Defer OSX output stream Start() around system supend and resume.

Defer OSX output stream Start() around system supend and resume.

Uses base::PowerMonitor to watch for suspend and resume events and
disables stream creation during suspend and for a second after a
resume is detected.

Since Start() is asynchronous anyways, this is pretty self contained
change.  If Stop() occurs before the deferred Start(), the pending
task is simply canceled.

This method has the advantage of preemptively preventing stream hangs
instead of trying to correct them after the fact.  It's also far less
invasive than the "restart all streams" workaround.  A risk is that
the post-resume delay is not long enough.

BUG=160920
TEST=repeated start/stop no longer fails.
NOTRY=true

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=247026

[DaleCurtis, 2014-01-16 23:04:46]

Here's the alternate workaround I discussed previously.  It resolves the issue
when tested with my unittest based repro case.

In the field it'll be easy to tell if this helps by watching the
Media.AudioOutputControllerPlaybackStartupSuccess UMA stat go down (which the
current workaround does not help).

[scherkus (not reviewing), 2014-01-17 17:53:01]

Is the idea to remove the other workaround, or to keep both in place as a safety
net?

[DaleCurtis, 2014-01-17 19:25:34]

Ideally I'd remove the other workaround, but they could live side by side.  As
mentioned previously, I was planning to watch the UMA stat to see how effective
this workaround is.  However I just looked and we're sitting at 99.20% success
on stable.  Which is great, but likely means the current workaround is reducing
failure rates by reducing the amount of time a given system is in the broken
state.

https://uma.googleplex.com/histograms/?version=32.0.1700.76-M&histograms=Medi...

There should still be an improvement if this workaround is better, it just won't
be as dramatic as I was looking for in the numbers.

[scherkus (not reviewing), 2014-01-17 23:04:45]

bombs away!

lgtm

[scherkus (not reviewing), 2014-01-17 23:04:46]

bombs away!

lgtm

[commit-bot: I haz the power, 2014-01-17 23:34:33]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dalecurtis@chromium.org/135853022/1

[commit-bot: I haz the power, 2014-01-18 00:15:46]

Retried try job too often on mac_rel for step(s) media_unittests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=mac_rel&nu...

[commit-bot: I haz the power, 2014-01-18 01:30:24]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dalecurtis@chromium.org/135853022/290001

[commit-bot: I haz the power, 2014-01-18 02:06:09]

Retried try job too often on mac_rel for step(s) media_unittests
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=mac_rel&nu...

[DaleCurtis, 2014-01-22 01:42:40]

PTAL.  I had to rewrite the AUHAL test to move them onto the audio thread to
avoid DCHECK(audio thread) failures.

[commit-bot: I haz the power, 2014-01-25 00:05:26]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dalecurtis@chromium.org/135853022/560001

[commit-bot: I haz the power, 2014-01-25 00:11:21]

Change committed as 247026