RE: navigationStart

Navigation Timing was the first consumer of this new framework...so it could have issues like you are bringing up.  There has been other issues that this suite has uncovered, so it's possible we haven't flushed them all out. 

Back to Tony's suggestion:
1. Verify whether we have a conformance test that covers this. I think we do not. Perhaps we could write one by inserting a large fixed delay in the beforeunload and unload handlers, navigating to a new page, and ensuring that the unload delay was counted but not the beforeunload.

I think this is a valid test to add.  We currently have a test for the unload event (http://w3c-test.org/webperf/tests/approved/navigation-timing/html5/test_timing_attributes_order.html), but not one for the beforeunload and definitely not one that does calculations of these times outside of just their timeline order and non-zero value.

I will take the action item to add this test (Action-75 has been created for this).  As an aside, outside of onbeforeunload being cancellable, why would we not want to capture the time spent here?

Thanks,
Karen

-----Original Message-----
From: Tony Gentilcore [mailto:tonyg@google.com] 
Sent: Wednesday, November 16, 2011 3:34 AM
To: Boris Zbarsky
Cc: olli@pettay.fi; Ricardo Oliveira; public-web-perf@w3.org
Subject: Re: navigationStart

On Wed, Nov 16, 2011 at 11:22 AM, Boris Zbarsky <bzbarsky@mit.edu> wrote:
> On 11/17/11 12:12 AM, Tony Gentilcore wrote:
>>
>> TBH, I don't see any tests that run in an<object>.
>
> http://w3c-test.org/framework/test/nav-timing-default/single/test_uniq
> ue_performance_objects/format/html/
> is very definitely running in an <object>.  The whole test is loaded 
> in an <object>.
>
> Same thing for
> http://w3c-test.org/framework/test/nav-timing-default/single/test_navi
> gate_within_document/format/html/
>
> Both are claimed to fail in Firefox; both pass completely if loaded 
> directly as 
> http://w3c-test.org/webperf/tests/approved/navigation-timing/html/test
> _unique_performance_objects.html
> and
> http://w3c-test.org/webperf/tests/approved/navigation-timing/html/test
> _navigate_within_document.html
> respectively.

Your are right. That use of <object> is bizarre. I didn't notice it because I was running the tests directly by clicking through http://w3c-test.org/webperf/tests/ instead of through that harness.

>
>> Nor does document.getElementsByTagName('object') turn anything up on 
>> the tests I sampled.
>
> The tests themselves don't have <object> in them.  The harness does, 
> as far as I can tell.
>
>> The sources along with revision history are here:
>>
>> https://dvcs.w3.org/hg/webperf/file/dbf0cc61824a/tests/approved/navig
>> ation-timing/html
>
> That doesn't include the sources for the test harness.

The test framework repos are here:
https://dvcs.w3.org/hg/testframework/
https://dvcs.w3.org/hg/testframework-core/

But I believe the easiest thing to do at this point is to file a bug at:
http://www.w3.org/Bugs/Public/enter_bug.cgi?product=Testing&component=Test%20Framework&priority=P3

Received on Wednesday, 16 November 2011 19:22:52 UTC