Broken textures

[Heath123]

model.zip Blender: In game:

[Heath123]

If you get very close using Free Look you can see this

[Heath123]

Converting it back to DAE and importing it, the UV map is way too big (as expected): https://user-images.githubusercontent.com/13787163/129484517-49ad95f7-8ea9-404a-80c1-3372526d1b3d.mp4

[Heath123]

Scaling it by 0.00001525 gets it back to about the right position, and 1 / 0.00001525 = 65573.7704918 which is very close to 2 ^ 16 or 65536

[Heath123]

So it's 65536 times too big

[Heath123]

Oh: ERROR: Geometry Cylinder#0 divisor 16 out of range That definitely seems like the problem

[Heath123]

if self.divisor >= 16:
    AutoFix.error('Geometry {} divisor {} out of range'.format(self.name, self.divisor))
    self.divisor = 0
    result = True

So you set divisor to 0, which would make everything 65536 times bigger than it would be Removing the check fixes it for me, why does it need to be there? (also the warning says geometry but point.py is also used for texture ccordinates)

[Robert-N7]

I believe this was an issue that I fixed in my development branch, still working to get it passing all checks though... There was an off by one bug detecting the precision needed for texture coordinates, which resulted in using the 16-bit format with a divisor of 16… as you see that fires the check that resets it to 0. You may think it's not necessary but in ctools models it's very common to have invalid divisor values. I'm not sure how it fixed it in your case, as shifting a 16 bit integer 16 times results in... Well nothing

[Heath123]

I believe this was an issue that I fixed in my development branch, still working to get it passing all checks though... There was an off by one bug detecting the precision needed for texture coordinates, which resulted in using the 16-bit format with a divisor of 16… as you see that fires the check that resets it to 0. You may think it's not necessary but in ctools models it's very common to have invalid divisor values. I'm not sure how it fixed it in your case, as shifting a 16 bit integer 16 times results in... Well nothing

If you want to work out why I attached a model that should reproduce it in the issue

[Heath123]

You may think it's not necessary

I know it's there for a reason, I just meant I'd need to know what it does to fix it properly