RE: MSE IDL tests

> PFA updated Chrome JSONs. I'll continue analysis on them tomorrow.

Can you confirm that there no work left to do on the MSE WebIDL tests?

Nearly half of the sub-test failures in the current MSE test suite are for tests on media-source/interfaces.html:

What is our explanation for these failures?


From: Matt Wolenetz []
Sent: Thursday, September 8, 2016 10:16 PM
To: Philippe Le Hégaret <>
Cc: Paul Cotton <>
Subject: Re: MSE IDL tests

Note, those JSONs are produced with the "--enable-experimental-web-platform-features" cmdline flag to Chrome.

On Thu, Sep 8, 2016 at 7:14 PM, Matt Wolenetz <<>> wrote:
#3662 is fixed (though FF and others may fail, shouldn't block PR though).

PFA updated Chrome JSONs. I'll continue analysis on them tomorrow.
These are from tip-of-tree Chrome local builds against<>'s runner. "no-isobmff" is a version for which I've forced my local build of Chrome to say "false" to any audio/mp4 or video/mp4 isTypeSupported() queries, so that the webm alternative (if any) is tested instead. Any tests that fail for mp4 but pass without mp4 support are likely due to mp4 PTS/DTS conflation in Chrome. IIUC, Edge also has such an issue, but FF has supposedly done this correctly.

On Thu, Sep 8, 2016 at 10:59 AM, Matt Wolenetz <<>> wrote:
Thank you, Philippe. landed last night, but is not yet in Canary. If I get some time today, I'll generate local Chrome with and without isobmff JSONs with builds including that fix, and I'll include links to the compliance (or test suite) bugs discovered as a result.

First though, I'll try to alleviate the problems FF is having with the current appendbuffer-quotaexceeded test:

On Thu, Sep 8, 2016 at 10:41 AM, Philippe Le Hégaret <<>> wrote:

I regenerated the test results.

I think we're good to go on that front.


On 9/8/2016 10:09 AM, Paul Cotton wrote:
Can you merge this PR today so we can refresh the MSE results?


From: Matt Wolenetz []
Sent: Thursday, September 8, 2016 9:48 AM
To: Francois Daoust <><>
Cc: Philippe Le Hegaret (<>) <><>; <><> <><>; Paul Cotton <><>
Subject: Re: MSE IDL tests (was: Re: HME WG meeting minutes, Tue Aug 30)

That makes sense. Thank you for finding and fixing the test problem.

On Sep 8, 2016 6:09 AM, "Francois Daoust" <<>> wrote:

In essence, I think that the test harness, called by the IDL harness, expects to get the list of tests before the "load" event fires on window. In other words, it expects all calls to "test" and "async_test" to have been made by then.

In the MSE case, we need to wait for the MediaSource object to be ready before we may call the IDL harness, and the "load" event will have already fired at that point. The IDL harness tells the test harness to run the first test, and since the "load" event already fired, the test harness believes it is the only test to run, and switches to an "I'm done!" mode, getting ready to report the results. Problem is this is when it receives a number of additional tests to perform. This makes the test harness go berserk.

The solution is relatively simple: we just need to be explicit as to when the list of tests is complete, which can be done with an initial call to "setup({ explicit_done: true });" followed by a call to "done()" after having run the IDL harness.

Done in the following PR:


Le 07/09/2016 à 23:22, Matt Wolenetz a écrit :
Thanks Philippe. Help with the IDL failures would be appreciated -- it
looks like it's not specific to just Chrome. As mentioned in narrower
thread, I've updated and landed #3655, and will rerun Chrome's JSON (for
builds with, and without, ISO-BMFF support claimed by isTypeSupported())
once I land (which should
fix a bunch of the TypeError-related Chrome failures). Landing that
Chrome patch may not come before EOD today, though.

On Wed, Sep 7, 2016 at 12:03 PM, Philippe Le Hégaret <<>
<<>>> wrote:

    On 9/2/2016 7:24 PM, Matt Wolenetz wrote:
    I've made some progress, not as much as I had hoped. 2 of my PRs
    are currently pending co-editor review. 5 pending PRs are still
    awaiting my full review.

    Thank you Matt for making progress!

    I merged one of your PRs, still looking at #3655.

    I regenerated test results and added a filter to eliminate
    VideoPlaybackQuality and TrackDefault related:


    We get a lot of errors on the idlharness tests which I'll look at as


      * Prepped PR to cover 4 gaps from #823

          o *Awaiting @jdsmith3000's review to land it*
          o crbug 623781 <http://crbug/623781> tracks fixing the
            Chrome compliance issue behind at least the failure of 2
            subtests (TypeError) in<>
          o Filed crbug 643788 <http://crbug/643788> to fix debug
            build assertion in w-p-t
            mediasource-removesourcebuffers.html and landed my fix for
            it <>

      * Reviewed and closed

          o Found only 1 trivial test coverage gap, prepped and merged
            trivial PR to close that

      * Deferred to VNext ancient appendStream coverage test

      * Deferred to VNext more strict isTypeSupported() coverage test


        Commented on mechanical

        <> (doesn't
        block MSE v1 PR, it's a cross-cutting infra PR)




          o Filed, and in-progress fixing (short-term Chrome MSE piece
            of) crbug 643846 <http://crbug/643846>

        TODO 5 more pending PRs...

    On Tue, Aug 30, 2016 at 6:42 PM, Matt Wolenetz
    <<> <<>>> wrote:

        On Tue, Aug 30, 2016 at 11:20 AM, Paul Cotton
        <<> <<>>>

            *MSE test suite results*

            <*/plh/*> Paul: still some outstanding pull requests


            is Matt's most recent update

            <*/plh/*> Matt: going through them...

            Outstanding PRs:


            <*/plh/*> ... resolving the merge conflict

            <*/plh/*> ... I'll send out an update later today

        I've worked through about 5 of the 15 pending PRs today, and
        will continue tomorrow. Current status:


          * Merged

          * Prepped and landed

          * Needed Chromium+W3C license stamp, so this

          * Updated and landed
          * Landed

          * Prepped PR to fix the test introduced in #3233 to comply

              o *Awaiting @jdsmith3000's review to land it*
              o crbug 639144 <http://crbug/639144> tracks fixing the
                Chrome compliance issue behind failure of subtest 1/3
                in this PR
              o crbug 373039 <http://crbug/373039> is one of a group
                of bugs tracking fixing the Chrome compliance issue
                behind failure of subtest 3/3 in this PR
          * Reviewed

              o Found 4 gaps, but otherwise closed this PR since most
                of the coverage it includes is already covered
                elsewhere in our current suite

            In-progress: put together PR to cover 4 gaps from #823

              o crbug 623781 <http://crbug/623781> tracks fixing the
                Chrome compliance issue behind at least the failure of
                2 subtests (TypeError) in
              o *In-progress: Investigating Chromium issue with PR I'm
                building to cover the 4 test gaps.*
          * TODO 10 more pending PRs, followed by call to regenerate
            reports and lists of known user agent bugs tracking test

Received on Monday, 19 September 2016 13:11:16 UTC