fix the fatal bug of the convolution calculation part of the Minkowski sum

[Garbage123King]

Summary of Changes

Modify the convolution calculation part of the Minkowski sum. In the 'elimination of antennas' phase, before I fix it, only the antennas of the first segment were eliminated, skipping the possible subsequent antennas. For more details, please refer to the explanation in the issue. I modified it so that if a new antenna is merged, the newly merged antenna is used as curr for the next iteration; otherwise, after_next is used as curr.

Release Management

• Affected package(s): Minkowski_sum_2
• Issue(s) solved: fix minkowski_sum has fatal bug for some special linestring and polygon #8551

[efifogel]

First, good job on spotting and fixing.
A more simple (and I guess efficient) fix is simply replacing Line 614:
- cycle.insert(curr,
+ after_next = cycle.insert(curr,

[sloriot]

I updated the code. We agree on the fix now?

[sloriot]

cc @efifogel

[efifogel]

BW, consider the Minkowski sum of two 2 segments, that is, two degenerate polygons.
I've noticed that now, while the full-convolution method works, the reduced convolution (which is the default) fails.
The decomposition method also works on this case.
I guess more rigorous tests are needed, but the manual does not say that such degenerate cases are expected to work though. As Garbase123King said, mathematically, they should all work...

[sloriot]

Successfully tested in CGAL 6.1-Ic-15

[github-actions]

This pull-request was previously marked with the label Tested, but has been modified with new commits. That label has been removed.