Re: master: Defensively code in type-width-in-bits

And taking interval-high makes it unsigned-type-width-in-bits. There's
already SB-C::INTEGER-TYPE-LENGTH, but it doesn't employ
type-approximate-interval.

On Sun, Apr 19, 2026 at 7:07 PM Stas Boukarev <[email protected]> wrote:
>
> I think type-approximate-interval should be used instead. (Which can
> still return NIL, and interval-high might also be NIL).
>
> On Sun, Apr 19, 2026 at 7:01 PM snuglas via Sbcl-commits
> <[email protected]> wrote:
> >
> > The branch "master" has been updated in SBCL:
> >        via  566883ab44ee474bfa5b12a00e8b6aeb2d946fdb (commit)
> >       from  c9317d951a80eab801e3a93e4edb346dc54fa302 (commit)
> >
> > - Log -----------------------------------------------------------------
> > commit 566883ab44ee474bfa5b12a00e8b6aeb2d946fdb
> > Author: Douglas Katzman <[email protected]>
> > Date:   Sun Apr 19 11:59:56 2026 -0400
> >
> >     Defensively code in type-width-in-bits
> >
> >     The receiving logic already potentially expected NIL
> > ---
> >  src/compiler/generic/utils.lisp | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/src/compiler/generic/utils.lisp b/src/compiler/generic/utils.lisp
> > index 85e6f5533..cdff56dab 100644
> > --- a/src/compiler/generic/utils.lisp
> > +++ b/src/compiler/generic/utils.lisp
> > @@ -593,7 +593,8 @@
> >    ;; Return (integer-length upper-bound) of numeric type. This is NOT a theoretical
> >    ;; smallest N bits needed to encode an element of type in a packed representation
> >    ;; (e.g. the interval 5..8 _could_ be stored in 2 bits) but we don't do that.
> > -  (integer-length (sb-c::interval-high (sb-c::numeric-type->interval ctype))))
> > +  (if (typep ctype 'numeric-union-type)
> > +      (integer-length (sb-c::interval-high (sb-c::numeric-type->interval ctype)))))
> >
> >  (defun env-system-tlab-p (env)
> >    #-system-tlabs (declare (ignore env))
> >
> > -----------------------------------------------------------------------
> >
> >
> > hooks/post-receive
> > --
> > SBCL
> >
> >
> > _______________________________________________
> > Sbcl-commits mailing list
> > [email protected]
> > https://lists.sourceforge.net/lists/listinfo/sbcl-commits


_______________________________________________
Sbcl-commits mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/sbcl-commits