We reached a (slight) milestone today with sbcl-0.9.15.37, in that the
mop-feature-tests module of the Closer project now reports[*]
| This MOP supports all known standard features.
| This MOP lacks no known standard features.
for SBCL. Now, lest I be accused of any kind of misinformation, I'm
absolutely not claiming that this means that SBCL's MOP is bug free
(it isn't), nor that all of the MOP is the ideal protocol (it probably
isn't) -- simply that I've run out of test cases. So if there are any
readers who have ideas for exciting or useful MOP code -- maybe you
have a flavors layer and a useful library, or a cunning use for
make-method-lambda, or something of that form -- it would be very good
to hear whether or not you run across problems, and if you do, whose
fault the problems are. (I believe that Espen has run into problems
with clg's use of the MOP even with sbcl-0.9.15, for instance; I don't
know exactly what those problems were, nor whose fault they were, but
this is already proof that it's not necessarily plain sailing just
because mop-feature-tests says that it can't find anything missing).
So, anyway, your feedback (with as much detail as you can provide, if
you encounter problems) would be appreciated.
[*] as well as, of perhaps lesser interest,
| This MOP supports the extra features
| This MOP metaclass structure has the leak
This "leak" is likely to stay for the long term; part of the
standard-class/funcallable-standard-class consistency rules has to do
with having the FUNCTION class in the superclasses of a given class,
otherwise the type system gets confused.