Dan Baston

Looks like the area of the result changed from 517747 in 3.9 to 517642 in 3.10. I can try bisecting on that.

Breaking commit is https://github.com/libgeos/geos/commit/0b2c29abd785bea95dc83f1777335ad48e744f8d

It seems that the commit produces a different sequence of pairwise unions, one of which produces an incorrect result both in JTS and GEOS: A ``` MULTIPOLYGON (((32716.64690000005 6953061.072500002, 32728.48959999997...

The second component of A is a tiny sliver that intersects B: 

> Is there a similar tolerance issue around arcs with infinite radius (aka straight lines (aka three control points in a line))? Good point. See the final (failing) test added...

A couple of recent notes on this: 1. I plugged in the following robust quadratic equation solver: https://github.com/archermarx/quadratic . It did not help with failing `CircularArcIntersector` test 40, but may...

Yes, I have to remind myself that the baseline is linearizing everything, and we're clearly doing better than that. Test 40 isn't a catastrophic failure in any case, it's just...

Current status of this: tests passing except for: - CircularArcIntersector test 40 - the computed intersection point is outside the arbitrary distance tolerance - CircularArcIntersector 42 and 53. These relate...

I poked at this a bit, writing a KdTree implementation that uses bulk loading so it can be fully balanced, with nodes in a contiguous block of memory. That gets...

> Why not merge? Merge conflict, modest benefits, general conservatism about differences from JTS