Re: PATCH: add an assert to flow.c

[Ian Lance Taylor <iant at google dot com>]

Ben Elliston <bje@au1.ibm.com> writes:

> On Mon, 2006-12-18 at 17:12 -0800, Ian Lance Taylor wrote:
> 
> > > Index: flow.c
> > > ===================================================================
> > > --- flow.c      (revision 120017)
> > > +++ flow.c      (working copy)
> > > @@ -1557,6 +1557,7 @@ initialize_uninitialized_subregs (void)
> > >               if (param.retval != NULL_RTX)
> > >                 {
> > >                   start_sequence ();
> > > +                 gcc_assert (CONST0_RTX (GET_MODE (param.retval)) != NULL_RTX);
> > >                   emit_move_insn (param.retval,
> > >                                   CONST0_RTX (GET_MODE (param.retval)));
> > >                   insn = get_insns ();
> > 
> > How could that assert ever fail?  What mode did you get?
> 
> CDImode, which has a NULL RTX for that entry in the constant zero table.

Would it make sense to add the complex modes to init_emit_once?
After, we initialize CONST0_RTX for the vector modes.

Ian