[Webkit-unassigned] [Bug 178076] [MSE][GStreamer] Insert parser elements in AppendPipeline when demuxing opus or Vorbis

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=178076

--- Comment #4 from Alicia Boya García <aboya at igalia.com> ---
Comment on attachment 323400
  --> https://bugs.webkit.org/attachment.cgi?id=323400
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=323400&action=review

>> Source/WebCore/platform/graphics/gstreamer/mse/AppendPipeline.cpp:955
>> +    GRefPtr<GstCaps> padCaps = adoptGRef(gst_pad_get_current_caps(GST_PAD(demuxerSrcPad)));
> 
> Is this GST_PAD() macro actually needed?

Not really, I should remove it.

>> Source/WebCore/platform/graphics/gstreamer/mse/AppendPipeline.cpp:965
>> +        return opusparse;
> 
> Just a comment for future reference without any action required on your side: opusparse is a floating reference. The return type is GRefPtr. This means that there's going to be a GRefPtr(opusparse) call under the hood which will call g_object_ref_sink() and convert the floating ref into a regular ref owned by the GRefPtr.

That is intended in this case, though I could be more explicit, right.

>> Source/WebCore/platform/graphics/gstreamer/mse/AppendPipeline.cpp:1034
>> +        // This is known to be an issue with YouTube webm files containing Opus audio as of YT2018.
> 
> YouTube/tv (isn't the same as YouTube)

Regular YouTube also uses Opus, I would be surprised if the files were different in regard to durations in the container.

>> Source/WebCore/platform/graphics/gstreamer/mse/AppendPipeline.cpp:-1094
>> -    // Must be done in the thread we were called from (usually streaming thread).
> 
> What happened to the black hole probe?

The condition in the `if` block was always false because by the time `pad-removed` is emitted the pad is already unlinked.
I'm not completely sure about probes lifecycle but I think they are removed automatically with the pad when the pad refcount hits zero (which should be soon after the `pad-removed` handler finishes unless it's re-linked there).

>> Source/WebCore/platform/graphics/gstreamer/mse/AppendPipeline.cpp:-1111
>> -        gst_element_unlink(m_demux.get(), m_decryptor.get());
> 
> No unlinking now? Why?

Removing the elements from the bin automatically unlinks them. It's easier to remove a series of optional elements than to remove links between a chain of optional elements then also removing the elements.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20171013/ecae5ed1/attachment.html>