO(3060) | #06: mozilla::ipc::MessagePumpForNonMainThreads::Run(base::MessagePump::Delegate *) [ipc/glue/MessagePump.cpp:334] 15:57:04 INFO - GECKO(3060) | #07: MessageLoop::RunHandler() [ipc/chromium/src/base/message_loop.cc:320] 15:57:04 INFO - GECKO(3060) | #08: MessageLoop::Run() [ipc/chromium/src/base/message_loop.cc:300] 15:57:04 INFO - GECKO(3060) | #09: nsThread::ThreadFunc(void *) [xpcom/threads/nsThread.cpp:427] 15:57:05 INFO - GECKO(3060) | #10: PR_NativeRunThread [nsprpub/pr/src/threads/combined/pruthr.c:406] 15:57:05 INFO - GECKO(3060) | #11: static unsigned int pr_root(void *) [nsprpub/pr/src/md/windows/w95thred.c:138] 15:57:05 INFO - GECKO(3060) | #12: ucrtbase.dll + 0x20369 15:57:05 INFO - GECKO(3060) | #13: KERNEL32.DLL + 0x12774 15:57:05 INFO - GECKO(3060) | #14: ntdll.dll + 0x70d51 15:57:05 ERROR - 566 INFO TEST-UNEXPECTED-FAIL | dom/media/test/test_mediarecorder_principals.html | assertion count 1 is more than expected 0 assertions 15:57:05 INFO - 567 INFO TEST-START | dom/media/test/test_mediarecorder_record_4ch_audiocontext.html 15:57:05 INFO - GECKO(3060) | ++DOMWINDOW == 20 (0000016FD322E800) [pid = 7660] [serial = 479] [outer = 0000016FD2B4E800] 15:57:05 INFO - GECKO(3060) | ++DOCSHELL 0000016FD233B000 == 5 [pid = 7660] [id = {63fdb4b1-9204-4731-aa92-15fdf83d300f}] 15:57:05 INFO - GECKO(3060) | ++DOMWINDOW == 21 (0000016FD3232400) [pid = 7660] [serial = 480] [outer = 0000000000000000] 15:57:05 INFO - GECKO(3060) | ++DOMWINDOW == 22 (0000016FD3233400) [pid = 7660] [serial = 481] [outer = 0000016FD3232400] 15:57:05 INFO - GECKO(3060) | --DOMWINDOW == 21 (0000016FD322A000) [pid = 7660] [serial = 467] [outer = 0000000000000000] [url = about:blank

There are 22 failures in the last 7 days. All of them on windows10-64-qr / debug I did some retriggers and it seems that the fail started from https://treeherder.mozilla.org/#/jobs?repo=mozilla-inbound&revision=db2973e7667c86f991162206506e64fdcd2d18d4&filter-searchStr=test-windows10-64-qr%2Fdebug-mochitest-media-e10s%20M-e10s(mda) :drno, can you please take a look?

I caught this in rr, but I also saw the error from bug 1468831. I think bug 1468831 is the failure mode of this bug in opt builds.

This happens because we naively assume in the track listeners that a MediaSegment is always all-null or all-non-null. But the bug occurs when it's a mix, so our check [1] is fooled and the track encoder advances too far. [1] https://searchfox.org/mozilla-central/rev/fcfb479e6ff63aea017d063faa17877ff750b4e5/dom/media/encoder/MediaEncoder.cpp#275-285

Pushed by pehrsons@gmail.com: https://hg.mozilla.org/integration/autoland/rev/54761b7fd534 Do not assume segments are all non-null when not all null. r=padenot

Is there a big risk to taking this on Beta to help lower the noise from intermittents? If not, please nominate for approval.

I'd say low risk.

Comment on attachment 9019399 [details] Bug 1453078 - Do not assume segments are all non-null when not all null. r?padenot [Beta/Release Uplift Approval Request] Feature/Bug causing the regression: Bug 1296531 User impact if declined: Mostly intermittent failures on try in debug builds. There could also be small a/v sync issues in recordings in release builds. Is this code covered by automated tests?: Yes Has the fix been verified in Nightly?: No Needs manual test from QE?: No If yes, steps to reproduce: List of other uplifts needed: None Risk to taking this patch: Low Why is the change risky/not risky? (and alternatives if risky): The amount of changed code is non-trivial but it's logically simple; we move to finer-grained checks and aggregate the results manually. String changes made/needed:

Comment on attachment 9019399 [details] Bug 1453078 - Do not assume segments are all non-null when not all null. r?padenot [Triage Comment] Low-risk fix which helps reduce intermittent oranges on Beta. Approved for 64.0b6.