Christophe Rhodes <csr21@cam.ac.uk> writes:
> I'm getting some errors from the test suite, which if I understand
> correctly may be disagreements about what TYPE-OF needs to do? If
> it's not this issue, then something else might be going wrong...
>
> For example (from rtest1.mac):
>
> /* ********************** Problem 22. *************** */
> %Input is
> P (1 + Y)
> 2
>
>
> The result is
> 2
> d 4 2
> --------- ((Y + 1) - 2 (Y + 1) + 1)
> 2
> d(Y + 1)
> -------------------------------------
> 8
>
> This differed from the expected result:
> 2
> 3 (1 + Y) - 1
> --------------
> 2
I should think so, too. Let me recall the recent threads (of which
you seem to be aware) `CVS CMUCL vs maxima' starting with
http://www.math.utexas.edu/pipermail/maxima/2003/005138.html
and `maxima breaks with CVS clisp' starting with
http://www.math.utexas.edu/pipermail/maxima/2003/005141.html
(I haven't contributed anything to this issue, just relating things.)
Like cmucl and clisp, sbcl changed (type-of 1). So I suggest to apply
R. Toy's (temporary?) patch. Actually, I think it might be preferable
to commit it to CVS soon, since it shouldn't break builds with older
versions of the supported Lisp systems (I checked SBCL 0.7.12.52) and
it is necessary for the more recent CVS versions of at least three of
them. We could then gradually replace all calls to MACLISP-TYPEP (as
suggested by Ray as being the better solution).
Thanks for pointing out that the most recent CVS sbcl doesn't show
the bug I described.
Wolfgang