Add new layout tests for the Web Notification API.

Add new layout tests for the Web Notification API.

This patch adds tests for the following scenarios:
  * Notification.permission should default to "default".
  * Notification.onshow() and Notification.onclose() should be called.
  * Notification.onclick() should be called when it gets clicked on.
  * Notification.tag should allow an existing notification to be replaced.
  * Sandboxed iframes should influence notification inheritance behavior.
  * Notifications can focus their opening window when it's in the background.

The expected output of "notifications-constructor-with-permission.html"
has been amended to reflect the updated output format.

Two tests have been updated to include the second argument to the
testRunner.grantWebNotificationPermission() call (whether to grant it).

BUG=349015

[Peter Beverloo, 2014-03-04 18:36:49]

Shiny new and updated tests :-). These depend on a few cleanups in the test
runner, which lives in Chromium:
  https://codereview.chromium.org/183663021/
  https://codereview.chromium.org/183853026/
  https://codereview.chromium.org/184313005/

Simple four-step plan for landing this CL is described here:
  http://crbug.com/349015

[abarth-chromium, 2014-03-04 19:07:13]

lgtm

[abarth-chromium, 2014-03-04 19:07:35]

https://codereview.chromium.org/186713002/diff/70001/LayoutTests/fast/notific...
File
LayoutTests/fast/notifications/notifications-constructor-with-permission.html
(right):

https://codereview.chromium.org/186713002/diff/70001/LayoutTests/fast/notific...
LayoutTests/fast/notifications/notifications-constructor-with-permission.html:13:
testRunner.grantWebNotificationPermission("file://", true);
What does |true| mean here?  Can we use a string or something more descriptive?

[Peter Beverloo, 2014-03-04 19:49:51]

https://codereview.chromium.org/186713002/diff/70001/LayoutTests/fast/notific...
File
LayoutTests/fast/notifications/notifications-constructor-with-permission.html
(right):

https://codereview.chromium.org/186713002/diff/70001/LayoutTests/fast/notific...
LayoutTests/fast/notifications/notifications-constructor-with-permission.html:13:
testRunner.grantWebNotificationPermission("file://", true);
On 2014/03/04 19:07:35, abarth wrote:
> What does |true| mean here?  Can we use a string or something more
descriptive?

Do you think grantWebNotificationPermission(boolean grantIt, string origin)
would be better?  I don't think "grantPermission(true)" is unclear like boolean
arguments usually are.

Alternatively I'd be happy to change the implementation to accept ["default",
"granted", "denied"] as well.

[abarth-chromium, 2014-03-04 21:04:39]

On 2014/03/04 19:49:51, Peter Beverloo wrote:
>
https://codereview.chromium.org/186713002/diff/70001/LayoutTests/fast/notific...
> File
> LayoutTests/fast/notifications/notifications-constructor-with-permission.html
> (right):
> 
>
https://codereview.chromium.org/186713002/diff/70001/LayoutTests/fast/notific...
>
LayoutTests/fast/notifications/notifications-constructor-with-permission.html:13:
> testRunner.grantWebNotificationPermission("file://", true);
> On 2014/03/04 19:07:35, abarth wrote:
> > What does |true| mean here?  Can we use a string or something more
> descriptive?
> 
> Do you think grantWebNotificationPermission(boolean grantIt, string origin)
> would be better?  I don't think "grantPermission(true)" is unclear like
boolean
> arguments usually are.
> 
> Alternatively I'd be happy to change the implementation to accept ["default",
> "granted", "denied"] as well.

Oh, I see.  I think this is fine.  If there were non-testing code, we'd probably
pick something slightly more readable, but for test code this is fine.

[Andrew T Wilson (Slow), 2014-03-04 21:26:37]

LGTM.

BTW, one other test that might be interesting is making sure that
requestPermission() can only be called in the context of a user gesture. Not
sure if testrunner has the facilities to let us test this case though.

https://codereview.chromium.org/186713002/diff/70001/LayoutTests/fast/notific...
File LayoutTests/fast/notifications/notification-close-expected.txt (right):

https://codereview.chromium.org/186713002/diff/70001/LayoutTests/fast/notific...
LayoutTests/fast/notifications/notification-close-expected.txt:8: PASS
notification.onshow() has been called.
BTW, weird to see the PASS lines show up in the output *after* "TEST COMPLETE"

https://codereview.chromium.org/186713002/diff/70001/LayoutTests/fast/notific...
File LayoutTests/fast/notifications/notification-replace.html (right):

https://codereview.chromium.org/186713002/diff/70001/LayoutTests/fast/notific...
LayoutTests/fast/notifications/notification-replace.html:31:
testPassed("updatedNotification.onshow() has been called.");
Does this test actually ensure that the first notification is closed/replaced?
Maybe we should register an onclose handler on the first one to ensure this? Or
is this just inherent in the test runner because it only shows one notification
at once?