test suite failures with cvs Maxima



Hello Robert,

First:
We have to renumber about 200 examples. It is really much more simple to add a
test.

If you don't like a dummy test perhaps you would like the following test (I
added this test before the test with the number 42):

/* First a test with a special value for the order of the Laguerre function.
   (Test is added to compensate a killed line to load orthopoly.) */

factor(specint(%e^(-s*t)*%l[10,0](t),t));
(s-1)^10/s^11;

Second: The following changes to rtesthyp.mac will work (GCL 2.6.8) and are
equivalent to the orginal tests:

Problem 13:
hgfred([-4],[1/2],z^2/2);
hermite(8,z/sqrt(2))/1680;

Problem 14:
ratsimp(hgfred([-4],[3/2],z^2/2));
hermite(9,z/sqrt(2))/(15120*sqrt(2)*z)$

Problem 66:
ratsimp(hgfred([-2],[3/2],z));
hermite(5,sqrt(z))/(120*sqrt(z));

Third:
The problem with the Pochhammer symbol has vanished. It simplifies again like
pochhammer(1,n) --> n!.

With the changes above we have no longer problems with the testsuite (GCL 2.6.8
and Windows XP). Should I commit the changes?

Dieter Kaiser

-----Urspr?ngliche Nachricht-----
Von: robert.dodier at gmail.com [mailto:robert.dodier at gmail.com] 
Gesendet: Sonntag, 27. Juli 2008 17:00
An: Dieter Kaiser
Cc: maxima at math.utexas.edu
Betreff: Re: [Maxima] test suite failures with cvs Maxima

On 7/27/08, Dieter Kaiser <drdieterkaiser at web.de> wrote:

>  First: The problems with rtest_hypgeo.test can almost be solved when
indroducing
>  a dummy test at the place where the loading of orthopoly is killed. All tests
>  are well numbered. This scheme and the expected failures get out of order
>  killing the line. Should I commit the dummy test, to reorder the numbering
>  scheme?

Well, I guess that works, but I think it would be better to just update
the numbers in tests/testsuite.lisp.

>  Second: There remains one open failure. Pochhammer(1,i) no longer simplifies
to
>  i!. After loading orthopoly by hand, it works again, the Pochhammer symbol
now
>  simplifies.

It seems we need a way to make simplifying functions autoload.
I guess this should do it: put the autoload declaration for simp-pochhammer
(and simp-unit-step and any others defined in orthopoly.lisp)
in src/max_ext.lisp, and move the defprop's for those from orthopoly
to src/max_ext.lisp.

>    ratsimp(hgfred([-4],[1/2],z^2/2)-hermite(8,z/sqrt(2))/1680);
>
>  hgfred returns an expression in terms of the Hermite function. But now the
>  Hermite function we use to compare will automatically expanded in a series.

Hmm, does the test succeed if it is changed to

hgfred([-4],[1/2],z^2/2);
hermite(8,z/sqrt(2))/1680;

??

best

Robert Dodier