Re: VRP wrapping MULT_EXPR

[Marc Glisse <marc dot glisse at inria dot fr>]

On Fri, 3 Aug 2012, Marc Glisse wrote:

> On Fri, 3 Aug 2012, Richard Guenther wrote:
>
> > > +/* Some quadruple precision helpers.  */
> > > +static int
> > > +quad_int_cmp (double_int l0, double_int h0,
> > > +             double_int l1, double_int h1, bool uns)
> > > +{
> > > +  int c = double_int_cmp (h0, h1, uns);
> > > +  if (c != 0) return c;
> > > +  return double_int_ucmp (l0, l1);
> > > +}
> >
> > I suppose that's appropriate for double-int.h as static inline function.
> >
> > Ok with or without moving it (we can do that as followup anyway) if
> > testing is ok.
>
> Thanks, I applied it without moving, and am now doing a quick bootstrap of 
> the following (to make sure I didn't forget a ';' during the copy-paste).
>
> 2012-08-03 Marc Glisse <marc.glisse@inria.fr>
>
> 	* tree-vrp.c (quad_int_cmp): Move ...
> 	* double-int.h (quad_int_cmp): ... here.

Hmm, that fails early. Apparently, the compilation of gengtype.o includes 
double-int.h and is done with -fkeep-inline-functions. The function 
quad_int_cmp, although it is static inline, then ends up in gengtype.o. 
But that function uses double_int_cmp, and gengtype doesn't link 
double-int.o, so I get an undefined reference failure. I guess I'll leave 
the function in tree-vrp.c for now.

--
Marc Glisse