Intersection error on barely touching polygons

[ZuiDet]

I found a very strange bug for the intersection of the following two polygons:

poly1 = GI.Polygon([[(0.0, 0.9), (0.09999999999999998, 0.9), (0.09999999999999998, 1.0), (0.0, 1.0), (0.0, 0.9)]]);
poly2 = GI.Polygon([[(-0.1000000000000057, 0.9000000000000031), (-4.761357969939932e-15, 0.8999999999999989), (1.0572338493419614e-30, 1.0), (-0.09999999999999998, 1.0), (-0.1000000000000057, 0.9000000000000031)]]);

poly_int = GO.intersection(poly2 poly1; target=GI.PolygonTrait()); returns an empty list as expected

However,
poly_int = GO.intersection(poly1 poly2; target=GI.PolygonTrait());

returns the left polygon completely.

I assume, this is connected to #193.

[asinghvi17]

Huh, on my machine and on PR #297 (which has no changes to predicates beyond adopting AdaptivePredicates.jl) I see the same result but with the order of polygons inverted.

Yeah it does look like this has something to do with tracing or maybe the bounce functions being bad on our end. They do intersect, so the intersection should return something, meaning that both of these results are wrong and it should probably return a triangle...

[asinghvi17]

LibGEOS gives me this:

julia> import GeometryOps as GO, LibGEOS

julia> poly_int = GO.intersection(GO.GEOS(), poly2, poly1)
GeoInterface.Wrappers.LineString{false, false}(LINESTRING (0 1, 1.0572338493419614e-30 1))

julia> poly_int = GO.intersection(GO.GEOS(), poly1, poly2)
GeoInterface.Wrappers.LineString{false, false}(LINESTRING (1.0572338493419614e-30 1, 0 1))

so there is that to consider as well