Reland: Use a shared thread for media operations (round 3).

Reland: Use a shared thread for media operations (round 3).

This reduces the number of threads created per media element by one.

TBR=jamesr,jochen
BUG=61293
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=204386

[scherkus (not reviewing), 2013-06-04 20:17:12]

acolwell: PS#1 is original CL, PS#2 contains my memory corruption fix as
detected by asan

check out http://crbug.com/245941 for some of the investigation

[scherkus (not reviewing), 2013-06-04 20:19:17]

also ... in The Real World using WMPI it's a race between the main thread waking
up and deleting pipeline, so while my unit test doesn't quite cover the exact
situation I feel it reliably reproduces the bug w/o getting into constructing a
test that involves multiple threads

tl;dr version: I'm open to ideas for tests and/or changes/improvements to
pipeline

[acolwell GONE FROM CHROMIUM, 2013-06-05 01:09:05]

LGTM % comment.

https://codereview.chromium.org/15993018/diff/2001/media/base/pipeline_unitte...
File media/base/pipeline_unittest.cc (right):

https://codereview.chromium.org/15993018/diff/2001/media/base/pipeline_unitte...
media/base/pipeline_unittest.cc:277: if (demuxer_)
nit: Add { } here and below since these take up multiple lines.

https://codereview.chromium.org/15993018/diff/2001/media/base/pipeline_unitte...
media/base/pipeline_unittest.cc:855: pipeline_->Stop(base::Bind(&DeletePipeline,
base::Passed(&pipeline_)));
Whoa... how is this not crashing? Isn't the Passed clearing pipeline_? I think
you need to make a local copy of the pipeline pointer to avoid undefined
compiler behavior.

[scherkus (not reviewing), 2013-06-05 18:40:43]

https://codereview.chromium.org/15993018/diff/2001/media/base/pipeline_unitte...
File media/base/pipeline_unittest.cc (right):

https://codereview.chromium.org/15993018/diff/2001/media/base/pipeline_unitte...
media/base/pipeline_unittest.cc:277: if (demuxer_)
On 2013/06/05 01:09:05, acolwell wrote:
> nit: Add { } here and below since these take up multiple lines.

actually all of these can fit on a single line

https://codereview.chromium.org/15993018/diff/2001/media/base/pipeline_unitte...
media/base/pipeline_unittest.cc:855: pipeline_->Stop(base::Bind(&DeletePipeline,
base::Passed(&pipeline_)));
On 2013/06/05 01:09:05, acolwell wrote:
> Whoa... how is this not crashing? Isn't the Passed clearing pipeline_? I think
> you need to make a local copy of the pipeline pointer to avoid undefined
> compiler behavior.

Turns out it _is_ crashing! (check out try bot results)

[commit-bot: I haz the power, 2013-06-05 18:42:01]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/scherkus@chromium.org/15993018/17001

[commit-bot: I haz the power, 2013-06-06 02:23:34]

Change committed as 204386