[MPlayer-dev-eng] Re: OpenDML Read/Write support

Ivan Kalvachev ivan at cacad.com
Mon Mar 22 15:01:13 CET 2004


Tobias Diedrich said:
> D Richard Felker III wrote:
>
>> > you'd have to look at the video dimensions at the filter chain's end
>> of
>> > course.  The current situation where it is specified in -lavcopts is
>> > not quite right IMHO.
>>
>> See -lavcopts autoaspect.
>
> Ok, the problem I see here is that d_width and d_height are only passed on
> as parameters in vf_next_config.
> I'd propose to add d_width and d_height fields to struct vf_instance_s
> and let vf_next_config set them.
> Then either the muxer can walk the filter chain and use the values from
> the last filter or we take the final values and store them in
> sh_video_t, so the muxer wouldn't have to know about video filters.
>
> Then add a big fat warning to ve_lavc if -lavcopts aspect is used
> instead of autoaspect.
>
> I'm going to grab something to eat now and write a patch after that.
>
Hey do you read your mailbox?

I have send an valid workarond.
Well. My solution is to add aspect in the muxer structures.
I have done this in the stream, as I sow that ve_lavc works\directly with
video stream muxer structure (mux_v).

Please take a look of my solution. I didn't say it is better,
but sh_video is used for input video.

And the video encoder should calculate the aspects, as it is last
in the chain. I also don't see problem the aviaspect option to
stay in case we want to change avi aspect with ovc/oac copy.

And we still don't know if WMP honor aspect info as we have never
produced stream with valid aspect info;)
Don't forget that different version of WMP may behave diffrently
(I think I sow an WMP9 to always resize fullscreen move to 4:3
breaking aspect)


Best Regards
   Ivan Kalvachev
  iive




More information about the MPlayer-dev-eng mailing list