Change g_webrtc_logging_delegate from process-level to RenderThread-level

Change g_webrtc_logging_delegate from process-level to RenderThread-level

--single-process mode may happen crash in file of webrtc_logging.cc where code
is CHECK(!g_webrtc_logging_delegate). The orginal code is to expect there to
be only one g_webrtc_logging_delegate instance in a single render process.
g_webrtc_logging_delegate is created when a RenderThreadImpl is created.
But in --single-process model all RenderThreadImpl are created in browser
process and so g_webrtc_logging_delegate may be created many times in a single
process. Then CHECK(!g_webrtc_logging_delegate) will fail.

In this patch will change g_webrtc_logging_delegate from process-level to
RenderThread-Level. Since every render process will only keep one RenderThread.

BUG=455573
Committed: https://crrev.com/1fb332dc1f81bb05716342704b7f969af6921f3e
Cr-Commit-Position: refs/heads/master@{#315131}

[jinlong.zhai, 2015-02-06 02:40:34]

jinlong.zhai@samsung.com changed reviewers:
+ brettw@chromium.org, dcheng@chromium.org, tommi@chromium.org

[jinlong.zhai, 2015-02-06 02:40:35]

Please help to review the patch.

[dcheng, 2015-02-06 20:10:24]

dcheng@chromium.org changed reviewers:
- brettw@chromium.org

[dcheng, 2015-02-06 20:10:25]

-brettw, -dcheng as reviewers.

I'd recommend picking just the reviewers you need (which in this case is
tommi@). With multiple reviewers, it increases the chances that the listed
reviewers all think that another reviewer will look at it =)

[dcheng, 2015-02-06 20:10:47]

dcheng@chromium.org changed reviewers:
- dcheng@chromium.org

[tommi (sloooow) - chröme, 2015-02-06 22:20:20]

The CQ bit was checked by tommi@chromium.org

[tommi (sloooow) - chröme, 2015-02-06 22:20:21]

lgtm

[commit-bot: I haz the power, 2015-02-06 22:20:39]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/889313003/1

[commit-bot: I haz the power, 2015-02-06 23:20:14]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-02-06 23:21:31]

Patchset 1 (id:??) landed as
https://crrev.com/1fb332dc1f81bb05716342704b7f969af6921f3e
Cr-Commit-Position: refs/heads/master@{#315131}

[tommi (sloooow) - chröme, 2015-02-16 21:12:42]

A revert of this CL (patchset #1 id:1) has been created in
https://codereview.chromium.org/931693002/ by tommi@chromium.org.

The reason for reverting is: This breaks webrtc logging.  It changes a global
pointer to be a tls pointer, which doesn't work..

[pbos, 2015-02-16 22:01:27]

On 2015/02/16 21:12:42, tommi wrote:
> A revert of this CL (patchset #1 id:1) has been created in
> https://codereview.chromium.org/931693002/ by mailto:tommi@chromium.org.
> 
> The reason for reverting is: This breaks webrtc logging.  It changes a global
> pointer to be a tls pointer, which doesn't work..

An idea for re-landing this: Shouldn't we be able to have a ref-counted global
instance instead, or does the logging delegate require internal locking that's
not present if the instance is to be shared?

This is just drive-by I'm not very familiar with global lazy initialization in
Chromium. :)

[jinlong.zhai, 2015-02-27 06:45:38]

On 2015/02/16 22:01:27, pbos wrote:
> On 2015/02/16 21:12:42, tommi wrote:
> > A revert of this CL (patchset #1 id:1) has been created in
> > https://codereview.chromium.org/931693002/ by mailto:tommi@chromium.org.
> > 
> > The reason for reverting is: This breaks webrtc logging.  It changes a
global
> > pointer to be a tls pointer, which doesn't work..
> 
> An idea for re-landing this: Shouldn't we be able to have a ref-counted global
> instance instead, or does the logging delegate require internal locking that's
> not present if the instance is to be shared?
> 
> This is just drive-by I'm not very familiar with global lazy initialization in
> Chromium. :)

I think the orginal logic is to expect there to be single instance for one
render process,
but it is not suitable for --single-process and also logging delegate can not to
be as
a RenderThread-level instance. I think the code is right execpt for
CHECK(!g_webrtc_logging_delegate).

So I suggest we just remove CHECK(!g_webrtc_logging_delegate).
in_reply_to: 5738600293466112
send_mail: 1
subject: Change g_webrtc_logging_delegate from process-level to
RenderThread-level