Re: [PATCH] Kill TREE_INVARIANT, take 2

[Paolo Bonzini <bonzini at gnu dot org>]

> > +static bool
> > +tree_invariant_p (tree t)
> > +{
> > +  tree inner = skip_simple_arithmetic (t);
> > +  return tree_invariant_p_1 (inner);
> > +}
>
> Should tree_invariant_p just handle the recursion directly?  For 
> example, if "a" and "b" are both invariant, then so is "a + b".

Actually skip_simple_arithmetic handless this nicely.  I didn't do full 
recursion in tree_invariant_p because Ada uses s_s_a.

> I think this should be "decl_address_invariant_p".  Otherwise, it sounds 
> like the declaration's value is invariant, when, in fact, it's the address.

Ok.

Paolo