Shotdown browser cooperatively with taskill on win platform.

Shotdown browser cooperatively with taskill on win platform.

This way makes sure that all the subprocess are also killed. 
(http://mackeblog.blogspot.com/2012/05/killing-subprocesses-on-windows-in.html)

Stress test gpu_process by setting --pageset-repeat=200 (this will be reverted upon submission)

BUG=424024

[nednguyen, 2014-11-14 14:22:36]

nednguyen@google.com changed reviewers:
+ kbr@chromium.org

[nednguyen, 2014-11-14 16:08:37]

Patchset #1 (id:1) has been deleted

[nednguyen, 2014-11-14 16:08:43]

Patchset #1 (id:20001) has been deleted

[tonyg, 2014-11-14 17:10:27]

tonyg@chromium.org changed reviewers:
+ tonyg@chromium.org

[tonyg, 2014-11-14 17:10:27]

https://codereview.chromium.org/723403003/diff/40001/tools/telemetry/telemetr...
File tools/telemetry/telemetry/core/platform/__init__.py (right):

https://codereview.chromium.org/723403003/diff/40001/tools/telemetry/telemetr...
tools/telemetry/telemetry/core/platform/__init__.py:334: def
CooperativelyShutdown(self, proc):
Should this fit into the *Application() set of methods naming-wise? For
instance:
LaunchApplication()
StopApplication()  # Graceful, cooperative
KillApplication()  # with vengence

[nednguyen, 2014-11-14 17:52:01]

https://codereview.chromium.org/723403003/diff/40001/tools/telemetry/telemetr...
File tools/telemetry/telemetry/core/platform/__init__.py (right):

https://codereview.chromium.org/723403003/diff/40001/tools/telemetry/telemetr...
tools/telemetry/telemetry/core/platform/__init__.py:334: def
CooperativelyShutdown(self, proc):
On 2014/11/14 17:10:27, tonyg wrote:
> Should this fit into the *Application() set of methods naming-wise? For
> instance:
> LaunchApplication()
> StopApplication()  # Graceful, cooperative
> KillApplication()  # with vengence

Done.

[Ken Russell (switch to Gerrit), 2014-11-14 18:51:38]

not lgtm

https://codereview.chromium.org/723403003/diff/60001/tools/telemetry/telemetr...
File tools/telemetry/telemetry/core/platform/win_platform_backend.py (right):

https://codereview.chromium.org/723403003/diff/60001/tools/telemetry/telemetr...
tools/telemetry/telemetry/core/platform/win_platform_backend.py:371:
subprocess.call(['taskkill', '/F', '/T', '/PID', str(proc.pid)])
This is far from a cooperative shutdown; it forcibly kills the process tree. The
approach in https://codereview.chromium.org/726563003 is much closer to the how
the browser shuts down on users' machines. I am continuing to refine that patch.

Further, this approach will mask any real bugs in the browser causing
sub-processes to leak.

[Ken Russell (switch to Gerrit), 2014-11-14 18:59:02]

Also note that this approach, for whatever reason, didn't actually fix the
problem.
http://build.chromium.org/p/tryserver.chromium.gpu/builders/win_gpu_triggered...
exhibits the same problem of a leaked sub-process (in this case, a GPU process).

[nednguyen, 2014-11-14 22:09:18]

On 2014/11/14 18:51:38, Ken Russell wrote:
> not lgtm
> 
>
https://codereview.chromium.org/723403003/diff/60001/tools/telemetry/telemetr...
> File tools/telemetry/telemetry/core/platform/win_platform_backend.py (right):
> 
>
https://codereview.chromium.org/723403003/diff/60001/tools/telemetry/telemetr...
> tools/telemetry/telemetry/core/platform/win_platform_backend.py:371:
> subprocess.call(['taskkill', '/F', '/T', '/PID', str(proc.pid)])
> This is far from a cooperative shutdown; it forcibly kills the process tree.
The
> approach in https://codereview.chromium.org/726563003 is much closer to the
how
> the browser shuts down on users' machines. I am continuing to refine that
patch.
> 
> Further, this approach will mask any real bugs in the browser causing
> sub-processes to leak.

I see. I will close this patch.