[Ecls-list] Two bugs in ECL's CLOS implementation

[Pascal Costanza]

On 27 Nov 2009, at 22:42, Juan Jose Garcia-Ripoll wrote:

>> Showstopper:
>> [...]
>>> (clos:set-funcallable-instance-function #'test (lambda (a b c) (* a b c)))
>> [...]
>>> (defmethod test (a b c) (+ a b c))
>> Detected access to an invalid or protected memory address.
> 
> CLOS:SET-FUNCALLABLE-INSTANCE-FUNCTION has to change the structure of
> the generic function to store the dispatch function somewhere.
> Unfortunately I forgot to copy a field, instance.sig,  which is used
> to determine when instances become obsolete and when they have been
> initialized.
> 
> Both are fixed in CVS. Thanks a lot for reporting and taking the time
> to write small test cases!

Thanks a lot indeed for fixing those. I now get much farther with Closer to MOP than before.

However, another issue occurred:

> (use-package :clos)

T
> (defgeneric foo (x y z))

#<STANDARD-GENERIC-FUNCTION FOO>
> (set-funcallable-instance-function #'foo (lambda (x y z) (+ x y z)))

#<STANDARD-GENERIC-FUNCTION FOO>
> (defun test ()
    (apply #'foo (list 1 2 3)))

TEST
> (test)

#<bytecompiled-closure #<bytecompiled-function 0000000101e00000>> is not of type FUNCTION.

Available restarts:

1. (RESTART-TOPLEVEL) Go back to Top-Level REPL.

Broken at TEST. In: #<process SI:TOP-LEVEL 00000001005c0f60>.
>> :r1

> (defun test ()
    (funcall #'foo 1 2 3))

TEST
> (test)

6


Apparently, ECL has problems with APPLYing generic functions that have been modified by way of set-funcallable-instance-function. This seems to be independent of whether code is compiled or byte-compliled.


Pascal

-- 
Pascal Costanza, mailto:pc at p-cos.net, http://p-cos.net
Vrije Universiteit Brussel
Software Languages Lab
Pleinlaan 2, B-1050 Brussel, Belgium