On Tue, Nov 8, 2022 at 12:59 PM Camm Maguire <c...@maguirefamily.org> wrote:
> The NaN report gives false now as you note, > 2.0^1024.1 still gives the inf for testing purposes. I think the false > is in simp-expt in maxima, but haven't got further. Most of his results > had been fixed some time ago. What? I don't understand. The false in SIMP-EXPT ... I don't understand what you mean. Also I'm not sure who "he" is, in reference to "his results". > The dribble error only omits error header output (which is sent to a > different file > descriptor) -- all newlines are identical addressing the original > complaint. I can get the error header in there too if this is > important. No, not important. It's likely that dribble is very rarely used, and the defect is minor. > >(setq b (- a) c (- a a)) > #<-nan> Minor wart here -- should be #<nan>, NaN doesn't have a sign. > >(dolist (l '(< <= = >= > /=)) (print (list (funcall l c c) (funcall (compile > >nil `(lambda (x y) (declare (long-float x y)) (,l x y))) c c)))) Looking over the results of float comparisons, looks right to me. best, Robert PS. I successfully built Version-2_6_13pre124 from a tarball on an ancient Ubuntu system (32 bit x86, gcc 4.8.4 iirc), it builds Maxima successfully, run_testsuite reports, I think, the same failures and unexpected successes as others have reported.