And set the output video context.

Versions older than 3.13 don't work well with clang to create win32+ucrt
targets.

Marvin Scholz authored 5 years ago and Thomas Guillem committed 5 years ago

macOS pthread implementation for pthread_cond_wait is buggy, causing
sometimes the mutex to not be locked when running thread cancellation
cleanup handlers.
This causes random failures of vlc_mutex_unlock, as the cancellation
cleanup handler does not hold a lock on the mutex in rare cases.

Signed-off-by: Thomas Guillem <thomas@gllm.fr>

Either it's initialized by the decoder device we require or we're not going to
use it at all.

The connection we'll use with this vdp/device will match the window configuration
as it comes from the decoder device which uses the same window configuration as
the window this display is tied to.

Thomas Guillem authored 5 years ago and Steve Lhomme committed 5 years ago

Signed-off-by: Steve Lhomme <robux4@ycbcr.xyz>

Thomas Guillem authored 5 years ago and Steve Lhomme committed 5 years ago

pictures, context and video context are now allocated from the mediacodec
module. This simplify a lot of things since MediaCodec is naturally a push API.

The mediacodec decoder module, via the decoder device will now decide of the vd
plugin used: android-display or gles2. It depends on how the LibVLC user setup
its surfaces or if the format need an OpenGL transformation (360 videos). The
decoder module need to know this informations before configuring MediaCodec in
order to choose the good Surface. The Surface will be either the SurfaceView
directly (AWindow_Video) or a SurfaceTexture that will be attached to a GL
texture (AWindow_SurfaceTexture) by the android gl converter.

Signed-off-by: Steve Lhomme <robux4@ycbcr.xyz>

Thomas Guillem authored 5 years ago and Steve Lhomme committed 5 years ago

Since p_sys will be referenced by the future video context and will be used to
do the final cleaning when the video context is not referenced anymore.

Signed-off-by: Steve Lhomme <robux4@ycbcr.xyz>

Thomas Guillem authored 5 years ago and Steve Lhomme committed 5 years ago

Signed-off-by: Steve Lhomme <robux4@ycbcr.xyz>

Thomas Guillem authored 5 years ago and Steve Lhomme committed 5 years ago

Signed-off-by: Steve Lhomme <robux4@ycbcr.xyz>

And set the output video context.

And set the output video context.

We can always get it from the video context which we now have on input of the filter.

Similar to how we use the intermediate output video format out of each chain.

Just like with the intermediat video format.

Similar to how we get the current output video format.

And the instance holder.

Rather than the legacy vlc_vaapi_FilterHoldInstance().

The VADisplay is stored in the decoder device opaque.

Rather than using vlc_vaapi_FilterHoldInstance().

The VADisplay is stored in the decoder device opaque.

Rather than a decoder device. In the end we need to store the video context,
not the decoder device in the pictures.