Update test.dart to do the same on all platforms for tests run

Update test.dart to do the same on all platforms for tests run
through selenium.

Remove the explicit exit(0) which can hide other issues. Cancel
the kill timer when the process terminates gracefully (which was
one of those issues hidden by the explicit exit(0)).

R=efortuna@google.com,gram@google.com,ahe@google.com,ricow@google.com
BUG=

Committed: https://code.google.com/p/dart/source/detail?r=13396

[Mads Ager (google), 2012-10-09 06:48:59]

Sorry for breaking the selenium tests on Windows, Emily. More proof that we all
(especially me) need careful reviews of test.dart changes. :-)

[ahe, 2012-10-09 07:03:17]

LGTM!

[Emily Fortuna, 2012-10-09 17:52:06]

lgtm... how do you explain the change in test status from 
http://chromegw.corp.google.com/i/client.dart/builders/dart2js-ie-win7-html/b...

and then this :
http://chromegw.corp.google.com/i/client.dart/builders/dart2js-ie-win7-html/b...

You can see in the previous one, no tests failed, but because it didn't
explicitly return an exit code of 0, the bot turned "red"

(??)

[Jennifer Messerly, 2012-10-09 17:54:56]

https://codereview.chromium.org/11086019/diff/1/tools/testing/dart/test_runne...
File tools/testing/dart/test_runner.dart (right):

https://codereview.chromium.org/11086019/diff/1/tools/testing/dart/test_runne...
tools/testing/dart/test_runner.dart:761: // the open browser processes. On
Windows, signals do not exist
I'm confused by this comment. See:
http://msdn.microsoft.com/en-us/library/dwwzkt4c(v=vs.100).aspx
and http://msdn.microsoft.com/en-us/library/xdkz3x12(v=vs.100).aspx

Granted, I haven't tried this in a long time, but IIRC Python on Windows has
basic signal support.

[Emily Fortuna, 2012-10-09 18:01:41]

https://codereview.chromium.org/11086019/diff/1/tools/testing/dart/test_runne...
File tools/testing/dart/test_runner.dart (right):

https://codereview.chromium.org/11086019/diff/1/tools/testing/dart/test_runne...
tools/testing/dart/test_runner.dart:761: // the open browser processes. On
Windows, signals do not exist
On 2012/10/09 17:54:56, John Messerly wrote:
> I'm confused by this comment. See:
> http://msdn.microsoft.com/en-us/library/dwwzkt4c%28v=vs.100%29.aspx
> and http://msdn.microsoft.com/en-us/library/xdkz3x12%28v=vs.100%29.aspx
> 
> Granted, I haven't tried this in a long time, but IIRC Python on Windows has
> basic signal support.

It does
http://docs.python.org/library/subprocess.html#subprocess.Popen.send_signal (and
also the Python win32 API)

Dart just has to send the right information when closing the process, which it
is currently not doing: http://code.google.com/p/dart/source/detail?r=8541

[Jennifer Messerly, 2012-10-09 18:06:28]

On 2012/10/09 18:01:41, Emily Fortuna wrote:
>
https://codereview.chromium.org/11086019/diff/1/tools/testing/dart/test_runne...
> File tools/testing/dart/test_runner.dart (right):
> 
>
https://codereview.chromium.org/11086019/diff/1/tools/testing/dart/test_runne...
> tools/testing/dart/test_runner.dart:761: // the open browser processes. On
> Windows, signals do not exist
> On 2012/10/09 17:54:56, John Messerly wrote:
> > I'm confused by this comment. See:
> > http://msdn.microsoft.com/en-us/library/dwwzkt4c%2528v=vs.100%2529.aspx
> > and http://msdn.microsoft.com/en-us/library/xdkz3x12%2528v=vs.100%2529.aspx
> > 
> > Granted, I haven't tried this in a long time, but IIRC Python on Windows has
> > basic signal support.
> 
> It does
> http://docs.python.org/library/subprocess.html#subprocess.Popen.send_signal
(and
> also the Python win32 API)
> 
> Dart just has to send the right information when closing the process, which it
> is currently not doing: http://code.google.com/p/dart/source/detail?r=8541

Makes me wonder what kind of spooky Win32 magic Python is doing to make that
work :)

in particular the note about creationflags: CREATE_NEW_PROCESS_GROUP makes me
wonder if they're using "CreateJobObject" or something similar. That's one of
the ways to get a process group to terminate together.

Sigh, Windows :\

[Mads Ager (google), 2012-10-09 18:56:17]

On 2012/10/09 18:06:28, John Messerly wrote:
> On 2012/10/09 18:01:41, Emily Fortuna wrote:
> >
>
https://codereview.chromium.org/11086019/diff/1/tools/testing/dart/test_runne...
> > File tools/testing/dart/test_runner.dart (right):
> > 
> >
>
https://codereview.chromium.org/11086019/diff/1/tools/testing/dart/test_runne...
> > tools/testing/dart/test_runner.dart:761: // the open browser processes. On
> > Windows, signals do not exist
> > On 2012/10/09 17:54:56, John Messerly wrote:
> > > I'm confused by this comment. See:
> > >
http://msdn.microsoft.com/en-us/library/dwwzkt4c%252528v=vs.100%252529.aspx
> > > and
http://msdn.microsoft.com/en-us/library/xdkz3x12%252528v=vs.100%252529.aspx
> > > 
> > > Granted, I haven't tried this in a long time, but IIRC Python on Windows
has
> > > basic signal support.
> > 
> > It does
> > http://docs.python.org/library/subprocess.html#subprocess.Popen.send_signal
> (and
> > also the Python win32 API)
> > 
> > Dart just has to send the right information when closing the process, which
it
> > is currently not doing: http://code.google.com/p/dart/source/detail?r=8541
> 
> Makes me wonder what kind of spooky Win32 magic Python is doing to make that
> work :)
> 
> in particular the note about creationflags: CREATE_NEW_PROCESS_GROUP makes me
> wonder if they're using "CreateJobObject" or something similar. That's one of
> the ways to get a process group to terminate together.
> 
> Sigh, Windows :\

Yes, I wonder what they are doing too to make it look to Python programmers like
they can send signals on an OS that has no signals. :-)

They might be using something like process groups. However, it is non-trivial to
decide what to put in a process group unless you allow the programmer to do it
himself. Which processes started from Dart do you want to be in the same process
group? Probably none of them because you want to kill them individually and not
have the other's die at the same time.

Anyway, we should look into making 'kill' a more graceful shutdown on Windows.
There *should* be a way to do that. :-)