The ffmpeg decoders require width/height/framerate on the sink template caps but really shouldn't for most cases. The container format/parser might not know these values but ffmpeg is still able to decode the stream.


In 0.10 this worked because there was only an intersection check between the caps but in 0.11 there is a subset check. In 0.10 it could've happened that caps without width/height/framerate were accepted although the template caps required these fields.


NOTE: The same might be true for the audio decoders and possibly and possibly in the encoders too