[Axiom-developer] Axiom on clisp

[Waldek Hebisch]

I can now report succesful build on clisp.  You can find a patch at:

http://www.math.uni.wroc.pl/~hebisch/prog/sbcl-clisp2.diff

Apply on top of sbcl patch.  Like in sbcl case replace
interp-proclaims.lips by safty declaration or empty file.

Clisp handles files and directories in rather strange way, to
make things sane I use cl-fad library from:

http://weitz.de/files/cl-fad.tar.gz

ATM one has to manually edit src/lisp/Makefile.pamphlet and
put there correct path to cl-fad.  

Another problem: in default C locale clisp rejects any non-ASCII
charaters.  Martin's guessing package contains some comments in
German and clisp chokes on them.  So, for succesfull build one
has to set charset to something allowing all 8-bit codes.
For example:

LC_CTYPE=ISO-8859-1 make

should work.

Like in sbcl case sockets are fake.  There is extra issue: some
test fail when printing long numbers.  Backtrace indicate floating
point overflow when computing log10 (needed to find width of a
number).  I am not sure what is the exact reason, but this points
out to another problem: clisp has "arbitrary precision" floats
and MOST-POSITIVE-LONG-FLOAT is really large:

> MOST-POSITIVE-LONG-FLOAT
8.8080652584198167656L646456992

In some places Axiom code seem to assume that long-float is the
same as double-float and allowing numbers comparable to
MOST-POSITIVE-LONG-FLOAT may lead to overflow in double-float
computations.

-- 
                              Waldek Hebisch
address@hidden