problems compiling pcl in Mac Allegro Common Lisp

[Kevin_Crowston@XV.MIT.EDU]

I'm trying to compile "no cute name" pcl in Allegro
Common Lisp 1.2.1 and having very little luck.  I would
be grateful for any advice from someone who has done this 
successfully.  (Someone must have, because the notes
say PCL has been tested in Coral 1.2, which I think is the same).  

I'm running Allegro Lisp on a MacPlus with a Gemini 020
accelerator board with 4M of memory.  The machine also
has an E-Machines Big Picture display.  It seems to
work fine running things like MultiFinder, Microsoft
Word, MacDraw, etc., so I don't think it's the hardware that's
flakey.  I haven't tried this yet on a stock Mac II,
but I guess that's an obvious next step.  If it works there for
other people, then I'm curious what it is about the accelerator
board that doesn't work.

I'm using System Version 6.0.2, with either Finder or
Multifinder.  Under MultiFinder, I give Allegro 2048K;
increasing this to 3000K didn't seem to help.  

I tried simply following the instructions in get-pcl.text,
i.e., load "defsys.lisp" and execute (pcl::compile-pcl).
This seems to work fine, except that sometimes the machine 
will simply crash while trying to load a compiled file.  It 
does this both under Finder and MultiFinder.  Usually a 
bomb box will appear, with a variety of IDs, typically 2, 8 
or 10 (I haven't noticed any patterns).  Sometimes the machine 
just freezes and has to be reset and rebooted.  

What's most annoying is that the behaviour is not repeatable.
Sometimes doing the exact same things two times in a row
will have different effects.  The files that seem to be the
problem most often include defs.fasl and macros.fasl, but
other files sometimes cause problems.  The furthest I've gotten is
to dcode-pre2.  If the process gets this far, loading this file will always
crash the machine. 

In looking at macros.lisp, I noticed that the file is supposed to
be loaded before it's compiled.  I tried doing that much by hand,
but the new compiled file didn't seem to make much difference
in the long run (ie., still crashes at dcode-pre2).  

Any advice will be appreciated.  

Kevin