Re: [Chicken-hackers] [PATCH] Do not assume all fixnum literals will be fixnums at runtime [was: Re: [PATCH][5] types.db entries for the overflow-aware ops]

[Peter Bex]

On Wed, May 24, 2017 at 09:25:15PM +1200, Evan Hanson wrote:
> Hi folks,
> 
> Good discussion, thank you Peter and Lemon for considering this.
> 
> > diff --git a/lfa2.scm b/lfa2.scm
> > index 0fd4612..4c7ff84 100644
> > --- a/lfa2.scm
> > +++ b/lfa2.scm
> > @@ -173,17 +173,14 @@
> >        ;; a simplified variant of the one in scrutinizer.scm
> >        (cond ((string? lit) 'string)
> >         ((symbol? lit) 'symbol)
> > +       ;; Do not assume fixnum width matches target platform's!
> > +       ((or (big-fixnum? lit)
> > +            (bignum? lit))
> > +        'integer)
> >         ((fixnum? lit) 'fixnum)
> > -       ((bignum? lit) 'bignum)
> >         ((flonum? lit) 'float)
> >         ((ratnum? lit) 'ratnum)
> >         ((cplxnum? lit) 'cplxnum)
> 
> You use `bignum?' for the non-fixnum part of the width check here in
> lfa2.scm, but `small-bignum?' in scrutinizer.scm -- why is that? (This
> is in the chicken-5 patch.)

Good catch, that's just an oversight.  Both should use small-bignum?,
and of course that means the bignum? check should be restored.

Cheers,
Peter

signature.asc
Description: Digital signature