22 May
2008
22 May
'08
10:25 a.m.
Good point. It would have to be restricted to either noncomplex types or side-effect free initializations, I guess.