Patch Set 1:

> > > My latest suggestion is to use memcmp at the end of the
 > cloning,
 > > > not memcpy :)
 > > >
 > > > The robustness is about someone adding a field to a
 > HInstruction
 > > > class and not updating its associated Clone method.
 > >
 > > So the cost of keeping this as it is - just an extra function
 > > definition in each instruction for getting more secure, C++
 > > friendly and less error-prone code?
 > >
 > > Anyway if you have a strong opinion on this I can rewrite it.
 >
 > Eso esta bien

aaaaaa