Avoid hanging the UI while rendering waveform about media-toc HOT 3 CLOSED

Getting position from the pipeline's clock is up to 4x faster than calling get_position.
There are still noticeable hangs in the waveform rendering, so any field borrowing or mutex in the listener & tracker should be avoided when possible.

from media-toc.

It turns out that cairo stroke is the time consuming part.
Stroking each segment individually is faster than stroking the whole waveform.
Next step: stroke as part of the double buffering mechanism... no surprise!
See: https://www.cairographics.org/threaded_animation_with_cairo/

Note: it would be great if we could draw incrementally in the same way as with current implementation in WaveformBuffer.update_samples. This would require being able to drop the beginning of the graphic and add a new part.

from media-toc.

Not totally satisfactory yet, but at least current state is such that the code in media-toc tries hard to minimize the time spent locking the UI thread and reduces the offset between the actual audio playback position and the waveform rendering.

from media-toc.