Tail padding, in sumary

[Mark Mitchell]

There are several separate tail-padding issues that have been discussed
over the last few days and I want to make sure that we are all on the
same page about all of them as I am trying to revise the specification
accordingly.

HP, Intel folks please confirm statements below.

1) Should we reuse tail padding at all?

   This was my original question; the specification was inconsistent.

   The answer was decided as yes.

2) Should padding be at the bit-field level, or only at the byte level?

   For example, given:

     struct A { virtual void f(); int i : 3; };
     struct B : public A { int j: 2; };

   Is it OK to pack B::j into the same byte as A::i?

   GCC does this (even though I think it a horrid idea due to the fact
   that you can no longer create efficient copy constructors).

   Neither Intel nor HP do this.

   Therefore, I believe the spec should be changed to clarify that --
   while the bytes following A::i are OK for use as padding -- the byte
   containing A::i is not.  (And, of course, GCC should be fixed.)

3) Should virtual bases be handled differently than non-virtual bases?

   For example, given:

     struct A { virtual void f(); char c1; };
     struct B { B(); char c2; };
     struct C : public A, public virtual B {};

   Is B supposed to be packed into the tail-padding of A?  In the case
   that it is non-virtual, it is.  However, in the virtual case GCC does
   not reuse the tail-padding.

   Probably, this is also a GCC bug, but I would like to know what HP
   and Intel do.

   HP, Intel?

-- 
Mark Mitchell                mark at codesourcery.com
CodeSourcery, LLC            http://www.codesourcery.com