bugs

Motion JPEG编解码器源代码

1.	MPEG2 support for progressive sequences frame (-F 0) and  
	field-interlaced (-F 3) is now well tested.  Field
	sequences (-F 1 and -F 2) appear to be working but aren't yet thoroughly
	tested and lack some recent optimisations available for
	progressive sequences.  

2. 	Currently the VCD HR Stills muxing stuff assumes *all* HR stills
  	are the same size which is given in the initial vbv_buffer_size...
	This will work with mpeg2enc (which does this) but will fail fail fail
	with other streams.

3.	The VBVbuffer delay code in the MPEG encoder is hopeless broken.  
	It is currently turned off
	and needs re-implementing from scratch.  Not urgent though since
	no decoder ever seems to take any notice of it.  Certainly I never
	figured out the the damn thing would be useful for once your
	mux-er has	checked for /ensured video buffer size bounding.


4.	mpeg2enc is currently broken for 422 and 444 video.   The problem is that
	the calculations for checking chrominance in the the motion compensation
	search assume 2-pel sub-sampled U and V in X and Y.