weird code generation differences without optimization [was Re: purify 4.1 and egcs still appear to be incompatible.]

[Jason Merrill]

>>>>> Ian Haggard <ian@shellus.com> writes:

> So I was wrong in saying that the single byte taken up by an empty class
> should be (or is even allowed to be) copied by a conforming C++
> implementation.

Fie.  The standard says nothing at all about padding bytes; the compiler is
free to do anything with them.  I will agree that the current behavior is
suboptimal, but it's not unconforming.

Someone could fix this in do_build_copy_constructor by checking to see if
the class is totally empty, and doing nothing in that case.

Jason