Am Sonntag, den 23.08.2009, 16:01 -0500 schrieb Barton Willis:
> I committed improved (I hope) functions float_approx_equal and
> bfloat_approx_equal. I used the method given in Knuth, "The Art of
> Computer Programming," 3rd edition, page 233.
>
> The test suite (Clozure CL) ran OK---let me know if you all see new
> test failures.
I have run the testsuite and the share_testsuite with CLISP 2.44 on a
Linux system. I had no problems with the testsuite.
I think the following failures in the share_testsuite are new:
********************** Problem 109 ***************
Input:
hypergeometric([1, 1], [2], 9.98b-1)
Result:
6.2270622228679275974b0
This differed from the expected result:
6.2270622228679275978b0
********************** Problem 111 ***************
Input:
hypergeometric([1, 1], [2], 9.99b-1)
Result:
6.9146699489310681197b0
This differed from the expected result:
6.9146699489310681202b0
********************** Problem 113 ***************
Input:
block([fpprec : 18], hypergeometric([1, 1], [2], 1.3b0))
Result:
9.26132926404566148b-1 - 2.41660973353061018b0 %i
This differed from the expected result:
9.26132926404566185b-1 - 2.4166097335306102b0 %i
Dieter Kaiser