Re: [PyOpenGL-Users] Textures on Bad Hardware

[Almar K. <alm...@gm...>]

On 1 August 2010 02:06, Ian Mallett <geo...@gm...> wrote:

> On Sat, Jul 31, 2010 at 3:05 PM, Dirk Reiners <dir...@gm...>wrote:
>
>> All of that looks pretty straightforward. The only thing I can think of,
>> but
>> that should give a different error, is textures that are not power-of-two
>> size.
>> What are the sizes of your textures?
>>
> I hadn't even thought of that!
>
> And after some testing, it seems that's the problem.  My test code can be
> found here:
>
> http://download577.mediafire.com/jnza7ybur4qg/ievqu4buekair6k/test_texture.zip
>
> I'm surprised at how ubiquitous using power-of-two textures must be.
> Basically no other applications seemed to have problems.  I never had
> problems, so I guess I assumed it was irrelevant.
>
> Uggggh I assume I'll have to rewrite my texture code so that it can take an
> argument to scale surfaces to the nearest factor of two.  Any other way?  I
> know of ARB_texture_rectangle, but that sounds horrible.  Blech.  I hate
> Intel.
>


The pretties way of doing this is checking the openGl version and if its <
2.0, you need power-of-two textures. However, I recently encountered a
system with an ATI card, that did have OpenGl >=2.0, but did NOT support
non-power-of-two textures. So my current implementation just tries resizing
the texture if it fails to initialize, or if OpenGl <2.0.

Cheers,
  Almar