Onerror return value handling in workers fixed as per HTML5 spec.

Onerror return value handling in workers fixed as per HTML5 spec.

This CL fixes the error propagation based on the return value
of the error event handler. As per the spec if the value returned is true,
the event should be cancelled, if false it should be propagated up till the Document.
This is based on the interpretation of the following spec sections:
https://html.spec.whatwg.org/multipage/webappapis.html#event-handlers
https://html.spec.whatwg.org/multipage/workers.html#runtime-script-errors-2

BUG=144593

[Mayur Kankanwadi, 2014-09-25 14:23:20]

mayurk.vk@samsung.com changed reviewers:
+ haraken@chromium.org, jochen@chromium.org, tkent@chromium.org

[Mayur Kankanwadi, 2014-09-25 14:23:21]

PTAL at the fix.
This CL currently breaks the following two layout tests:
fast/workers/worker-onerror-04.html
storage/indexeddb/transaction-complete-workers.html
I am working on to fix these tests, so that the modified behavior passes. 
Thanks.

[Mike West, 2014-09-26 11:31:15]

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

[Mike West, 2014-09-26 11:31:16]

What does Firefox do in these scenarios?

[Mayur Kankanwadi, 2014-09-26 11:38:50]

On 2014/09/26 11:31:16, Mike West wrote:
> What does Firefox do in these scenarios?

FF cancel's the error event on 'true' return value from the error handler, not
letting the error to be propagated further till the document.
The spec(https://html.spec.whatwg.org/multipage/webappapis.html#event-handlers)
has this to say as
--
Process return value as follows:
1) If the event type is mouseover
2) If the event type is error and E is an ErrorEvent object
    If return value is a Web IDL boolean true value, then cancel the event.
--
Hope, the interpretation is correct.

[Mike West, 2014-09-26 11:49:07]

On 2014/09/26 11:38:50, Mayur Kankanwadi wrote:
> On 2014/09/26 11:31:16, Mike West wrote:
> > What does Firefox do in these scenarios?
> 
> FF cancel's the error event on 'true' return value from the error handler, not
> letting the error to be propagated further till the document.
> The
spec(https://html.spec.whatwg.org/multipage/webappapis.html#event-handlers)
> has this to say as
> --
> Process return value as follows:
> 1) If the event type is mouseover
> 2) If the event type is error and E is an ErrorEvent object
>     If return value is a Web IDL boolean true value, then cancel the event.
> --
> Hope, the interpretation is correct.

So this patch brings us into line with Firefox's implementation, correct? If so,
please add a test to verify the behavior.

[Mayur Kankanwadi, 2014-09-26 11:50:24]

On 2014/09/26 11:49:07, Mike West wrote:
> On 2014/09/26 11:38:50, Mayur Kankanwadi wrote:
> > On 2014/09/26 11:31:16, Mike West wrote:
> > > What does Firefox do in these scenarios?
> > 
> > FF cancel's the error event on 'true' return value from the error handler,
not
> > letting the error to be propagated further till the document.
> > The
> spec(https://html.spec.whatwg.org/multipage/webappapis.html#event-handlers)
> > has this to say as
> > --
> > Process return value as follows:
> > 1) If the event type is mouseover
> > 2) If the event type is error and E is an ErrorEvent object
> >     If return value is a Web IDL boolean true value, then cancel the event.
> > --
> > Hope, the interpretation is correct.
> 
> So this patch brings us into line with Firefox's implementation, correct? If
so,
> please add a test to verify the behavior.

Yes, I will add the extra layout test for this scenario. I am also checking two
existing 
layout test failures due to this patch.
Thanks.

[jochen (gone - plz use gerrit), 2014-10-07 14:32:55]

any updates on this? The CL looks good per se, but needs tests

[Mayur Kankanwadi, 2014-10-07 14:55:52]

On 2014/10/07 14:32:55, jochen wrote:
> any updates on this? The CL looks good per se, but needs tests

Yes, I am working on it. 
By tomorrow you should be able to see some updates on it.

[Mayur Kankanwadi, 2014-10-08 15:17:02]

On 2014/10/07 14:55:52, Mayur Kankanwadi wrote:
> On 2014/10/07 14:32:55, jochen wrote:
> > any updates on this? The CL looks good per se, but needs tests
> 
> Yes, I am working on it. 
> By tomorrow you should be able to see some updates on it.

Hi jochen,
PTAL.
Added new layout tests for the change.
I am not so sure of the changes done in
/storage/indexeddb/resources/transaction-complete-workers.js.
It will fix the test case, but am not sure if it still serves the right cause.
Thanks.

[jochen (gone - plz use gerrit), 2014-10-09 11:27:29]

https://codereview.chromium.org/607513002/diff/40001/LayoutTests/fast/workers...
File LayoutTests/fast/workers/worker-onerror-04-expected.txt (left):

https://codereview.chromium.org/607513002/diff/40001/LayoutTests/fast/workers...
LayoutTests/fast/workers/worker-onerror-04-expected.txt:1: CONSOLE ERROR: line
7: Uncaught ReferenceError: foo is not defined
why is this error gone? the onerror handler in the test returns false, shouldn't
that not cancel the event?

[jochen (gone - plz use gerrit), 2014-11-12 15:05:02]

any updates on this?

[Mayur Kankanwadi, 2014-11-13 14:34:17]

On 2014/11/12 15:05:02, jochen (slow) wrote:
> any updates on this?

Hi jochen,
Sorry for the delay in updates.
I am currently stuck with some other matters.
Hope to resolve them soon.
So please bear with me.
Best regards,
Mayur Kankanwadi.