I grabbed yesterdays sources and there are a few problems.
o First, it fails in rtest3a with trying to eval some ill-formed
expression. The changes looked fairly minimal, but I think the
problem is in mstuff.lisp. The definition of mfunction1 changed.
If I change it back to the original, that error doesn't occur
anymore.
o The following patch is required. Otherwise problems 2 and 5 of
rtest1a returns 0 where FALSE is expected.
--- maxima-clean/maxima-5.5/src/ar.lisp Mon May 8 02:09:41 2000
+++ maxima-devel/maxima-5.5/src/ar.lisp Fri Apr 20 09:19:29 2001
@@ -60,7 +60,7 @@
($FIXNUM . FIXNUM)))))
(COND ((NOT LTYPE)
(COND ((EQ TYPE '$ANY)
- #+cl (make-array diml)
+ #+cl (make-array diml :initial-element nil)
#-cl
(MAKE-MGENARRAY #+cl :type #-cl type '$ANY
#+cl :CONTENT #-cl CONTENT (APPLY '*ARRAY NIL T DIML)))
o A typo in compile-cmulisp: (setq *init-run*) should probably be
(setq *init-run* t).
o In generr.lisp, the errset macro used for clisp can also be used for
CMUCL. If not, then errcatch doesn't catch errors as desired.
(Note: If you change generr, the files that use errset don't get
recompiled by make. I think they should.)
o Finally, a more serious problem. Problems 6 and 23 in rtest6a
are wrong. For problem 23, the expected answer is abs(y). However,
I get just plain y. Similarly in problem 23, the abs is missing.
For some reason radexpand is ALL instead of TRUE. (Note: the
documentation for SQRT is wrong. It says if RADEXPAND is TRUE,
sqrt(x^2) is x. This only happens if RADEXPAND is ALL.)
With these minor changes, all of the tests pass on CMUCL, except for
the last noted problem. These go away too if RADEXPAND is set to
TRUE.
Ray