Currently the MediaFormatReader and Demuxing share the same TaskQueue. In trying to implement main-thread-blocking in drawImage() (bug 1295921) by storing the last keyframe, we're hitting an issue where tasks dispatched before recovery seek at blocking recovery because MediaCache waits on main thread.

Split MFR and decoders to use independent TaskQueues.

Comment on attachment 8815147 [details] Bug 1319992: P1. Run demuxing operations on its own task queue. https://reviewboard.mozilla.org/r/96162/#review96350 ::: dom/media/MediaFormatReader.h:56 (Diff revision 2) > RefPtr<SeekPromise> > Seek(const SeekTarget& aTarget, int64_t aUnused) override; > > protected: > void NotifyDataArrivedInternal() override; > + void UpdateBuffered() override {}; It is confusing to have an empty implementation. We should make MediaDecoderReader::NotifyDataArrived virtual: class MediaDecoderReader { virtual void NotifyDataArrived() { MOZ_ASSERT(OnTaskQueue()); NS_ENSURE_TRUE_VOID(!mShutdown); UpdateBuffered(); } } class MediaFormatReader { void NotifyDataArrived() override { NotifyDataArrivedInternal(); } } ::: dom/media/MediaFormatReader.h:488 (Diff revision 2) > bool NeedInput(DecoderData& aDecoder); > > DecoderData& GetDecoderData(TrackType aTrack); > > // Demuxer objects. > - RefPtr<MediaDataDemuxer> mDemuxer; > + class DemuxerFactory; Call it "DemuxerProxy" since it is not a factory but a proxy to delegate jobs to another thread. ::: dom/media/MediaFormatReader.cpp:445 (Diff revision 2) > + , mData(new Data(aDemuxer)) > + { } > + > + ~DemuxerFactory() > + { > + RefPtr<Data> data = mData; mData.forget(). ::: dom/media/MediaFormatReader.cpp:458 (Diff revision 2) > + > + RefPtr<MediaDataDemuxer::InitPromise> Init(); > + > + MediaTrackDemuxer* GetTrackDemuxer(TrackType aTrack, uint32_t aTrackNumber) > + { > + MOZ_RELEASE_ASSERT(mData && mData->mInitDone); We just need to check mData once in the constructor to ensure it is not null. ::: dom/media/MediaFormatReader.cpp:526 (Diff revision 2) > + class Wrapper; > + struct Data > + { > + NS_INLINE_DECL_THREADSAFE_REFCOUNTING(Data) > + > + Data(MediaDataDemuxer* aDemuxer) : mInitDone(false), mDemuxer(aDemuxer) {} explicit Data(...); ::: dom/media/MediaFormatReader.cpp:528 (Diff revision 2) > + { > + NS_INLINE_DECL_THREADSAFE_REFCOUNTING(Data) > + > + Data(MediaDataDemuxer* aDemuxer) : mInitDone(false), mDemuxer(aDemuxer) {} > + > + Atomic<bool> mInitDone; Can't use in-class initialization? ::: dom/media/MediaFormatReader.cpp:569 (Diff revision 2) > + } > + > + RefPtr<SeekPromise> Seek(const media::TimeUnit& aTime) override > + { > + RefPtr<Wrapper> self = this; > + return InvokeAsync( I prefer: return InvokeAsync(mTaskQueue, __func__, [self, aTime]() { return self->mTrackDemuxer->Seek(aTime); })->Then(mTaskQueue, __func__, [self]() { self->UpdateRandomAccessPoint(); }, [self]() { self->UpdateRandomAccessPoint(); }) ->CompletionPromise(); to avoid nested lambda. ::: dom/media/MediaFormatReader.cpp:649 (Diff revision 2) > + }); > + SyncRunnable::DispatchToThread(mTaskQueue, runnable); > + } > + > +private: > + Monitor mMonitor; a Mutex will do the job. ::: dom/media/MediaFormatReader.cpp:2583 (Diff revision 2) > + mBuffered = TimeIntervals(); > + return; > + } > + > + int64_t startTime = 0; > + if (!ForceZeroStartTime()) { Need to rebase on bug 1313635.

Comment on attachment 8815147 [details] Bug 1319992: P1. Run demuxing operations on its own task queue. https://reviewboard.mozilla.org/r/96162/#review96350 > It is confusing to have an empty implementation. > > We should make MediaDecoderReader::NotifyDataArrived virtual: > > class MediaDecoderReader > { > virtual void NotifyDataArrived() > { > MOZ_ASSERT(OnTaskQueue()); > NS_ENSURE_TRUE_VOID(!mShutdown); > UpdateBuffered(); > } > } > > class MediaFormatReader > { > void NotifyDataArrived() override > { > NotifyDataArrivedInternal(); > } > } i'm not sure it's less confusing to have default implementation that aren't actually called. but sure... > Call it "DemuxerProxy" since it is not a factory but a proxy to delegate jobs to another thread. it does create MediaTrackDemuxer , but sure.. > We just need to check mData once in the constructor to ensure it is not null. hmmm.. the aim is to ensure that this code is never called before the init promise has completed. checking in the constructor serves no purpose, seeing that data there will always be non-null > Can't use in-class initialization? not in an elegant fashion no. > a Mutex will do the job. when do you use a mutex over a monitor? mutexes seem seldom use in the media code.

Comment on attachment 8815147 [details] Bug 1319992: P1. Run demuxing operations on its own task queue. https://reviewboard.mozilla.org/r/96162/#review96350 > I prefer: > > return InvokeAsync(mTaskQueue, __func__, [self, aTime]() { > return self->mTrackDemuxer->Seek(aTime); > })->Then(mTaskQueue, __func__, > [self]() { self->UpdateRandomAccessPoint(); }, > [self]() { self->UpdateRandomAccessPoint(); }) > ->CompletionPromise(); > > to avoid nested lambda. it does look nicer. But doesn't makes it go through like this: T1 = reader's taskqueue T2 = demuxer taskqueue with nested lambda: T1 -> T2 -> T2 -> T1 with nice rewrite T1 -> T2 -> T1 -> T2 -> T1 InvokeAsync will resolve back to the reader's taskqueue, which will then queue a new task on demuxer's taskqueue to complete the task.

changing to non-nested lambda gives me assertion failure about running in the wrong thread. Assertion failure: mResponseTarget->IsCurrentThreadIn(), at /Users/jyavenard/Work/Mozilla/obj-ff-dbg/dist/include/mozilla/MozPromise.h:340 7 INFO SourceBuffer buffered ranges grew from TimeRanges: to TimeRanges: 8 INFO got durationchange event 9 INFO got loadedmetadata event LLVM ERROR: IO failure on output stream. fatal error: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/otool: internal objdump command failed #01: mozilla::MozPromise<RefPtr<mozilla::MediaTrackDemuxer::SamplesHolder>, mozilla::MediaResult, true>::ThenValueBase::CompletionPromise() (MozPromise.h:340, in XUL) #02: mozilla::MediaFormatReader::DemuxerProxy::Wrapper::GetSamples(int) (MediaFormatReader.cpp:584, in XUL) 10 INFO TEST-PASS | dom/media/mediasource/test/test_AudioChange_mp4.html | got loadedmetadata event #03: mozilla::MediaFormatReader::DoDemuxAudio() (MediaFormatReader.cpp:1327, in XUL) #04: mozilla::MediaFormatReader::RequestDemuxSamples(mozilla::TrackInfo::TrackType) (MediaFormatReader.cpp:1582, in XUL) #05: mozilla::MediaFormatReader::Update(mozilla::TrackInfo::TrackType) (MediaFormatReader.cpp:1960, in XUL) #06: decltype(*(fp).*fp0(Get<0ul>(fp1).PassAsParameter())) mozilla::detail::RunnableMethodArguments<mozilla::TrackInfo::TrackType>::applyImpl<mozilla::MediaFormatReader, void (mozilla::MediaFormatReader::*)(mozilla::TrackInfo::TrackType), StoreCopyPassByConstLRef<mozilla::TrackInfo::TrackType>, 0ul>(mozilla::MediaFormatReader*, void (mozilla::MediaFormatReader::*)(mozilla::TrackInfo::TrackType), mozilla::Tuple<StoreCopyPassByConstLRef<mozilla::TrackInfo::TrackType> >&, mozilla::IndexSequence<0ul>) (nsThreadUtils.h:779, in XUL) #07: _ZN7mozilla6detail23RunnableMethodArgumentsIJNS_9TrackInfo9TrackTypeEEE5applyINS_17MediaFormatReaderEMS6_FvS3_EEEDTcl9applyImplfp_fp0_dtdefpT10mArgumentscvNS_13IndexSequenceIJLm0EEEE_EEEPT_T0_ (nsThreadUtils.h:785, in XUL) #08: mozilla::detail::RunnableMethodImpl<void (mozilla::MediaFormatReader::*)(mozilla::TrackInfo::TrackType), true, false, mozilla::TrackInfo::TrackType>::Run() (nsThreadUtils.h:814, in XUL) #09: mozilla::AutoTaskDispatcher::TaskGroupRunnable::Run() (TaskDispatcher.h:196, in XUL) #10: mozilla::TaskQueue::Runner::Run() (TaskQueue.cpp:232, in XUL) #11: nsThreadPool::Run() (nsThreadPool.cpp:226, in XUL) #12: non-virtual thunk to nsThreadPool::Run() (nsThreadPool.cpp:153, in XUL) #13: nsThread::ProcessNextEvent(bool, bool*) (nsThread.cpp:1214, in XUL) #14: NS_ProcessNextEvent(nsIThread*, bool) (nsThreadUtils.cpp:361, in XUL) #15: mozilla::ipc::MessagePumpForNonMainThreads::Run(base::MessagePump::Delegate*) (MessagePump.cpp:338, in XUL) #16: MessageLoop::RunInternal() (message_loop.cc:233, in XUL) #17: MessageLoop::RunHandler() (message_loop.cc:226, in XUL) #18: MessageLoop::Run() (message_loop.cc:205, in XUL) #19: nsThread::ThreadFunc(void*) (nsThread.cpp:469, in XUL) #20: _pt_root (ptthread.c:219, in libnss3.dylib) #21: _pthread_body (in libsystem_pthread.dylib) + 180 #22: _pthread_body (in libsystem_pthread.dylib) + 0 the code: RefPtr<SamplesPromise> GetSamples(int32_t aNumSamples) override { RefPtr<Wrapper> self = this; return InvokeAsync(mTaskQueue, __func__, [self, aNumSamples]() { return self->mTrackDemuxer->GetSamples(aNumSamples); }) ->Then(mTaskQueue, __func__, [self]() { self->UpdateRandomAccessPoint(); }, [self]() { self->UpdateRandomAccessPoint(); }) ->CompletionPromise(); } :gerald, :jwwang any ideas?

Not from me I'm afraid, I don't really understand the workings of MozPromise. JW should know better hopefully, or you can corner Bobby next week! Just looking at ThenValueBase::CompletionPromise(), it enforces its use on the target thread of the 'Then', so you'll just have to work around that, unless it can safely be changed to work on any thread, but that's where you need a real MozPromise expert.

Ok.. This is a limitation of CompletionPromise, it must be set on the previous request, and in the target thread of the previous Request. Discussing with :bholley, he also mentioned what I mentioned in comment #6, that you now have to dispatch two extra tasks and is an unnecessary overhead. A workaround could be: RefPtr<SamplesPromise> GetSamples(int32_t aNumSamples) override { RefPtr<Wrapper> self = this; return InvokeAsync(mTaskQueue, __func__, [self, aNumSamples]() { return self->mTrackDemuxer->GetSamples(aNumSamples); }) ->Then(mTaskQueue, __func__, [self]() { self->UpdateRandomAccessPoint(); }, [self]() { self->UpdateRandomAccessPoint(); }) ->Then(AbstractThread::GetCurrent()->AsTaskQueue(), __func__, [] {}) ->CompletionPromise(); } The problem is setting that the last request is a CompletionPromise in a thread safe manner, and there's no guarantee that the previous tasks didn't already run, making it too late to make it a completion promise. As such, I'll go back to using nested lambda.

(In reply to Jean-Yves Avenard [:jya] from comment #15) > Ok.. > > This is a limitation of CompletionPromise, it must be set on the previous > request, and in the target thread of the previous Request. http://searchfox.org/mozilla-central/rev/d98418da69edeb1f2f8e6f3840157fae1512f89b/xpcom/threads/MozPromise.h#417 This is because mCompletionPromise is read on the target thread http://searchfox.org/mozilla-central/rev/d98418da69edeb1f2f8e6f3840157fae1512f89b/xpcom/threads/MozPromise.h#342 and created in CompletionPromise() which must also run on the target thread to avoid race. I am think about creating a ThenWithCompletion() function so we can do: p->ThenWithCompletion(thread1, __func__, onResolve1, onReject1) ->ThenWithCompletion(thread2, __func__, onResolve2, onReject2); The syntax is more similar to that of JS.

Comment on attachment 8815549 [details] Bug 1319992: P0. Have CompletionPromise returns a RefPtr<MozPromise>. https://reviewboard.mozilla.org/r/96434/#review96636 Can you elaborate how that makes a difference?

Comment on attachment 8815264 [details] Bug 1319992: P2. Update MediaDecoderReader documentation. https://reviewboard.mozilla.org/r/96274/#review96638

Comment on attachment 8815549 [details] Bug 1319992: P0. Have CompletionPromise returns a RefPtr<MozPromise>. https://reviewboard.mozilla.org/r/96434/#review96636 Yes.. You can do something like: return InvokeAsync(mTaskQueue, __func__, [self, aTime]() { return self->mTrackDemuxer->Seek(aTime) ->Then(self->mTaskQueue, __func__, [self]() { self->UpdateRandomAccessPoint(); }, [self]() { self->UpdateRandomAccessPoint(); }) ->CompletionPromise(); }); otherwise you can't as InvokeAsync static_assert that the type returned is a RefPtr<MozPromise> You can currently do: return InvokeAsync(mTaskQueue, __func__, [self, aTime]() { return self->mTrackDemuxer->Seek(aTime) ->Then(self->mTaskQueue, __func__, [self]() { self->UpdateRandomAccessPoint(); }, [self]() { self->UpdateRandomAccessPoint(); }); }); Then can be chained, as all Then returns RefPtr<MozPromise> but not CompletionPromise which returns MozPromise*

Comment on attachment 8815265 [details] Bug 1319992: P3. Remove no longer used seeking argument. https://reviewboard.mozilla.org/r/96276/#review96642

Comment on attachment 8815266 [details] Bug 1319992: P4. Use Mutex in place of Monitor. https://reviewboard.mozilla.org/r/96278/#review96644

Comment on attachment 8815549 [details] Bug 1319992: P0. Have CompletionPromise returns a RefPtr<MozPromise>. https://reviewboard.mozilla.org/r/96434/#review96650

Comment on attachment 8815549 [details] Bug 1319992: P0. Have CompletionPromise returns a RefPtr<MozPromise>. https://reviewboard.mozilla.org/r/96434/#review96818

JW, something very dodgy since rebasing to use ThenPromise https://treeherder.mozilla.org/logviewer.html#?job_id=32535724&repo=try the promise returned is null. I don't see how that could ever happen. Haven't dwelved into it yet.

See bug 1323155.

Comment on attachment 8818471 [details] Bug 1319992: P5. Don't attempt to estimate readyState when ended. https://reviewboard.mozilla.org/r/98530/#review98804 LGTM.

Pushed by jyavenard@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/781555873370 P1. Run demuxing operations on its own task queue. r=jwwang https://hg.mozilla.org/integration/autoland/rev/d3366444625b P2. Update MediaDecoderReader documentation. r=jwwang https://hg.mozilla.org/integration/autoland/rev/ab31c60d46d7 P3. Remove no longer used seeking argument. r=jwwang https://hg.mozilla.org/integration/autoland/rev/49877fd60322 P4. Use Mutex in place of Monitor. r=jwwang https://hg.mozilla.org/integration/autoland/rev/f982627067f7 P5. Don't attempt to estimate readyState when ended. r=jwwang

backed out for frequent test failures like https://treeherder.mozilla.org/logviewer.html#?job_id=7998841&repo=autoland starting with this push

Backout by cbook@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/6d6e627bb384 Backed out changeset f982627067f7 https://hg.mozilla.org/integration/autoland/rev/9782a71736b2 Backed out changeset 49877fd60322 https://hg.mozilla.org/integration/autoland/rev/b81f3cb4782b Backed out changeset ab31c60d46d7 https://hg.mozilla.org/integration/autoland/rev/df6043463ade Backed out changeset d3366444625b https://hg.mozilla.org/integration/autoland/rev/be7270495e69 Backed out changeset 781555873370 for frequent test failurs in mediasource-endofstream.html

The "fix" to the intermittent timeout that cause those changes to be backed out is extremely suspicious. Rather than having DemuxerProxy::GetTrackDemuxer return a already_AddRefed<Wrapper> it returns a Wrapper* This causes the DemuxerProxy to always own the last reference to the MediaTrackDemuxer rather than passing the ownership to the MediaFormatReader. The only difference is that in MediaFormatReader::Shutdown, the MediaTrackDemuxer are deleted shortly later. It makes no sense on why it would work in one way but not another. But at this stage, I'm beyond caring :(

Pushed by jyavenard@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/f3ac066c3741 P1. Run demuxing operations on its own task queue. r=jwwang https://hg.mozilla.org/integration/autoland/rev/45afcd8cf9ef P2. Update MediaDecoderReader documentation. r=jwwang https://hg.mozilla.org/integration/autoland/rev/3cfdc09bf851 P3. Remove no longer used seeking argument. r=jwwang https://hg.mozilla.org/integration/autoland/rev/2ae0864c76bc P4. Use Mutex in place of Monitor. r=jwwang https://hg.mozilla.org/integration/autoland/rev/48b968826893 P5. Don't attempt to estimate readyState when ended. r=jwwang

Backed out for failing test_eme_stream_capture_blocked_case1.html: https://hg.mozilla.org/integration/autoland/rev/98dade0102954e835c4acf99d50e41717dd8cb3f https://hg.mozilla.org/integration/autoland/rev/9b90a8fec5ac840f05d4d67b9799ed0a7a3d7e58 https://hg.mozilla.org/integration/autoland/rev/6726d2d57b405ad4280203aab665fdaab98dc506 https://hg.mozilla.org/integration/autoland/rev/4248712f23ce750c15ebf89bdc43d7578f75e47e https://hg.mozilla.org/integration/autoland/rev/6ba59bfd211d616c3c6dc1be7fff5b15cd6839ec Push with failures: https://treeherder.mozilla.org/#/jobs?repo=autoland&revision=48b9688268938ffe6ad33ea84891a76170a9a597 Failure log: https://treeherder.mozilla.org/logviewer.html#?job_id=8105008&repo=autoland 03:07:31 INFO - [03:07:31.157] 640x480@959kbps audio&video tracks, each with its key-6_case1 created MediaKeys object ok 03:07:31 INFO - TEST-PASS | dom/media/test/test_eme_stream_capture_blocked_case1.html | [03:07:31.157] 640x480@959kbps audio&video tracks, each with its key-6_case1 setMediaKeys failed as expected. 03:07:31 INFO - [03:07:31.209] 640x480 then 400x300, same key (1st) per track-7_case1 video: fetch of bipbop_480_624kbps-cenc-video-key1-init.mp4 complete, appending 03:07:31 INFO - [03:07:31.220] 640x480 then 400x300, same key (1st) per track-7_case1 audio: fetch of bipbop_480_624kbps-cenc-audio-key1-init.mp4 complete, appending 03:07:31 INFO - Buffered messages finished 03:07:31 ERROR - TEST-UNEXPECTED-FAIL | dom/media/test/test_eme_stream_capture_blocked_case1.html | application terminated with exit code 1 03:07:31 INFO - runtests.py | Application ran for: 0:02:47.516931 03:07:31 INFO - zombiecheck | Reading PID log: /var/folders/q8/1vnm3qf56qs7z6b9j3w0gwl800000w/T/tmpyDN_vGpidlog 03:07:31 INFO - ==> process 2018 launched child process 2020 03:07:31 INFO - ==> process 2018 launched child process 2021 03:07:31 INFO - ==> process 2018 launched child process 2022 03:07:31 INFO - ==> process 2018 launched child process 2023 03:07:31 INFO - zombiecheck | Checking for orphan process with PID: 2020 03:07:31 INFO - zombiecheck | Checking for orphan process with PID: 2021 03:07:31 INFO - zombiecheck | Checking for orphan process with PID: 2022 03:07:31 INFO - zombiecheck | Checking for orphan process with PID: 2023 03:07:31 INFO - mozcrash Downloading symbols from: https://queue.taskcluster.net/v1/task/RG8ZjSV2QaChovm6HQj3Kw/artifacts/public/build/firefox-53.0a1.en-US.mac.crashreporter-symbols.zip 03:07:38 INFO - mozcrash Copy/paste: /builds/slave/test/build/macosx64-minidump_stackwalk /var/folders/q8/1vnm3qf56qs7z6b9j3w0gwl800000w/T/tmp1rcLWL.mozrunner/minidumps/D5C5099C-5407-4648-AD49-B850D72C758D.dmp /var/folders/q8/1vnm3qf56qs7z6b9j3w0gwl800000w/T/tmpiEbdUg 03:07:52 INFO - mozcrash Saved minidump as /builds/slave/test/build/blobber_upload_dir/D5C5099C-5407-4648-AD49-B850D72C758D.dmp 03:07:52 INFO - mozcrash Saved app info as /builds/slave/test/build/blobber_upload_dir/D5C5099C-5407-4648-AD49-B850D72C758D.extra 03:07:52 INFO - PROCESS-CRASH | dom/media/test/test_eme_stream_capture_blocked_case1.html | application crashed [@ mozilla::MozPromise<bool, mozilla::MediaResult, true>::ThenValueBase::AssertIsDead()] 03:07:52 INFO - Crash dump filename: /var/folders/q8/1vnm3qf56qs7z6b9j3w0gwl800000w/T/tmp1rcLWL.mozrunner/minidumps/D5C5099C-5407-4648-AD49-B850D72C758D.dmp 03:07:52 INFO - Operating system: Mac OS X 03:07:52 INFO - 10.10.5 14F27 03:07:52 INFO - CPU: amd64 03:07:52 INFO - family 6 model 69 stepping 1 03:07:52 INFO - 4 CPUs 03:07:52 INFO - 03:07:52 INFO - GPU: UNKNOWN 03:07:52 INFO - 03:07:52 INFO - Crash reason: EXC_BAD_ACCESS / KERN_INVALID_ADDRESS 03:07:52 INFO - Crash address: 0x0 03:07:52 INFO - Process uptime: 167 seconds 03:07:52 INFO - 03:07:52 INFO - Thread 66 (crashed) 03:07:52 INFO - 0 XUL!mozilla::MozPromise<bool, mozilla::MediaResult, true>::ThenValueBase::AssertIsDead() [MozPromise.h:48b968826893 : 357 + 0x0] 03:07:52 INFO - rax = 0x0000000108f87913 rdx = 0x00007fff797911f8 03:07:52 INFO - rcx = 0x000000010486a598 rbx = 0x00007fff79791c50 03:07:52 INFO - rsi = 0x0000a9000000a900 rdi = 0x0000a8000000a903 03:07:52 INFO - rbp = 0x00000001312c9a10 rsp = 0x00000001312c9a00 03:07:52 INFO - r8 = 0x00000001312c99b0 r9 = 0x00000001312ca000 03:07:52 INFO - r10 = 0x00007fff9027d3ef r11 = 0x00007fff9027d3c0 03:07:52 INFO - r12 = 0x000000010a33a998 r13 = 0x000000010a33ab98 03:07:52 INFO - r14 = 0x0000000000000000 r15 = 0x0000000000000001 03:07:52 INFO - rip = 0x0000000106d2ea7e 03:07:52 INFO - Found by: given as instruction pointer in context 03:07:52 INFO - 1 XUL!mozilla::MozPromise<bool, mozilla::MediaResult, true>::ThenValueBase::ResolveOrRejectRunnable::~ResolveOrRejectRunnable() [MozPromise.h:48b968826893 : 321 + 0x6] 03:07:52 INFO - rbx = 0x000000012cc40c10 rbp = 0x00000001312c9a30 03:07:52 INFO - rsp = 0x00000001312c9a20 r12 = 0x000000010a33a998 03:07:52 INFO - r13 = 0x000000010a33ab98 r14 = 0x0000000000000000 03:07:52 INFO - r15 = 0x0000000000000001 rip = 0x0000000106d2b498 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 2 XUL!<name omitted> [nsThreadUtils.cpp:48b968826893 : 45 + 0xb] 03:07:52 INFO - rbx = 0x0000000000000000 rbp = 0x00000001312c9a50 03:07:52 INFO - rsp = 0x00000001312c9a40 r12 = 0x000000010a33a998 03:07:52 INFO - r13 = 0x000000010a33ab98 r14 = 0x0000000000000000 03:07:52 INFO - r15 = 0x0000000000000001 rip = 0x000000010502698e 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 3 XUL!mozilla::TaskQueue::Dispatch(already_AddRefed<nsIRunnable>, mozilla::AbstractThread::DispatchFailureHandling, mozilla::AbstractThread::DispatchReason) [nsCOMPtr.h:48b968826893 : 294 + 0x6] 03:07:52 INFO - rbx = 0x000000012cf2e7d0 rbp = 0x00000001312c9a80 03:07:52 INFO - rsp = 0x00000001312c9a60 r12 = 0x000000010a33a998 03:07:52 INFO - r13 = 0x000000010a33ab98 r14 = 0x0000000000000000 03:07:52 INFO - r15 = 0x0000000000000001 rip = 0x0000000104ff842e 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 4 XUL!mozilla::MozPromise<bool, mozilla::MediaResult, true>::ThenValueBase::Dispatch(mozilla::MozPromise<bool, mozilla::MediaResult, true>*) [MozPromise.h:48b968826893 : 376 + 0x9] 03:07:52 INFO - rbx = 0x000000012cc40c10 rbp = 0x00000001312c9ad0 03:07:52 INFO - rsp = 0x00000001312c9a90 r12 = 0x000000010a33a998 03:07:52 INFO - r13 = 0x000000010a33ab98 r14 = 0x000000012c1fc5b0 03:07:52 INFO - r15 = 0x000000012655fe40 rip = 0x0000000106d2b1e3 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 5 XUL!mozilla::MozPromise<bool, mozilla::MediaResult, true>::DispatchAll() [MozPromise.h:48b968826893 : 735 + 0x8] 03:07:52 INFO - rbx = 0x0000000000000000 rbp = 0x00000001312c9b10 03:07:52 INFO - rsp = 0x00000001312c9ae0 r12 = 0x000000010a33a998 03:07:52 INFO - r13 = 0x000000010a33ab98 r14 = 0x0000000127952308 03:07:52 INFO - r15 = 0x000000012655fe40 rip = 0x0000000106d2b038 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 6 XUL!void mozilla::MozPromise<bool, mozilla::MediaResult, true>::Private::Resolve<bool const&>(bool const&&&, char const*) [MozPromise.h:48b968826893 : 792 + 0x8] 03:07:52 INFO - rbx = 0x000000012655fe40 rbp = 0x00000001312c9b40 03:07:52 INFO - rsp = 0x00000001312c9b20 r12 = 0x000000010a33a998 03:07:52 INFO - r13 = 0x000000010a33ab98 r14 = 0x0000000127952308 03:07:52 INFO - r15 = 0x0000000108ec2daf rip = 0x0000000106d2b834 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 7 XUL!mozilla::MozPromise<bool, mozilla::MediaResult, true>::ChainTo(already_AddRefed<mozilla::MozPromise<bool, mozilla::MediaResult, true>::Private>, char const*) [MozPromise.h:48b968826893 : 749 + 0xf] 03:07:52 INFO - rbx = 0x000000010a33a998 rbp = 0x00000001312c9b80 03:07:52 INFO - rsp = 0x00000001312c9b50 r12 = 0x00000001279522e0 03:07:52 INFO - r13 = 0x000000010a33ab98 r14 = 0x000000012655fe40 03:07:52 INFO - r15 = 0x0000000108ffd899 rip = 0x0000000106d2adef 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 8 XUL!mozilla::detail::ProxyFunctionRunnable<mozilla::MediaFormatReader::DemuxerProxy::NotifyDataArrived()::$_8, mozilla::MozPromise<bool, mozilla::MediaResult, true> >::Run() [MozPromise.h:48b968826893 : 1146 + 0x13] 03:07:52 INFO - rbx = 0x0000000125e41bc8 rbp = 0x00000001312c9bc0 03:07:52 INFO - rsp = 0x00000001312c9b90 r12 = 0x6200ee9782d059e3 03:07:52 INFO - r13 = 0x000000010a33ab98 r14 = 0x000000012cc40a00 03:07:52 INFO - r15 = 0x00000001279522e0 rip = 0x0000000106d2abff 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 9 XUL!mozilla::TaskQueue::Runner::Run() [TaskQueue.cpp:48b968826893 : 232 + 0x9] 03:07:52 INFO - rbx = 0x000000012cfd5550 rbp = 0x00000001312c9c60 03:07:52 INFO - rsp = 0x00000001312c9bd0 r12 = 0x6200ee9782d059e3 03:07:52 INFO - r13 = 0x000000010a33ab98 r14 = 0x000000012cc40a00 03:07:52 INFO - r15 = 0x000000012cc40a30 rip = 0x0000000104ff05ba 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 10 XUL!nsThreadPool::Run() [nsThreadPool.cpp:48b968826893 : 226 + 0x6] 03:07:52 INFO - rbx = 0x000000012fad7d78 rbp = 0x00000001312c9d00 03:07:52 INFO - rsp = 0x00000001312c9c70 r12 = 0x0000000000000000 03:07:52 INFO - r13 = 0x00000001312c9cb8 r14 = 0x000000012fad7d60 03:07:52 INFO - r15 = 0x00000001312c9cc0 rip = 0x0000000104fffc1b 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 11 XUL!non-virtual thunk to nsThreadPool::Run() [nsThreadPool.cpp:48b968826893 : 153 + 0x9] 03:07:52 INFO - rbx = 0x0000000000000000 rbp = 0x00000001312c9d10 03:07:52 INFO - rsp = 0x00000001312c9d10 r12 = 0x0000000130856ac0 03:07:52 INFO - r13 = 0x0000000000000000 r14 = 0x00000001312c9dd7 03:07:52 INFO - r15 = 0x0000000130856af0 rip = 0x0000000104fffd3d 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 12 XUL!nsThread::ProcessNextEvent(bool, bool*) [nsThread.cpp:48b968826893 : 1213 + 0x6] 03:07:52 INFO - rbx = 0x0000000000000000 rbp = 0x00000001312c9dc0 03:07:52 INFO - rsp = 0x00000001312c9d20 r12 = 0x0000000130856ac0 03:07:52 INFO - r13 = 0x0000000000000000 r14 = 0x00000001312c9dd7 03:07:52 INFO - r15 = 0x0000000130856af0 rip = 0x0000000104ff705b 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 13 XUL!NS_ProcessNextEvent(nsIThread*, bool) [nsThreadUtils.cpp:48b968826893 : 381 + 0xd] 03:07:52 INFO - rbx = 0x0000000000000000 rbp = 0x00000001312c9de0 03:07:52 INFO - rsp = 0x00000001312c9dd0 r12 = 0x00000001312c9df8 03:07:52 INFO - r13 = 0x00000000000008ff r14 = 0x000000011bedd800 03:07:52 INFO - r15 = 0x0000000130856ac0 rip = 0x0000000105027523 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 14 XUL!mozilla::ipc::MessagePumpForNonMainThreads::Run(base::MessagePump::Delegate*) [MessagePump.cpp:48b968826893 : 338 + 0xa] 03:07:52 INFO - rbx = 0x000000012cfe5e00 rbp = 0x00000001312c9e40 03:07:52 INFO - rsp = 0x00000001312c9df0 r12 = 0x00000001312c9df8 03:07:52 INFO - r13 = 0x00000000000008ff r14 = 0x000000011bedd800 03:07:52 INFO - r15 = 0x0000000130856ac0 rip = 0x000000010557dffb 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 15 XUL!MessageLoop::Run() [message_loop.cc:48b968826893 : 232 + 0x8] 03:07:52 INFO - rbx = 0x0000000130856ac0 rbp = 0x00000001312c9e70 03:07:52 INFO - rsp = 0x00000001312c9e50 r12 = 0x0000000130856b20 03:07:52 INFO - r13 = 0x00000000000008ff r14 = 0x000000012cfe5e00 03:07:52 INFO - r15 = 0x0000000000000001 rip = 0x0000000105514859 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 16 XUL!nsThread::ThreadFunc(void*) [nsThread.cpp:48b968826893 : 467 + 0x8] 03:07:52 INFO - rbx = 0x0000000130856ac0 rbp = 0x00000001312c9ec0 03:07:52 INFO - rsp = 0x00000001312c9e80 r12 = 0x0000000130856b20 03:07:52 INFO - r13 = 0x00000000000008ff r14 = 0x000000012cfe5e00 03:07:52 INFO - r15 = 0x0000000000000001 rip = 0x0000000104ff523f 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 17 libnss3.dylib!_pt_root [ptthread.c:48b968826893 : 216 + 0x3] 03:07:52 INFO - rbx = 0x000000012f1eb7b0 rbp = 0x00000001312c9ef0 03:07:52 INFO - rsp = 0x00000001312c9ed0 r12 = 0x0000000000015983 03:07:52 INFO - r13 = 0x00000000000008ff r14 = 0x00000001312ca000 03:07:52 INFO - r15 = 0x0000000000000000 rip = 0x0000000104c35e6a 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 18 libsystem_pthread.dylib!_pthread_body + 0x83 03:07:52 INFO - rbx = 0x00000001312ca000 rbp = 0x00000001312c9f10 03:07:52 INFO - rsp = 0x00000001312c9f00 r12 = 0x0000000000015983 03:07:52 INFO - r13 = 0x00000000000008ff r14 = 0x000000012f1eb7b0 03:07:52 INFO - r15 = 0x0000000104c35d90 rip = 0x00007fff9028005a 03:07:52 INFO - Found by: call frame info 03:07:52 INFO - 19 libsystem_pthread.dylib!_pthread_start + 0xb0 03:07:52 INFO - rbp = 0x00000001312c9f50 rsp = 0x00000001312c9f20 03:07:52 INFO - rip = 0x00007fff9027ffd7 03:07:52 INFO - Found by: previous frame's frame pointer 03:07:52 INFO - 20 libsystem_pthread.dylib!thread_start + 0xd 03:07:52 INFO - rbp = 0x00000001312c9f78 rsp = 0x00000001312c9f60 03:07:52 INFO - rip = 0x00007fff9027d3ed 03:07:52 INFO - Found by: previous frame's frame pointer 03:07:52 INFO - 21 libnss3.dylib + 0x135d90 03:07:52 INFO - rsp = 0x00000001312ca030 rip = 0x0000000104c35d90 03:07:52 INFO - Found by: stack scanning

Pushed by jyavenard@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/19c468c32d03 P1. Run demuxing operations on its own task queue. r=jwwang https://hg.mozilla.org/integration/autoland/rev/7df4d36392e7 P2. Update MediaDecoderReader documentation. r=jwwang https://hg.mozilla.org/integration/autoland/rev/909eee913f30 P3. Remove no longer used seeking argument. r=jwwang https://hg.mozilla.org/integration/autoland/rev/4ea0f7d805d9 P4. Use Mutex in place of Monitor. r=jwwang https://hg.mozilla.org/integration/autoland/rev/816eca779bfa P5. Don't attempt to estimate readyState when ended. r=jwwang

https://hg.mozilla.org/mozilla-central/rev/19c468c32d03 https://hg.mozilla.org/mozilla-central/rev/7df4d36392e7 https://hg.mozilla.org/mozilla-central/rev/909eee913f30 https://hg.mozilla.org/mozilla-central/rev/4ea0f7d805d9 https://hg.mozilla.org/mozilla-central/rev/816eca779bfa