Freeze media time and audio rendering when the system suspends.

Freeze media time and audio rendering when the system suspends.

Allows us to avoid unexpected differences between what the audio
renderer and video renderer think the current time is. Causes the
wall clock time calculations to return false for "is time moving."

With this information the video renderer (and other callers) can
accurately act upon frozen moments and not expire frames due to
a difference of opinion on what the current time is (audio only
updates its base every Render() call).

BUG=620478
TEST=new unittests

Committed: https://crrev.com/e3fb958aa538c5b146afea7f9df498a579d3e5a4
Cr-Commit-Position: refs/heads/master@{#400454}

[DaleCurtis, 2016-06-16 19:36:54]

Description was changed from

==========
Freeze time when the system suspends.

Allows us to avoid unexpected differences between what the audio
renderer and video renderer think the current time is. Causes the
wall clock time calculations to return false for "is time moving."

With this information the video renderer (and other callers) can
accurately act upon frozen moments and not expire frames due to
a difference of opinion on what the current time is (audio only
updates its base every Render() call).

BUG=620478
TEST=new unittests
==========

to

==========
Freeze media time and audio rendering when the system suspends.

Allows us to avoid unexpected differences between what the audio
renderer and video renderer think the current time is. Causes the
wall clock time calculations to return false for "is time moving."

With this information the video renderer (and other callers) can
accurately act upon frozen moments and not expire frames due to
a difference of opinion on what the current time is (audio only
updates its base every Render() call).

BUG=620478
TEST=new unittests
==========

[DaleCurtis, 2016-06-16 19:37:00]

Patchset #1 (id:1) has been deleted

[DaleCurtis, 2016-06-16 19:37:32]

dalecurtis@chromium.org changed reviewers:
+ xhwang@chromium.org

[DaleCurtis, 2016-06-16 19:38:14]

dalecurtis@chromium.org changed reviewers:
+ sandersd@chromium.org

[DaleCurtis, 2016-06-16 19:38:15]

xhwang@ is out for GoogleServe, so +sandersd@ for review.

[sandersd (OOO until July 31), 2016-06-16 19:46:11]

lgtm

[DaleCurtis, 2016-06-16 20:21:19]

The CQ bit was checked by dalecurtis@chromium.org

[commit-bot: I haz the power, 2016-06-16 20:22:07]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2072843002/20001

[commit-bot: I haz the power, 2016-06-16 22:06:18]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-06-16 22:06:19]

Try jobs failed on following builders:
  linux_chromium_rel_ng on tryserver.chromium.linux (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_...)

[DaleCurtis, 2016-06-16 22:12:33]

The CQ bit was checked by dalecurtis@chromium.org

[commit-bot: I haz the power, 2016-06-16 22:13:39]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2072843002/20001

[xhwang, 2016-06-16 22:16:38]

nice, LGTM

I am gonna test this on Netflix as well.

[DaleCurtis, 2016-06-16 23:24:16]

Hmm, looks like this is breaking WebGL conformance tests somehow... will take a
closer look.

[DaleCurtis, 2016-06-16 23:40:37]

dalecurtis@chromium.org changed reviewers:
+ kbr@chromium.org

[DaleCurtis, 2016-06-16 23:40:38]

Hmm, I can't repro any failures locally with the conformance tests on my mac.

+kbr: are there any known issues with the GPU bots and suspend/resume behavior?
This CL suspends the audio/video clock when the power observer indicates that
the system is suspending.

[xhwang, 2016-06-16 23:43:15]

On 2016/06/16 23:40:38, DaleCurtis wrote:
> Hmm, I can't repro any failures locally with the conformance tests on my mac.
> 
> +kbr: are there any known issues with the GPU bots and suspend/resume
behavior?
> This CL suspends the audio/video clock when the power observer indicates that
> the system is suspending.

Actually, how about video-only? Do we need to do something similar in
WallClockTimeSource?

[DaleCurtis, 2016-06-16 23:44:35]

I don't think video-only can suffer from the same problems, but I'll be sure to
test before landing.

[Ken Russell (switch to Gerrit), 2016-06-16 23:54:49]

On 2016/06/16 23:40:38, DaleCurtis wrote:
> Hmm, I can't repro any failures locally with the conformance tests on my mac.
> 
> +kbr: are there any known issues with the GPU bots and suspend/resume
behavior?
> This CL suspends the audio/video clock when the power observer indicates that
> the system is suspending.

No, the GPU bots shouldn't suspend or resume.

Did you look at the stack trace in the log? It looks like a basic logic bug.

[644:1299:0616/151708:FATAL:observer_list.h(164)] Check failed: false. Observers
can only be added once!
0   Chromium Framework                  0x000000010cc15ce3
_ZN4base5debug10StackTraceC1Ev + 19
1   Chromium Framework                  0x000000010cc347c7
_ZN7logging10LogMessageD2Ev + 71
2   Chromium Framework                  0x000000010cc5c309
_ZN4base16ObserverListBaseINS_13PowerObserverEE11AddObserverEPS1_ + 217
3   Chromium Framework                  0x000000010cc5b6b2
_ZN4base22ObserverListThreadSafeINS_13PowerObserverEE11AddObserverEPS1_ + 354
4   Chromium Framework                  0x000000010df69983
_ZN5media17AudioRendererImplC2ERK13scoped_refptrIN4base22SingleThreadTaskRunnerEEPNS_17AudioRendererSinkE12ScopedVectorINS_12AudioDecoderEERKS1_INS_8MediaLogEE
+ 963
5   Chromium Framework                  0x000000010df6f5e0
_ZN5media22DefaultRendererFactory14CreateRendererERK13scoped_refptrIN4base22SingleThreadTaskRunnerEERKS1_INS2_10TaskRunnerEEPNS_17AudioRendererSinkEPNS_17VideoRendererSinkERKNS2_8CallbackIFvRKNSF_IFviELNS2_8internal8CopyModeE1EEEELSI_1EEE
+ 176
6   Chromium Framework                  0x0000000111ed2336
_ZN5media18WebMediaPlayerImpl14CreateRendererEv + 86
7   Chromium Framework                  0x0000000111ed945a
_ZN4base8internal7InvokerINS_13IndexSequenceIJLm0EEEENS0_9BindStateINS0_15RunnableAdapterIMN5media18WebMediaPlayerImplEFNSt3__110unique_ptrINS6_8RendererENS8_14default_deleteISA_EEEEvEEEFSD_PS7_EJNS0_17UnretainedWrapperIS7_EEEEELb0EFSD_vEE3RunEPNS0_13BindStateBaseE
+ 42
8   Chromium Framework                  0x000000010df39f6d
_ZN5media18PipelineController5StartEPNS_7DemuxerEPNS_8Pipeline6ClientEbb + 301
9   Chromium Framework                  0x0000000111ed3f62
_ZN5media18WebMediaPlayerImpl13StartPipelineEv + 1954
10  Chromium Framework                  0x0000000111ed419b
_ZN5media18WebMediaPlayerImpl21DataSourceInitializedEb + 219
11  Chromium Framework                  0x000000010c8fbe8c
_ZN4base8internal7InvokerINS_13IndexSequenceIJLm0EEEENS0_9BindStateINS_8CallbackIFvbELNS0_8CopyModeE1EEES6_JRbEEELb0EFvvEE3RunEPNS0_13BindStateBaseE
+ 28
12  Chromium Framework                  0x000000010cc1620b
_ZN4base5debug13TaskAnnotator7RunTaskEPKcRKNS_11PendingTaskE + 187
13  Chromium Framework                  0x0000000111079aab
_ZN9scheduler16TaskQueueManager24ProcessTaskFromWorkQueueEPNS_8internal9WorkQueueEPNS1_13TaskQueueImpl4TaskE
+ 811
14  Chromium Framework                  0x00000001110786bb
_ZN9scheduler16TaskQueueManager6DoWorkEN4base9TimeTicksEb + 507
15  Chromium Framework                  0x000000010cc1620b
_ZN4base5debug13TaskAnnotator7RunTaskEPKcRKNS_11PendingTaskE + 187
16  Chromium Framework                  0x000000010cc453dc
_ZN4base11MessageLoop7RunTaskERKNS_11PendingTaskE + 572
17  Chromium Framework                  0x000000010cc456dc
_ZN4base11MessageLoop21DeferOrRunPendingTaskERKNS_11PendingTaskE + 44
18  Chromium Framework                  0x000000010cc45a2b
_ZN4base11MessageLoop6DoWorkEv + 299
19  Chromium Framework                  0x000000010cc07943
_ZN4base24MessagePumpCFRunLoopBase7RunWorkEv + 51
20  Chromium Framework                  0x000000010cc359fa
_ZN4base3mac15CallWithEHFrameEU13block_pointerFvvE + 10
21  Chromium Framework                  0x000000010cc072f4
_ZN4base24MessagePumpCFRunLoopBase13RunWorkSourceEPv + 68
22  CoreFoundation                      0x00007fff97f90a01
__CFRUNLOOP_IS_CALLING_OUT_TO_A_SOURCE0_PERFORM_FUNCTION__ + 17
23  CoreFoundation                      0x00007fff97f82b8d __CFRunLoopDoSources0
+ 269
24  CoreFoundation                      0x00007fff97f821bf __CFRunLoopRun + 927
25  CoreFoundation                      0x00007fff97f81bd8 CFRunLoopRunSpecific
+ 296
26  Foundation                          0x00007fff93daeb29
-[NSRunLoop(NSRunLoop) runMode:beforeDate:] + 278
27  Chromium Framework                  0x000000010cc0801e
_ZN4base20MessagePumpNSRunLoop5DoRunEPNS_11MessagePump8DelegateE + 126
28  Chromium Framework                  0x000000010cc0777f
_ZN4base24MessagePumpCFRunLoopBase3RunEPNS_11MessagePump8DelegateE + 127
29  Chromium Framework                  0x000000010cc44f37
_ZN4base11MessageLoop10RunHandlerEv + 215
30  Chromium Framework                  0x000000010cc65333
_ZN4base7RunLoop3RunEv + 51
31  Chromium Framework                  0x000000010cc440e7
_ZN4base11MessageLoop3RunEv + 103
32  Chromium Framework                  0x0000000111d8cd0a
_ZN7content12RendererMainERKNS_18MainFunctionParamsE + 730
33  Chromium Framework                  0x000000010cbbf3e2
_ZN7content21ContentMainRunnerImpl3RunEv + 802
34  Chromium Framework                  0x000000010cbbe426
_ZN7content11ContentMainERKNS_17ContentMainParamsE + 54
35  Chromium Framework                  0x000000010c59b90a ChromeMain + 58
36  Chromium Helper                     0x000000010c534d62 main + 530
37  Chromium Helper                     0x000000010c534b44 start + 52

[commit-bot: I haz the power, 2016-06-16 23:59:33]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-06-16 23:59:34]

Try jobs failed on following builders:
  mac_chromium_rel_ng on tryserver.chromium.mac (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_rel_...)

[DaleCurtis, 2016-06-17 00:13:35]

No, I didn't see those in the logs, thanks for pointing them out. Even after
you've pointed that error out, I don't see how you found that in those massive
logs... The error didn't seem listed around the test failures when I looked. How
did you find it?

[Ken Russell (switch to Gerrit), 2016-06-17 00:17:22]

On 2016/06/17 00:13:35, DaleCurtis wrote:
> No, I didn't see those in the logs, thanks for pointing them out. Even after
> you've pointed that error out, I don't see how you found that in those massive
> logs... The error didn't seem listed around the test failures when I looked.
How
> did you find it?

I opened "stdio" for "webgl_conformance_tests on Intel GPU on Mac (with patch)
on Mac-10.10" and searched for one of the failing test names
("conformance_textures_video_tex_2d_rgba_rgba_unsigned_byte").

[DaleCurtis, 2016-06-17 00:20:47]

dalecurtis@chromium.org changed reviewers:
- kbr@chromium.org

[DaleCurtis, 2016-06-17 00:20:48]

I did the same one with a different test, but didn't pull that crash up. Thanks
in any case. Sorry for the noise.

The crash is still a mystery though since we add/remove the observer in
construction/destruction. It seems like the state of PowerMonitor::Get() must be
changing...

[DaleCurtis, 2016-06-17 01:21:45]

dalecurtis@chromium.org changed reviewers:
+ kbr@chromium.org

[DaleCurtis, 2016-06-17 01:21:45]

Issue was adding the observer on the render thread while removing it on the
media thread. The observer list used by the power observer indexes by thread id.

[DaleCurtis, 2016-06-17 01:22:30]

The CQ bit was checked by dalecurtis@chromium.org

[DaleCurtis, 2016-06-17 01:22:31]

The patchset sent to the CQ was uploaded after l-g-t-m from xhwang@chromium.org,
sandersd@chromium.org
Link to the patchset: https://codereview.chromium.org/2072843002/#ps40001
(title: "Add observer on |task_runner_|.")

[commit-bot: I haz the power, 2016-06-17 01:22:50]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2072843002/40001

[commit-bot: I haz the power, 2016-06-17 01:37:13]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2016-06-17 01:37:14]

Try jobs failed on following builders:
  chromeos_daisy_chromium_compile_only_ng on tryserver.chromium.linux
(JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.linux/builders/chromeos_daisy_...)

[xhwang, 2016-06-17 07:12:52]

OOC, did you test the video-only case?

https://chromiumcodereview.appspot.com/2072843002/diff/40001/media/renderers/...
File media/renderers/audio_renderer_impl_unittest.cc (right):

https://chromiumcodereview.appspot.com/2072843002/diff/40001/media/renderers/...
media/renderers/audio_renderer_impl_unittest.cc:765: renderer_->OnSuspend();
Can you actually use TestPowerMonitorSource so we are closer to the real use
case? e.g.

https://cs.chromium.org/chromium/src/net/url_request/url_request_unittest.cc?...

[DaleCurtis, 2016-06-17 16:41:52]

Did you ever see the video only case hang? I could not reproduce issues.

I don't think the test power source adds anything interesting it'll just proxy
to our methods and we still need the null check for other tests using ari
indirectly.

[DaleCurtis, 2016-06-17 17:47:40]

The CQ bit was checked by dalecurtis@chromium.org

[commit-bot: I haz the power, 2016-06-17 17:48:15]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/2072843002/40001

[commit-bot: I haz the power, 2016-06-17 18:03:35]

Committed patchset #2 (id:40001)

[commit-bot: I haz the power, 2016-06-17 18:04:57]

Description was changed from

==========
Freeze media time and audio rendering when the system suspends.

Allows us to avoid unexpected differences between what the audio
renderer and video renderer think the current time is. Causes the
wall clock time calculations to return false for "is time moving."

With this information the video renderer (and other callers) can
accurately act upon frozen moments and not expire frames due to
a difference of opinion on what the current time is (audio only
updates its base every Render() call).

BUG=620478
TEST=new unittests
==========

to

==========
Freeze media time and audio rendering when the system suspends.

Allows us to avoid unexpected differences between what the audio
renderer and video renderer think the current time is. Causes the
wall clock time calculations to return false for "is time moving."

With this information the video renderer (and other callers) can
accurately act upon frozen moments and not expire frames due to
a difference of opinion on what the current time is (audio only
updates its base every Render() call).

BUG=620478
TEST=new unittests

Committed: https://crrev.com/e3fb958aa538c5b146afea7f9df498a579d3e5a4
Cr-Commit-Position: refs/heads/master@{#400454}
==========

[commit-bot: I haz the power, 2016-06-17 18:04:58]

Patchset 2 (id:??) landed as
https://crrev.com/e3fb958aa538c5b146afea7f9df498a579d3e5a4
Cr-Commit-Position: refs/heads/master@{#400454}