Quantizing depth texture information
Right now if you create a GL_DEPTH_COMPONENT24 texture and use glCopyTexImage2D with GL_DEPTH_COMPONENT to fill that texture from a 24bit depth buffer, only the top 16bits of the 24bit texture get copied.
Right now you have to you have to explicitly say GL_DEPTH_COMPONENT24 to copy all 24bits into the destination texture. This seems like incorrect behavior (and differs from what the NVidia OpenGL implementation does).
Example:
glBindTexture(GL_TEXTURE_2D, 1);
glTextureParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAX_LEVEL, 0);
glTexImage2D(GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT24, w, h, 0, GL_DEPTH_COMPONENT, GL_FLOAT, NULL);
glCopyTexImage2D(GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT, 0, 0, w, h, 0);
Workaround:
glCopyTexImage2D(GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT24, 0, 0, w, h, 0);