@jschueller thanks for the fast PR!

As far as I understand it, for the LLVM bug to trigger, link-time optimisation must be enabled - at least judging from this bugreport:

llvm/llvm-project#71196

I was thus wondering if it would make sense to just disable LTO if clang>=17 is detected, rather than injecting the -fno-assume-unique-vtables flag. I am not sure though, what do you think?

EDIT: I am talking about the PAGMO_ENABLE_IPO option, just to be clear.

I propose to keep the new flag, but only use it when IPO is enabled
your solution would be way shorter though, you decide

will close #579

@jschueller sorry for getting back to this only now...

As far as I understand from #586, switching off IPO actually did not prevent the bug from manifesting right? If that is the case and it is ok with you, I would slightly edit this PR to enable -fno-assume-unique-vtables regardless of whether or not IPO is enabled.

oh, right, I did not take that into account
done

@jschueller thank you!