Hi,
My comment isn't about the technical aspect of the patch itself -- it
is a simple patch. But we don't seem to understand yet the
implications of "final". As was observed on the standard reflectors,
the appropriate trait might actually need to be binary instead of
unary as this patch implements -- for the purpose of empty base
optimization which is where the problems pop up.
Interesting the binary vs unary thing, makes me curious to re-read the
whole thread. Then indeed, makes sense to be safe and not documenting
for 4.7 the unary version which we are going to use right now for the
immediate needs of the library.
Paolo.