[Gamester17]

I am still having a slight rendering issue in video-playback in the latest XBMC for Windows (SDL/OpenGL Win32 version of XBMC) with my (HP Compaq 2510p) laptop that has Intel x3100 graphics (GMA965).

The problem is a 1-pixel thick horizontal line just at the lower edge of the cropping that contstantly keeps flashing in a greenish/pinkish color. Here is a debug log and some screenshots if you guys get the opertunity to look at it:
http://pastebin.com/m3389d196
http://s3.supload.com/free/screenshot000.JPG/view/
http://s3.supload.com/free/screenshot001....jpg/view/
http://s3.supload.com/free/screenshot002....jpg/view/
http://s3.supload.com/free/screenshot003....jpg/view/
http://s3.supload.com/free/screenshot004.JPG/view/

Note! Before the rendering fix for Intel GMA950 was put into XBMC SVN I only got a black screen from video playback under XBMC for Windows.

PS! This could possibly be a graphic drivers issue as it is an HP laptop and it will only let me install the latest HP approved video graphic drivers (and not the very latest drivers direcly downloaded from intel.com, and the HP drivers are about 6-months old, ...when I try to to install the latest drivers from intel.com it sais that those do not support my hardware).

[WiSo]

Almost the same on Mac mini GM945. I have the line at the bottom and a line right from the video. I guess it's one pixel thick.
d4rk, elupus any ideas?

[elupus]

Likely the same issue I just fixed for skin textures. By subtracting 0.5 from the texture cordinates width, it aught to go away.

[WiSo]

Unfortunately recent svn doesn't fix the problem on the mac mini:
http://s3.supload.com/free/screenshot000.png/view/
http://s3.supload.com/free/screenshot003.png/view/

[jmarshall]

Recent SVN doesn't contain the fix elupus alludes to. The fix he did is specifically for skin textures. The same fix needs to be applied to the video textures.

Essentially: Either alter the texture coords a bit (might possibly be troublesome for scaling - not sure) or clear out the textures to transparent on creation.

[WiSo]

Ok, just saw that d4rk also changes something:

Code:

fixed: projection matrix was off by 1 pixel

But I'm really no expert to see if it's for this problem or not

[WiSo]

Forget my last post. Saw that it was also related to GraphicContext.cpp.

[WiSo]

Ok, I subtract 0.01 of texcoord_x and texcoord_y and the video looks fine now. Where would be the appropriated place to put this in?
At the moment and just for testing purpose I do this in CLinuxRendererGL::RenderSinglePass but this is done for every render frame.
Maybe in CLinuxRendererGL::CreateYV12Texture? But then for every texture wether npot or not?

[jmarshall]

I'm not sure if that's the correct fix or not. The reason is that that will technically affect the scaling of the texture. I'm not sure whether the effect would be significant or not. The only case to care about is if no scaling occurs (as then it should be a 1:1 mapping). Clearing the texture to transparent on creation would fix it as well, without any scaling issues being present.

d4rk would be the one to know

[WiSo]

Clearing the texture to transparent on creation would fix it as well, without any scaling issues being present.

d4rk would be the one to know

Damn and I wanted to ask where to clear as I dunno

[d4rk]

Sorry, didn't see this thread.

I'm not able to reproduce this but here's one thing to try: change

Code:

im.texcoord_x = ((float)im.width / (float)np2x);
im.texcoord_y = ((float)im.height / (float)divfactor / (float)np2y);

in 2344 in LinuxRendererGL.cpp to

Code:

im.texcoord_x = ((float)(im.width-1.0) / (float)np2x);
im.texcoord_y = ((float)(im.height-1.0) / (float)divfactor / (float)np2y);

[WiSo]

I'm a little puzzled here. The fix you suggest seems to work even though the line you mentioned belongs to software rendering (RGB textures) which I don't use on the Mac mini.
Also I never see the log line from line 2342 which should be in the log How does it work?

[WiSo]

Oh forget what I wrote I think I shouldn't test when I'm ill. I think I forgot to revert my changes and therefore it worked.
I'll have to test it again but I doubt that it worked when changing the texcoords in the software render.
I'll also test it some lines below which should work then

[elupus]

Hold of on that fix. I'm rewriting the code d4rk quoted, so i'll try to get this fixed at the same time.

[WiSo]

Ok great.