Change PlatformDecoderModule::Create*Decoder() to take an
mp4_demuxer::{Audio,Video}DecoderConfig parameter, instead of enumerating all
parameters. This means the platform decoders can have more data if need be,
like the AACAudioConfig.
Change MediaDataDecoder::Input() to take an nsAutoPtr<MP4Sample>&. The sample
will be deleted by the caller (MP4Reader) if Input() returns DECODE_STATUS_OK,
but if the MediaDataDecoder wants to assume responsibility of the
lifecycle of the sample (say to enqueue it), it can forget() on the
nsAutoPtr& passed in and assume responsibility.
Call PlatformDecoderModule::Create() on the decode thread. This is a step
towards making these classes decode-thread only.
Add PlatformDecoderModule::Init(), which caches the pref's we need, since
PlatformDecoderModule::Create() is no longer called on the main thread, we can
no longer access them in there.
Add Init() method to MediaDataDecoder interface. This is so that we can call
MediaDataDecoder::Shutdown() to unblock the initialization of a decoder, if
that init needs to block.
Pass LayersBackend type to WMFVideoDecoder, so it knows whether to init DXVA.
Adds comments to PlatformDecoderModule.h, and describes the call model
better. Renames CreateAudioDecoder and CreateVideoDecoder to
CreateAACDecoder and CreateH264Decoder. We may change them to accept a
codec parameter and have more general names in future when we need to
support more codecs here.
We should restrict the output of MP4Reader to H.264 and AAC so that we
don't end up invoking rogue platform decoders (if that's even possible).
This means that we won't play MP3 audio inside MP4, that's deliberate.
This ensures that if the MP4Reader fails to get a decoder owner, we
don't abort the media load, but instead we will assume that we don't
have an advanced layers backend, and so fallback to software decoding on
Windows.
