As announced in
http://www.math.utexas.edu/pipermail/maxima/2004/007260.html
I have removed some package-related kludges. In particular, no LISP
resp. USER nicknames are created anymore for the corresponding
standard packages CL resp. CL-USER and all standard symbols come from
the CL package. The latter point is of some importance for GCL since,
as Vadim pointed out, there used to be a certain confusion between
different symbols of the same name CL:FOO vs. LISP:FOO. I hope that
the present maxima-package.lisp gives a complete picture of
compile-time symbol issues (except for packages defined under
src/numerical).
Further, all lisps now use the condition-system based ERRSET in
generr (we can remove src/kclmac.lisp).
I tested this with recent CVS versions of CMUCL (19a+ 2004-09-09) and
CLISP, with GCL 2.6.5 and with the current CVS SBCL 0.8.15.14 (with
package-locking).
All lisps pass the testsuite. Also, xmaxima seems to work for all of
them (in the sense that the examples and plots in the xmaxima tutorial
seem to work as expected, except that the last line `describe(log);
all' works only for GCL and CLISP; for CMUCL and SBCL, you have to
type in the `all' at the prompt). I also checked the tests in
share/contrib/diffequations/tests. All lisps behave in the same way,
viz. all tests are passed except test 143 in rtest_ode1_riccati.mac,
where the result misses some Bessel function terms.
I still have to check the various lisp files in share, so please don't
be surprised if the tensor package or other out-of-core lisp packages
are slightly broken (because of references to the LISP or USER
packages) for the next two days or so. In any case, ready-to-run
testsuites for such packages, in the style of our standard testsuite,
would be welcome :-)
Wolfgang