[Bf-committers] Re: [Bf-blender-cvs] CVS commit: blender/source/blender/src drawobject.c drawview.c editmesh_mods.c

[Stephane SOPPERA]

Ton Roosendaal wrote:

> Hi,

Hi,

> Hrms, I've removed the code already for zbuffer value checks, it was  
> just too unreliable with edges

I noticed that.

> or slightly non-planar quads. I also  wanted to have proper edge 
> selection for subsurf-optimal draw, not  possible with previous 
> method. Plus the zbuf method didnt even work on  linux with newer ATIs.

What do you think of the method I gave in my previous mail?

> I rather check on what exactly is wrong with your gfx setup. Whats 
> your  card?

A Matrox MGA G200 (quite old card) with (only) 8Mo Video RAM.
(maybe this problem won't last long for me since I'm seriously thinking 
upgrading my configuration ;-)

> OS?

The problem is reproduced on both Windows and Linux as long as hardware 
acceleration is used (on window, in 24 or 32bit mode, software rendering 
is used so it works and on linux hardware acceleration is possible only 
in 16bit)

> I can give some hints for (or a patch) with prints in the  code.

It seems that sample_backbuf function does not return correct values.
In framebuffer_to_index function, 16bits is used.

What kind of print do you need? Maybe should it be possible to dump the 
content of the backbuffer into an image file? Is it possible with 
blender API?

Thanks,

-- 
Stephane SOPPERA
http://perso.wanadoo.fr/stephane.soppera