Replace MessageLoop::current() with ThreadTaskRunnerHandle::Get() in GCM

Replace MessageLoop::current() with ThreadTaskRunnerHandle::Get() in GCM

Usages of MessageLoop::current() in GCM production code have been
replaced with ThreadTaskRunnerHandle::Get(). Usages of MessageLoop
in some test codes have also been replaced, but not all since we
don't have alternative to replace RunLoop.

BUG=none
TEST=none

Committed: https://crrev.com/1389abf29e75c80041b80dbbf7a2f7e6bb8233fb
Cr-Commit-Position: refs/heads/master@{#338401}

[jianli, 2015-07-10 22:22:45]

jianli@chromium.org changed reviewers:
+ fgorski@chromium.org, zea@chromium.org

[fgorski, 2015-07-10 22:30:24]

lgtm

[Nicolas Zea, 2015-07-10 22:39:38]

LGTM with a comment

https://codereview.chromium.org/1232193002/diff/1/google_apis/gcm/base/mcs_me...
File google_apis/gcm/base/mcs_message_unittest.cc (right):

https://codereview.chromium.org/1232193002/diff/1/google_apis/gcm/base/mcs_me...
google_apis/gcm/base/mcs_message_unittest.cc:31:
task_runner_handle_(task_runner_) {
Do we need the handle here (and in the other tests)?

[jianli, 2015-07-10 23:04:15]

https://codereview.chromium.org/1232193002/diff/1/google_apis/gcm/base/mcs_me...
File google_apis/gcm/base/mcs_message_unittest.cc (right):

https://codereview.chromium.org/1232193002/diff/1/google_apis/gcm/base/mcs_me...
google_apis/gcm/base/mcs_message_unittest.cc:31:
task_runner_handle_(task_runner_) {
On 2015/07/10 22:39:38, Nicolas Zea wrote:
> Do we need the handle here (and in the other tests)?

Yes, this is to replace the task runner set in MessageLoop that is being
removed.

[Nicolas Zea, 2015-07-10 23:06:39]

https://codereview.chromium.org/1232193002/diff/1/google_apis/gcm/base/mcs_me...
File google_apis/gcm/base/mcs_message_unittest.cc (right):

https://codereview.chromium.org/1232193002/diff/1/google_apis/gcm/base/mcs_me...
google_apis/gcm/base/mcs_message_unittest.cc:31:
task_runner_handle_(task_runner_) {
On 2015/07/10 23:04:15, jianli wrote:
> On 2015/07/10 22:39:38, Nicolas Zea wrote:
> > Do we need the handle here (and in the other tests)?
> 
> Yes, this is to replace the task runner set in MessageLoop that is being
> removed.

AFAICT this is just storing another reference to task_runner_ though, which
isn't being used. task_runner_ itself will keep the task runner alive, why do we
need a second reference?

[jianli, 2015-07-10 23:11:43]

https://codereview.chromium.org/1232193002/diff/1/google_apis/gcm/base/mcs_me...
File google_apis/gcm/base/mcs_message_unittest.cc (right):

https://codereview.chromium.org/1232193002/diff/1/google_apis/gcm/base/mcs_me...
google_apis/gcm/base/mcs_message_unittest.cc:31:
task_runner_handle_(task_runner_) {
On 2015/07/10 23:06:39, Nicolas Zea wrote:
> On 2015/07/10 23:04:15, jianli wrote:
> > On 2015/07/10 22:39:38, Nicolas Zea wrote:
> > > Do we need the handle here (and in the other tests)?
> > 
> > Yes, this is to replace the task runner set in MessageLoop that is being
> > removed.
> 
> AFAICT this is just storing another reference to task_runner_ though, which
> isn't being used. task_runner_ itself will keep the task runner alive, why do
we
> need a second reference?

This is to make ThreadTaskRunnerHandle::Get() work. Otherwise the reference to
TaskRunner instance is not set in TLS.

[Nicolas Zea, 2015-07-10 23:16:16]

On 2015/07/10 23:11:43, jianli wrote:
>
https://codereview.chromium.org/1232193002/diff/1/google_apis/gcm/base/mcs_me...
> File google_apis/gcm/base/mcs_message_unittest.cc (right):
> 
>
https://codereview.chromium.org/1232193002/diff/1/google_apis/gcm/base/mcs_me...
> google_apis/gcm/base/mcs_message_unittest.cc:31:
> task_runner_handle_(task_runner_) {
> On 2015/07/10 23:06:39, Nicolas Zea wrote:
> > On 2015/07/10 23:04:15, jianli wrote:
> > > On 2015/07/10 22:39:38, Nicolas Zea wrote:
> > > > Do we need the handle here (and in the other tests)?
> > > 
> > > Yes, this is to replace the task runner set in MessageLoop that is being
> > > removed.
> > 
> > AFAICT this is just storing another reference to task_runner_ though, which
> > isn't being used. task_runner_ itself will keep the task runner alive, why
do
> we
> > need a second reference?
> 
> This is to make ThreadTaskRunnerHandle::Get() work. Otherwise the reference to
> TaskRunner instance is not set in TLS.

Ah, makes sense. Thanks for clearing that up!

[jianli, 2015-07-10 23:25:58]

The CQ bit was checked by jianli@chromium.org

[commit-bot: I haz the power, 2015-07-10 23:27:43]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1232193002/1

[commit-bot: I haz the power, 2015-07-10 23:59:41]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-07-11 00:00:37]

Patchset 1 (id:??) landed as
https://crrev.com/1389abf29e75c80041b80dbbf7a2f7e6bb8233fb
Cr-Commit-Position: refs/heads/master@{#338401}