Comment 13 for bug 1753224

@Tomasz, that equation is constant for r (they drop out of the numerator & denominator). It seems to be OK for larger arcs (78 segments in hi-res, 45 in low), but probably returns way more than necessary for smaller ones.

Is the 1% / 0.33% error based on circle area? Seems like the error needs to be based on mils or something not relative to the circle size.