Re: GCC CVS mainline does not build on i686-linux (rtlcheck failure)

[Richard Henderson <rth at redhat dot com>]

On Mon, Oct 14, 2002 at 09:46:20AM +0200, Andreas Jaeger wrote:
> Richard Henderson <rth@redhat.com> writes:
> 
> > On Sat, Oct 12, 2002 at 03:46:46PM +0100, Graham Stott wrote:
> >>     if ((TREE_CODE (decl) == VAR_DECL || TREE_CODE (decl) == PARM_DECL)
> >>         && DECL_REGISTER (decl)
> >> +       && REG_P (DECL_RTL (decl))
> >>         && REGNO (DECL_RTL (decl)) < FIRST_PSEUDO_REGISTER)
> >
> > Lemme guess: the problem ocurrs with a variable that was declared
> > as "register int foo;", but this is a non-optimizing compile, so we
> > dropped it to the stack anyway?
> 
> register long double x;
> 
> But it's an optimizing compile,

That's strange.  I must be missing something.
Oh well.  It's clear at least what _one_ problem
with the code as-is would be.

Patch approved.


r~