Re: HME WG meeting minutes, Tue Aug 30

I'm sharing updated Chrome JSONs with @plehegar shortly. I'll continue
analysis on this tomorrow.

Matt

On Thu, Sep 8, 2016 at 5:46 PM, Matt Wolenetz <wolenetz@google.com> wrote:

> I landed a fix for #3662 (appendbuffer-quota-exceeded).
> I'll produce new Chrome JSON now (with and without isobmff-support, on
> tip-of-tree local build). These results won't be "filtered" vs
> at-risk/removed features and won't include the interfaces.html sync to
> current editor's draft that Jerry's working on.
>
> On Thu, Sep 8, 2016 at 4:47 PM, Matt Wolenetz <wolenetz@google.com> wrote:
>
>> That makes sense, Jerry - Chrome and FF were complying and firing load
>> event earlier; the fix is to test when the tests are ready. If the lack of
>> load event firing in Edge causes failures in the MSE IDL test, that would
>> need to be fixed in Edge probably.
>>
>> Also, syncing interfaces.html to editor's draft of IDL makes sense.
>> Please prepare a PR for that.
>>
>> On Thu, Sep 8, 2016 at 3:40 PM, Jerry Smith (WPT) <jdsmith@microsoft.com>
>> wrote:
>>
>>> I don’t see any issue between the IDL test change and Edge.  It tells
>>> the harness that the tests aren’t over until an explicit done is executed.
>>> That prevents a quick load event from confusing the test.  Apparently only
>>> Canary was setting the delay-the-load-event-flag to false on attachment,
>>> and that was causing an earlier load that tripped the failure condition.
>>> Edge doesn’t support this flag currently and so on this test at least
>>> should be okay.  Correct?
>>>
>>>
>>>
>>> Is there any reason we don’t update the interface.html to reflect the
>>> current editors draft?  That would eliminate the need for filtering on the
>>> IDL test at least.  I’m willing to bug that change and create a PR.
>>>
>>>
>>>
>>> Jerry
>>>
>>>
>>>
>>> *From:* Matt Wolenetz [mailto:wolenetz@google.com]
>>> *Sent:* Thursday, September 8, 2016 11:05 AM
>>> *To:* Philippe Le Hégaret <plh@w3.org>
>>> *Cc:* Jean-Yves Avenard <jyavenard@mozilla.com>; Paul Cotton <
>>> Paul.Cotton@microsoft.com>; public-html-media@w3.org
>>> *Subject:* Re: HME WG meeting minutes, Tue Aug 30
>>>
>>>
>>>
>>> Also, the IDL failures have been fixed by @tidoust (who made an
>>> excellent find). The fix may expose issues in implementations which do not
>>> yet stop delaying the load event immediately on attachment (Edge?).
>>>
>>>
>>>
>>> I'm working on addressing https://github.com/w3c/web-pla
>>> tform-tests/issues/3662#issuecomment-245683923 (which, IIUC, FF is
>>> failing or timing out against) as well as updating Chrome's results with a
>>> tip-of-tree build for each of with-isobmff and without-isobmff support
>>> today.
>>>
>>>
>>>
>>> Matt
>>>
>>>
>>>
>>> On Thu, Sep 8, 2016 at 11:03 AM, Philippe Le Hégaret <plh@w3.org> wrote:
>>>
>>>
>>>
>>> On 9/7/2016 11:27 PM, Jean-Yves Avenard wrote:
>>>
>>> Hi
>>>
>>> On 8 Sep 2016, at 5:03 AM, Philippe Le Hégaret <plh@w3.org> wrote:
>>>
>>> I regenerated test results and added a filter to eliminate
>>> VideoPlaybackQuality and TrackDefault related:
>>>    http://w3c.github.io/test-results/media-source/less-than-2.html
>>>
>>> We get a lot of errors on the idlharness tests which I'll look at as
>>> well.
>>>
>>> Philippe
>>>
>>> I’d love to know how you generate those reports, in particular the
>>> Firefox results one.
>>>
>>> I used the online version of WPT
>>> http://w3c-test.org/tools/runner/index.html
>>>
>>> with /media-source/
>>>
>>> using Firefox 51.0a1 (2016-09-08) (64-bit).
>>>
>>> I obtained http://w3c.github.io/test-results/media-source/FF51.json
>>>
>>> Because with the exception of any tracks related methods and the
>>> VideoQuality::totalFrameDelay (which we will not add as we believe it
>>> serves no purpose); all tests are PASS here using firefox 51.
>>>
>>> If you have the JSON file, I'm more than happy to replace mine with
>>> yours. It's my preference in fact and I didn't think of asking you, sorry.
>>>
>>> It's highly possible I didn't set the proper flags when I ran the tests.
>>> Just make sure you get the latest commits from WPT (or use the online
>>> version).
>>>
>>> Philippe
>>>
>>>
>>>
>>
>>
>