chaser for cp-tree.h

[Zack Weinberg]

On Fri, Jan 21, 2000 at 01:34:46PM -0800, Zack Weinberg wrote:
> On Fri, Jan 21, 2000 at 03:23:00PM -0500, Kaveh R. Ghazi wrote:
> >  > From: Zack Weinberg <zack@wolery.cumb.org>
> >  >  
> >  > > Zack,
> >  > > 
> >  > >       What happened to this patch?
> >  >  
> >  > No one ever responded to me about it.
> >  >  
> >  > It may be that TYPE_ALIAS_SET should be changed to unsigned int
> >  > instead.  I don't *think* the alias code uses negative set numbers...
> >  > zw
> > 
> > What is the default promotion when ?: has mixed signed-ness?
> > 
> > I think your cast doesn't change the code behavior if its being promoted
> > to unsigned anyway.  So it should be safe IMHO.
> 
> The default promotion is to unsigned, so yes, it's a safe patch.  I am
> not sure it is the right patch.  It may be that this is the right
> patch - I am testing it now.

It is an effective patch (with some more tweaks) but after some
thought, I believe the original patch for cp-tree.h was the correct
one.  Alias sets are normally positive, but -1 has special meaning,
and it is silly to force the language-independent code to use unsigned
ints just because the C++ front end wants to stick bitfields in that
slot.

I'm not going to commit it without an ack from the C++ people, though.

zw