Re: RFA: PATCH to initializer_constant_valid_p for c++/38880

[Jason Merrill <jason at redhat dot com>]

Richard Guenther wrote:
> On Thu, 19 Feb 2009, Jason Merrill wrote:
>
> > > > And to then wonder why
> > > > POINTER_PLUS_EXPR wasn't used instead of PLUS_EXPR.
> > > Because http://gcc.gnu.org/viewcvs?view=rev&revision=143442 made the
> > > compiler stop folding to POINTER_PLUS_EXPR.
> > Do you two think it would make sense to put back that folding for the case of
> > PTR+small constant?
>
> I don't see how a small constant makes that folding more correct.

Yeah, I guess unless both operands are constants, you can get corner 
cases that will break, and reintroducing the POINTER_PLUS_EXPR fold just 
for two constant operands seems a bit silly.

Ian, do you still have reservations about my patch?

Jason