Suggestions for new tests and infrastructure improvements. Not in any particular order. ------------------------------------------------------------------------ Description: For common extensions, verify that GetProcAddress returns non-NULL pointers for all the functions that are supposed to belong to each extension. Submitted by: Allen Akin ------------------------------------------------------------------------ Description: Change the common.mak defaults for Linux to match the OpenGL ABI standard. Submitted by: Allen Akin ------------------------------------------------------------------------ Description: Test glDrawBuffer, especially the GL_FRONT_AND_BACK and GL_NONE options with glDrawPixels, points, lines, triangles. Submitted by: Brian Paul ------------------------------------------------------------------------ Description: Test glColorMask on front buffer and back buffer, with glDrawPixels, points, lines, triangles. Submitted by: Brian Paul ------------------------------------------------------------------------ Description: Test glPolygonMode in all permutations. Perhaps also combine this with glCullFace. Submitted by: Brian Paul ------------------------------------------------------------------------ Description: Test two sided lighting, also with glPolygonMode permutations. Submitted by: Brian Paul ------------------------------------------------------------------------ Description: Test glTexEnv modes. For extra credit, test with multitexture. Submitted by: Brian Paul ------------------------------------------------------------------------ Description: Mipmap filtering tests. See: http://www.zdnet.com/etestinglabs/stories/bi/0,8829,2408223,00.html Submitted by: Allen Akin , in response to observation by Michael Jones ------------------------------------------------------------------------ Description: Support standalone Voodoo cards under Win32. Suggest using the glsetup API. Neal says: One problem which you may not be aware of is that its hard to determine whether or not 3dfxvgl.dll can actually be used safely. The problem is that even if the driver's been installed, the 3dfx card might have been taken out later, leaving a driver which faults as soon as you load the 3dfxvgl dll. As far as I know, the only truly safe way to deal with this is to check for the presence of the Glide dlls, load glide if they're present, and then use the Glide API (grGetNumBoards or something with a vaguely similar name) to actually detect the presence of the hardware. If its there, you should be able to load 3dfxvgl.dll safely. Submitted by: Johan Smet , Neal Tringham ------------------------------------------------------------------------ Description: Support dynamic display resolution changes under Win32. Win95 doesn't really support this well, and some drivers don't handle it correctly for Win98. One workaround would be to document that users should change the display resolution manually before running glean. Submitted by: Johan Smet ------------------------------------------------------------------------ Description: Add an option to allow overwriting the results directory. Submitted by: Neal Tringham ------------------------------------------------------------------------ Description: Add an option to allow selection of a particular OpenGL driver (for systems on which multiple drivers may coexist). Submitted by: Neal Tringham ------------------------------------------------------------------------ Description: Support language localization. Submitted by: Neal Tringham ------------------------------------------------------------------------ Description: Add a test that determines the mipmap LOD bias. Gary says: I would recommend someone write a simple test subroutine that determines the LOD bias. This could be done by rendering a small texture map, e.g. 2x2 or 4x4, with mipmaps at varying texel:pixel ratios and then reading the pixels back from the framebuffer. The code could then determine exactly where the LOD changes, and thus what the implicit LOD bias is, at least for non-perspective textures. Submitted by: Gary Tarolli ------------------------------------------------------------------------ Description: Write a test to check operations on shared dlists and textures. Submitted by: Kevin Martin ------------------------------------------------------------------------ Description: The ttexcombine test sometimes fails because a texture format has less precision than the framebuffer format, and the test's error tolerances are based solely on the framebuffer format. Should modify tolerances to take texture format into account. Submitted by: Allen Akin