MauriceHubain
MauriceHubain
> Note that BOOST_GEOMETRY_ROBUSTNESS_ALTERNATIVE is deprecated, and will be removed This is definitely fine. We just tried that flag, because it seemed to fix our problems with the difference function.
If you change your example in a way, that poly2 is a polygon, not a multipolygon (as in our test cases), the test fails for double and long double precision:...
Nevertheless, I will check, if changing both input geometries to multipolygon, would solve all our issues.
Unfortunately I found a first example (from our "real life" issues), where also using mutlipolygon does not give the expected results: mp_t poly1; bg::read_wkt("MULTIPOLYGON(((1.2549999979079400 0.85000000411847698, -1.2550000020920500 0.84999999897038103, -1.2549999999999999 -0.85000000102961903, 1.2549999999999999...
Sure, I can file a new ticket. But this scenario was already part of the provided test bench.
> We can really not work with such a bench, sorry. Because things are fixed one by one, there can (sometimes) be months in between and handled by different persons....
I tested it on branch "boost-1.89.0.beta1" and got the same result.