[MPlayer-dev-eng] [PATCH] Implement stride support in ve_raw and vf_vo wrapper

[John Koleszar]

Ivan Kalvachev wrote:

>>From the patch it looks like you get everything right.
>  
>
After more testing, it looks like I didn't. :-(

>I had some doubts about using the video system to allocate the
>temporal buffer, but on second thought I guess it is OK.  
>
This doesn't actually work in all cases. The model for calling 
vf_get_image is that you get get destination images by using vf->next, 
which I gather is how DR works. Since the ve's and vo's are at the end 
of the chain, there is no next filter. Since the decoder (or other 
filters) render into the ve's context, the ve can't allocate an image in 
it's own context.

This problem manifests itself when a filter or decoder requests a temp 
image either via MP_IMGTYPE_TEMP or MP_IMGTYPE_IPB. In this case the mpi 
passed to the ve's put_image is the same instance as the temp mpi that 
was allocated at init time, which obviously doesn't work. One decoder 
that does this is SVQ3 as seen in the  clip below:

ftp://opensource:opensource@ftp.on2.com/Narnia_LWR_Trailer1_87d2_3000.mov
Sorry for the large size, I tried to mux a clip of it into an avi to 
make it a little smaller, but mplayer segv'd when trying to play it back.

So the question becomes how to best allocate the temp buffer. I could 
just memalign it and calculate the dimensions and strides, but it seems 
like that duplicates an awful lot of the mpi interface. Based on the 
number of formats this is supposed to support, I think it makes sense to 
stick with the mpi. My current workaround is a hack that copies about 35 
lines from vf_get_image to set up the mpi, but I don't think this is a 
good general solution, since that's a big chunk of code to duplicate in 
three places (vf_get_image, ve_raw, vf_vo). Would people here be 
amenable to moving the common code to a  new function for this and using 
it in those three places? Or at least the ve and vo? Just need a little 
guidance on how to proceed.

To see the issue with the ve:
$ mencoder -nosound -ovc raw -frames 240  -o  test.avi  
Narnia_LWR_Trailer1_87d2_3000.mov

To see the issue with the vo:
$ mplayer -vf format=bgra -vo sdl Narnia_LWR_Trailer1_87d2_3000.mov
(need to use a rgb format to get into the draw_frame code path. This one 
isn't as big a deal as the one above, since it's only one plane, but you 
can see that the dimensions are wrong)

Cheers,

John