[clpython-devel] Re: CLPython on SBCL

[Jason Nielsen]

On Tue, 6 May 2008, Willem Broekema wrote:

> On Tue, May 6, 2008 at 6:19 AM, Jason Nielsen <jdn at math.carleton.ca> wrote:
>
>> I can confirm that compilation fails using SBCL 1.0.16 on a Ubuntu 7.10
>> system with a:
>>
>> "has no suitable method for STREAM-ELEMENT-TYPE, and so has fallen through
>> to this method...."
>>
>> error that crashes the whole session.  You can reproduce this behaviour
>> with the following:
>>
>> (defclass foo (standard-class) ())
>> (defclass bar () () (:metaclass foo))
>>
>
> Thanks for the confirmation. Actually I hoped this would not occur in the
> latest version, as Nikodemus Siivola said on 2008-01-15, in reply to what
> seems a very similar issue reported by Pascal Costanza: "This seems to have
> been fixed at some point during the last 10 months": <
> http://article.gmane.org/gmane.lisp.steel-bank.devel/10603>
>
> Custom metaclasses are crucial for CLPython. Hopefully someone feels
> inclined to fix this SBCL issue.
>
> - Willem
>

Interesting, if you run in an sbcl console without slime:

(defclass foo (standard-class) ())
(defclass bar () () (:metaclass foo))

you get a different error message:

debugger invoked on a SIMPLE-ERROR in thread #<THREAD "initial thread" 
{A884551}>:
   The class #<STANDARD-CLASS STANDARD-OBJECT> was specified as a 
super-class of
   the class #<FOO BAR>, but the meta-classes #<STANDARD-CLASS 
STANDARD-CLASS>
   and #<STANDARD-CLASS FOO> are incompatible.  Define a method for
   SB-MOP:VALIDATE-SUPERCLASS to avoid this error.

so thinking this may be some sort of slime/sbcl problem I decided to try 
to asdf-install without slime.  The compilation fails much earlier with:

; registering #<SYSTEM #:LW-COMPAT {A8F48F9}> as LW-COMPAT
WARNING: change in instance length of class SB-PRETTY:PRETTY-STREAM:
   current length: 25
   compile time length: 27

debugger invoked on a SIMPLE-ERROR in thread #<THREAD "initial thread" 
{A884551}>:
   The class SB-PRETTY:PRETTY-STREAM was not changed, and there's no 
guarantee that
the loaded code (which expected another layout) will work.

This error also doesn't mean much to me.

Jason