[MPlayer-dev-eng] fixed resolution/window size on streams

[Soeren Sonnenburg]

Dear list,

when playing the demoscene realtime stream
(via mplayer http://91.121.4.152:8000 ) I noticed that the resolution is
frequently on-the-fly switched, which ruins any viewing experience (as
mplayer drops out of full screen mode and in window mode the window
jumps around and is rescaled to different sizes).

As non of the options -fixed-vo -nokeepaspect -monitoraspect 4:3 etc
seem to help I am wondering what one could do about this (although I am
using -vo sdl, it also does not work with -vo x11 where it even causes
crashes and on -vo mpegpes mplayer dies with cannot find codec when that
happens).

Looking at the output I see things like:

VDec: vo config request - 512 x 480 (preferred colorspace: Planar YV12)% 
VDec: using Planar YV12 as output csp (no 0)
Movie-Aspect is undefined - no prescaling applied.
VO: [sdl] 512x480 => 512x480 Planar YV12 
VDec: vo config request - 512 x 384 (preferred colorspace: Planar YV12)% 
VDec: using Planar YV12 as output csp (no 0)
Movie-Aspect is undefined - no prescaling applied.
VO: [sdl] 512x384 => 512x384 Planar YV12 
VDec: vo config request - 384 x 384 (preferred colorspace: Planar YV12)% 
VDec: using Planar YV12 as output csp (no 0)
Movie-Aspect is undefined - no prescaling applied.
VO: [sdl] 384x384 => 384x384 Planar YV12 
VDec: vo config request - 384 x 288 (preferred colorspace: Planar YV12)% 
VDec: using Planar YV12 as output csp (no 0)
Movie-Aspect is undefined - no prescaling applied.
VO: [sdl] 384x288 => 384x288 Planar YV12 
VDec: vo config request - 320 x 240 (preferred colorspace: Planar YV12)% 
VDec: using Planar YV12 as output csp (no 0)
Movie-Aspect is undefined - no prescaling applied.
VO: [sdl] 320x240 => 320x240 Planar YV12 
VDec: vo config request - 384 x 288 (preferred colorspace: Planar YV12)% 
VDec: using Planar YV12 as output csp (no 0)
Movie-Aspect is undefined - no prescaling applied.

So the *input resolution* changes all the time.

Any comments and ideas how this should be fixed? Should fixed-vo always
append the scale filter, as this is not enough (I tested), should the
input be normalized to the initial resolution?

Soeren