It seems that CCL is more liberal/comprehensive than other Lisps in
its usage of MAKE-INSTANCES-OBSOLETE. I haven't dived further into
this but the following patch fixes the issue specifically for OpenMCL:
diff -rN -u old-elephant-1.0/src/elephant/classes.lisp
new-elephant-1.0/src/elephant/classe
> I'm not sure where VALUE would be defined. CCL 1.3 apparently doesn't
> know it; TRIVIAL-GARBAGE claims that CCL doesn't support weak data
> structures at all (it simulates weak pointers with a hash table).
Ah, I see this has been fixed in latest 1.0; apparently this
is present in OpenMCL but h
Fix proposed:
--- old-elephant-1.0/src/elephant/schema-evolution.lisp 2009-05-03
19:18:29.277598554 +0200
+++ new-elephant-1.0/src/elephant/schema-evolution.lisp 2009-05-03
19:18:29.284267720 +0200
@@ -120,8 +120,7 @@
upgrade-instance. This should be called after a redefinition."
(loop
FWIW here's the backtrace:
(B7AA67FC) : 0 (FUNCALL #'#<#> # NIL NIL NIL)
271
(B7AA682C) : 1 (%%STANDARD-COMBINED-METHOD-DCODE (# #) -303392234) 679
(B7AA686C) : 3 (UPDATE-OBSOLETE-INSTANCE #) 1919
(B7AA6884) : 4 (%%ONE-ARG-DCODE #((#<#> #<#>) NIL NIL 0
# ...) #) 95
(B7AA6898) : 5 (FUNCALL #
Just got this on CCL:
If the schemas mismatch, a derived controller schema should
have been computed
Dataset and code work fine with SBCL, though.
Maybe there's a missing implementation bit for CCL?
Leslie
--
http://www.linkedin.com/in/polzer
_
