share testsuite



Am Montag, den 27.07.2009, 10:58 -0500 schrieb Barton Willis:
> I moved (defmvar $multiple_value_return nil) to before the first
> usage of multiple_value_return. That seems to fix the problem for
> me...So try it now.
> 
> There were a few parts of the hypergeometric code that were
> easier to write in Maxima than CL, but I needed (wanted) a
> Common Lisp-like scheme for returning multiple values. So:
> 
>  (%i1) multiple_value_return : true$
>  (%o1) true
> 
>  (%i2) multiple_values(a,b,c);
>  (%o2) multiple_values(a, b, c)
> 
>  (%i3) multiple_value_return : false$
> 
>  (%i4) multiple_values(a,b,c);
>  (%o4) a
> 
> Now that I'm unlikely to change this (there is some grueling code
> surrounding all this), you all can tell me if it my scheme was a
> big mistake :)

I have committed a change to troper.lisp. Now the exp function compiles
correctly again (a silly error).

These are the failures I get with CLISP on a Linux.

Error summary:
Errors found
in /usr/local/share/maxima/5.18post/share/contrib/rtest_lsquares.mac,
problems:
(20 21 22 37 38 39 41 42)
Error found
in /usr/local/share/maxima/5.18post/share/contrib/stringproc/rtestprintf.mac, problem:
(27)
Error found
in /usr/local/share/maxima/5.18post/share/orthopoly/rtest_pochhammer.mac, problem:
(29)
Error found
in /usr/local/share/maxima/5.18post/share/contrib/rtest_to_poly_solver.mac, problem:
(288)
Error found
in /usr/local/share/maxima/5.18post/share/hypergeometric/rtest_hg.mac,
problem:
(159)
Error found
in /usr/local/share/maxima/5.18post/share/hypergeometric/rtest_nfloat.mac, problem:
(22)

I think the failures in rtest_to_poly_solver and rtest_hg are due to my
slow computer. It is very old and has not a lot of memory. 

In rtest_nfloat I get the following failure:

Running tests in rtest_nfloat: 
********************** Problem 22 ***************
Input:
(declare(x, complex), nfloat(conjugate(x), [x = %i + 1.3], 10))


Result:
1.3 - %i

This differed from the expected result:
1.3 - 1.0 %i

37/38 tests passed.

I will have a look at the failures in rtest_lsquares.mac. It is only a
problem of the accuracy of the expected float results.

Dieter Kaiser