https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79671
--- Comment #25 from rguenther at suse dot de <rguenther at suse dot de> --- On Tue, 28 Feb 2017, bernd.edlinger at hotmail dot de wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79671 > > --- Comment #24 from Bernd Edlinger <bernd.edlinger at hotmail dot de> --- > (In reply to Richard Biener from comment #23) > > (In reply to Bernd Edlinger from comment #22) > > > (In reply to Jakub Jelinek from comment #20) > > > > which fails also on x86_64-linux at -O2. And that testcase regressed > > > > with > > > > r223126. Now whether this is valid C++, no idea, placement new is > > > > messy. > > > > > > This test case can't be valid, suppose the A has a copy constructor > > > that that is also not called when B is moved around. > > > > The canonical fix is to put the type you placement new into the union > > storage > > into the union as regular member rather than having a char[] member in the > > union. > > Yes. Of course you cannot put a non-POD type in a union, > but maybe a pointer to A, that is probably what boost should > do in their functor class. You can put a non-POD into a union but then you need to provide copy/move/etc. constructors as they are otherwise default deleted.
