Rendering Faults on Windows

This was reported during Ludlum Dare 48. Unfortunately I have no way to verify this at the moment…

The symptoms were:

1. A game was being made on a linux box, and all effects were showing up correctly. Specifically WebGL 2.0 renderer things like shapes.
2. When the game was passed over to the folks doing their art, they could never see the results, and it sounded to me like Indigo was falling back to WebGL 1.0.
3. However! According to Indigo’s logs it was still using WebGL 2.0!
4. When the devs on the linux boxes set the config.useWebGL1 flag, they got the same experience as their Windows based team mates.

Anything more to add @kag0?

[nrktkt]

yup, looks to be working for me on windows with ff

Also, unrelated, I like the game!

haha, thanks, glad to hear it

[davesmith00000]

Copying the description from the PR in case anyone is looking for the details:

Essentially the max number of buffer bindings on the Windows machines we have access to is 24,
but 24 is the _minimum_ number that the WebGL 2.0 detection process in Indigo looks for.
So I was foolishly (oversight) checking for 24 to be available, and then instantly setting two
pointers outside that range to 32 and 64 for the blend and custom UBO data bindings respectively.

This was not a problem on Mac or Linux, which is why it went undetected, but was immediately a
problem when we tried it on Windows, even a Windows VM. At least, on the machines we have for
testing.

The solution is simply to reduce Indigo's requirements by lowering the pointer offsets.