novice problems (was: KCL on a SPARCstation 1)

[mic@theory.LCS.MIT.EDU (Michelangelo Grigni)]

I haven't had all the luck of Greg P in making akcl on a SPARC.  I'm also
running SunOS 4.0.3; I ftped kcl.tar.Z and akcl-1-243.tar.Z from cli.com
(he had version 222 from rascal).  The only changes I made to the sun4.defs
were to reset MAINDIR (of course) and to extend INITFORM to load the
on-line documentation file setdoc.lsp.  akcl makes and starts fine, but
there are some immediate problems with the online documentation:

vvvvvvvvvvvvvvvvvvvvvv
AKCL (Austin Kyoto Common Lisp)  Version(1.243) Sun Oct 22 17:35:15 EDT 1989
Contains Enhancements by W. Schelter

Changes in version 1-225 definitely require recompilation of user files.
[To suppress this message remove /a/kcl/akcl/unixport/message]
>(help 'progn)
-----------------------------------------------------------------------------
PROGN                                                          [Special form]

Syntax: (progn {form}*)
Evaluates FORMs in order, and returns whatever the last FORM returns.
-----------------------------------------------------------------------------

>(help* "progn")

Error: 251 is an illgal index to a package hashtable.
Error signalled by SYSTEM::APROPOS-DOC.

Broken at HELP*.  Type :H for Help.
>>:q

Top level.
>(si:package-size (find-package 'lisp))
509
251

>
^^^^^^^^^^^^^^^^^^^^^^

Function help* (autoload.lsp) calls function si::apropos-doc (describe.lsp)
which dies somewhere in macro do-symbols (packlib.lsp).  do-symbols loops
over all symbols in the LISP package hash tables (first the table of
internals and then the table of externals).  Apparently the macro is
looking too far in the internal-symbol hash table.

The immediate problem can be fixed by recompiling apropos-doc, briefly:

 (in-package "system")
 (defun apropos-doc ... [identical to original])
 (help* "progn") ; works fine
 (proclaim '(optimize (safety 2) (space 3)))
 (compile 'si:apropos-doc)
 (help* "progn") ; still works fine

So I have fixed the immediate problem, but I have low confidence in the
rest of the system.  I am considering re-making the whole thing with lower
cc optimization; meanwhile has anyone already encountered this problem and
found a fix or explanation?  Any suggestions appreciated.

-- 
Michelangelo (Mic) Grigni -- Gradual Student, MIT Applied Math