Re: [PyOpenGL-Users] Reading Depth Component - not working anymore?

[Mike C. F. <mcf...@vr...>]

Hart's Antler wrote:
> --- "Mike C. Fletcher" <mcf...@vr...> wrote:
>   
...
> i dont have a stripped down example, but the code is just 3 lines you can insert into any pyopengl
> demo.
> data = glReadPixels(0,0,x,y, GL_DEPTH_COMPONENT, GL_UNSIGNED_BYTE) # note x,y are screen size
> img = Image.fromstring( 'L', (x,y), data )  # requires PIL
> img.save( 'z.bmp' )  # when you open z.bmp you will find that it is blank - all white.
>   
I can reproduce the effect you are reporting if I have a large frustum
where the near and far clipping plans are at .01 and 1000 respectively. 
If I bring those down so that the geometry is just bracketed by the
clipping planes, the depth buffer "image" shows up as black for that
touching the front clipping plane (lowest depth) and white for that
touching the back clipping plane (highest depth).

That is, the wrapper *seems* to be working fine (though the extremely
high (.999+) values in the float mode with a large frustum surprise me,
they *seem* to represent real depth values).  The code I'm using to test
is available in the CVS here:

http://pyopengl.cvs.sourceforge.net/pyopengl/OpenGLContext/tests/saveimage.py?view=markup

You can see the commented-out code that uses GL_FLOAT type and maps the
values to the 0-255 range using array operations as well.  I'm open to
the possibility that something is going wrong in the wrapper, but it
*looks* like something unexpected in the GL, rather than something wrong
with the wrapper.

This URL

    http://www.opengl.org/resources/faq/technical/depthbuffer.htm

would seem to suggest that the likely culprit for this kind of error is
specifying a zNear which is too near to the camera.

Hope this helps,
Mike

-- 
________________________________________________
  Mike C. Fletcher
  Designer, VR Plumber, Coder
  http://www.vrplumber.com
  http://blog.vrplumber.com