(2021-03-24, 22:41)Magnus Forsberg Wrote: But if an author is rounding of the values of a rotated primitive in his design, it will introduce a miscalculation, and cause thin, hairline gaps. Right?
I don't know. In theory, yes, but I'm not sure if it happens in practice. (Note: when I say I'm not sure, I really mean that I'm not sure; I'm not saying I think it doesn't happen.) If it does happen, then it might be better to relax the 3-decimal-place restriction for transformation matrices in parts.