[content shell] add support for invoking testRunner.notifyDone in a different process

[content shell] add support for invoking testRunner.notifyDone in a different process

Some layout tests, e.g.
http/tests/navigation/no-referrer-target-blank.html create a new
process, and notifyDone() will be invoked in that new process. Add
support for forwarding that signal to the main test runner.

BUG=111316
R=mkwst@chromium.org
TEST=http/tests/navigation/no-referrer-target-blank.html does not time out

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=192819

[Mike West, 2013-04-08 13:50:50]

Thanks for this! One question, two nits:

https://codereview.chromium.org/13795002/diff/1/content/shell/shell.cc
File content/shell/shell.cc (right):

https://codereview.chromium.org/13795002/diff/1/content/shell/shell.cc#newcod...
content/shell/shell.cc:267: if
(CommandLine::ForCurrentProcess()->HasSwitch(switches::kDumpRenderTree))
Nit: I assume it's intentional that we're using kDumpRenderTree throughout this
file? Might be worth renaming the switch at some point once you've killed DRT.

https://codereview.chromium.org/13795002/diff/1/content/shell/webkit_test_run...
File content/shell/webkit_test_runner.cc (right):

https://codereview.chromium.org/13795002/diff/1/content/shell/webkit_test_run...
content/shell/webkit_test_runner.cc:532: 
Nit: Extra newline.

https://codereview.chromium.org/13795002/diff/1/content/shell/webkit_test_run...
content/shell/webkit_test_runner.cc:539:
interfaces->configureForTestWithURL(GURL(), false);
When does this kick in? I don't understand when the main test runner wouldn't be
in the main window.

[jochen (gone - plz use gerrit), 2013-04-08 13:53:39]

https://codereview.chromium.org/13795002/diff/1/content/shell/webkit_test_run...
File content/shell/webkit_test_runner.cc (right):

https://codereview.chromium.org/13795002/diff/1/content/shell/webkit_test_run...
content/shell/webkit_test_runner.cc:539:
interfaces->configureForTestWithURL(GURL(), false);
On 2013/04/08 13:50:50, Mike West (chromium) wrote:
> When does this kick in? I don't understand when the main test runner wouldn't
be
> in the main window.

The main test runner is per process, while the main window is globally unique,
i.e. this is true for the first RV in all but the first process.

[Mike West, 2013-04-08 13:58:10]

On 2013/04/08 13:53:39, jochen wrote:
>
https://codereview.chromium.org/13795002/diff/1/content/shell/webkit_test_run...
> File content/shell/webkit_test_runner.cc (right):
> 
>
https://codereview.chromium.org/13795002/diff/1/content/shell/webkit_test_run...
> content/shell/webkit_test_runner.cc:539:
> interfaces->configureForTestWithURL(GURL(), false);
> On 2013/04/08 13:50:50, Mike West (chromium) wrote:
> > When does this kick in? I don't understand when the main test runner
wouldn't
> be
> > in the main window.
> 
> The main test runner is per process, while the main window is globally unique,
> i.e. this is true for the first RV in all but the first process.

Ah, I see. That makes sense, thanks.

LGTM, if you fix the whitespace nit. :)

[jochen (gone - plz use gerrit), 2013-04-08 14:15:36]

Committed patchset #2 manually as r192819 (presubmit successful).