I wanted to thank you, because pointing to unsafeCoerce was a nice bit of serendipity. I'll admit I thought it was nitpicking initially, but I did a quick pass at cleaning things up last week after your mail (not perfect, but this was a prototype so I'm not looking for pretty): https://github.com/elisehuard/game-prototype/commit/5cf85fa60f325dbe610b5e1f... and that fixed the texture mapping! Consider me informed on that issue :) (as for which example, I used Redbook4 but I admit to extrapolating a little, this was my mistake) Thank you, On 2 July 2014 13:11, Jason Dagit wrote:

On Wed, Jul 2, 2014 at 9:33 AM, Elise Huard wrote:

...

Hi Jason, thanks for taking the time to answer. So trying to read between the lines: - do you have any _working_ examples of texture mapping with OpenGLRaw and low-level Haskell I could have a look at? (anything would help at this point)

I do. I converted some of the old nehe tutorials: https://github.com/dagit/nehe-tuts

If you look through the history you can find examples that use a very very old version of OpenGL (instead of OpenGLRaw). The OpenGL style in those examples is old and no longer the recommended way. The examples use immediate mode which is now deprecated in favor of shaders.

I think lesson 6 makes a cube and throws some textures on it. The lesson numbers are meant to match these: * http://nehe.gamedev.net/tutorial/lessons_01__05/22004/ * http://nehe.gamedev.net/tutorial/lessons_06__10/17010/

...

- do you think the unsafeCoerce (which I borrowed from examples) that transform CpFloat to GLdouble might be the issue, or is this just a general stylistic comment?

Not just stylistic. As for correctness, I think the answer varies by platform. Specifically, I think CFloat is not required to match GLdouble in size (in fact, doubles should be expected to be twice as big). While both types should correspond to IEEE 754 floats the bit width is very likely different and that could easily cause the interpretation of the values to go wonky. Imagine passing 64bits when only the first 32bits correspond to a valid number or truncating a 64bit value to 32bits.

Which examples use unsafeCoerce for this? I'm just sort of curious so I can recommend people not to use them :)