[PyOpenGL-Devel] [ pyopengl-Bugs-1979002 ] Segfault with glReadPixelsf

[SourceForge.net <no...@so...>]

Bugs item #1979002, was opened at 2008-05-30 01:34
Message generated for change (Comment added) made by nobody
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105988&aid=1979002&group_id=5988

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: GL
Group: v3.0.0
Status: Open
Resolution: None
Priority: 5
Private: No
Submitted By: Nobody/Anonymous (nobody)
Assigned to: Nobody/Anonymous (nobody)
Summary: Segfault with glReadPixelsf

Initial Comment:
PyOpenGL version: 3.0.0b2
Numpy version: 1.0.4
Platform: Windows XP SP 2
GPU: NVidia 8800GTS, 169.21 driver version


Attempting to perform a glReadPixelsf on a 512x512 frame buffer causes a crash. In fact, every format I tried besides glReadPixelsub crashes.

Thanks,

Eddy Talvala, ta...@st...

Minimum testcase that shows crash on my system attached.


----------------------------------------------------------------------

Comment By: Nobody/Anonymous (nobody)
Date: 2008-06-01 12:49

Message:
Logged In: NO 

Further experimentation appears to show that the bug is in
OpenGL.images.createTargetArray

createTargetArray attempts to create, based on the pixel format and the
color channel type, a zero array of sufficient size.  However, the
calculation seems incorrect for the case of an RGB float array, at least.

After adding some prints inside createTargetArray, and calling
glReadPixels with:

  readback_image1 = glReadPixelsub(0,0,512,512,GL_RGB)

I get the following values for the variables inside createTargetArray:
  formatBits: 24
  typeBits: 8
  lastDim, remainder: 3, 0
  dims: (512, 512, 3)
  arrayType: <class 'OpenGL.arrays.arraydatatype.GLubyteArray'>

Which is correct, and works fine.  However, the same variables for the
call

  readback_image2 = glReadPixelsf(0,0,width,height,GL_RGB)

aren't correct:

  formatBits: 24
  typeBits: 32
  dims: (512, 512)
  arrayType: <class 'OpenGL.arrays.arraydatatype.GLfloatArray'>

Most importantly, the dims tuple is wrong - the data has three color
channels, each of which is a float. A 512x512 float array is 3 times too
small to contain it, and explains why the code then crashes when
simple.glReadPixels is called using the array returned by
createTargetArray.

Also, formatBits=24 makes little sense for a RGB float texture (where a
pixel takes up 96 bits) unless formatBits=24 simply means '3 components'.

It would seem at first glance that lastDim should simply be set based on
format, not based on comparisons between formatBits and typeBits. Perhaps
this code is holdover from an earlier version, where multiple color
channels were stuffed into a single array entry, which no longer appears to
be the case.

Hope this helps,

Eddy Talvala, <lastname> at stanford.edu

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=105988&aid=1979002&group_id=5988