Adam Wulkiewicz

@tinko92 I think we should have another level of abstraction between a CS-specific part of what is now a strategy and algorithmic part. Note that currently the CS-specific part is...

There are 4 parts of the randomization: - distribution - randomizing algorithm - CS-specific strategy - generator Randomizing algorithm has to know the domain/geometry and may need to preprocess data...

> I think the policy solution is strictly superior to the tag solution This is my suggestion number 2, correct? You differentiate between policies and tags while I considered policies...

@uncleysh @ayushgupta138 Thanks for the report and trying to reproduce it. > Does boost geometry only support Windows? No, with different compilers floating point operations can give different results. There...

Yes, this is caused by the fact that too small coordinate_type is used. And currently Boost.Geometry does nothing about it. You could try passing a CalculationType of greater size only...

> Does `variant=debug` vs `variant=release` make a difference? @mloskot No, the failures are the same in both cases. The last one is compilation error so it doesn't matter. `algorithms_line_interpolate` fails...

@vissarion I don't know why exactly but it seems that after merging https://github.com/boostorg/geometry/pull/840 `line_interpolate` test also started to fail with gcc-6.3 the same way: https://app.circleci.com/pipelines/github/boostorg/geometry/165/workflows/5d3af105-4f5c-49b5-a4b6-7a128f94347c/jobs/12439 It also fails for me...

The issue was automatically closed after merging a fix for one of the failures. So I reopened it and edited the issue description.

Your assumption is correct. I.e. the strategy should define the coordinate system used by the operation. However currently some core operations like e.g. points comparison are performed using the coordinate...

@barendgehrels I'm debugging the case with robustness enabled. `sym_difference()` calculates difference two times and then union of the results here: https://github.com/boostorg/geometry/blob/develop/include/boost/geometry/algorithms/sym_difference.hpp#L134-L152 In all of these calculations the same rescale policy...