Subject: ECL 9.12.2 causes multiple regressions in Maxima
From: Raymond Toy
Date: Fri, 04 Dec 2009 15:52:19 -0500
Aleksej Saushev wrote:
> Hello!
>
> I observe this with Maxima 5.19.2 and CVS HEAD (snapshot as of today
> midnight UTC):
>
>
> Using Lisp ECL 9.12.2
> Distributed under the GNU Public License. See the file COPYING.
> Dedicated to the memory of William Schelter.
> The function bug_report() provides bug reporting information.
> (%i1) run_testsuite()
> Running tests in rtest16:
> ********************** Problem 173 ***************
> Input:
> 60
> 2 - 1
> float(-------) - 1
> 60
> 2
>
>
> Result:
> - 1.110223024625157e-16
>
> This differed from the expected result:
> 0.0
>
> ********************** Problem 174 ***************
> Input:
> 1000
> 2 - 1
> float(---------) - 1
> 1000
> 2
>
>
> Result:
> - 1.110223024625157e-16
>
> This differed from the expected result:
> 0.0
>
These are known "issues". Maxima expects the conversion from rationals
to floats to be done by rounding the rational to the nearest float. All
other Lisps I've tried do this, but ecl does this by truncating so the
answer is not zero. This seems fairly deeply tied to ecl's use of gmp
to do the conversion, which only does truncation.
For the others failures, those seem to be new, and suggest something has
changed in ecl. With ecl 9.8.4, I only get a handful of test failures.
Ray