Has the discussion in the linked issue been resolved?

@jensmaurer: any thoughts on this?

I think the main gist here is that using a pointer-to-base when actually pointing to a derived class is fine in most cases, but not in the array indexing case. Changing the example as suggested is losing that point.

I think the main gist here is that using a pointer-to-base when actually pointing to a derived class is fine in most cases, but not in the array indexing case.

Using a pointer-to-base when actually pointing to an object of derived class type is UB in most (if not all) cases (member access, member function call etc.).

Using pointer-to-base which points to the base subobject is fine in all cases (even in the array indexing case). What you can violate with such a pointer, is [expr.add]/4, not [expr.add]/6.

@jensmaurer I'm not sure what to do here. Should we move this example to p4 and edit the example in p6 like proposed? Or just keep the status quo?

I like the int / unsigned int example in p6, because it is mildly surprising.

I also like to retain the base / derived situation, but maybe it's worthwhile to formulate that in terms of "base class subobject" in p4, as a special case for when the p4.2 precondition isn't satisfied.

@languagelawyer Could you take another stab at this then?

@jensmaurer not sure what you meant, but I did this. Note, the Note is outside the itemize environment, unlike in the next paragraph.

@jensmaurer Could you please take another look here?

@languagelawyer could you please have a look?