[Sbcl-devel] Re: First PCL patch batch

[William H. N. <wil...@ai...>]

On Mon, Oct 21, 2002 at 12:39:41PM +0100, Christophe Rhodes wrote:
> The attached are some fairly trivial PCL fixups:
> 
> * make some DEFMETHOD things signal PROGRAM-ERROR rather than ERROR (not
>   required by ANSI, but probably sensible);
> * delete an "optimization" that appeared to have no effect at all;
> * add a comment on COMPUTE-EFFECTIVE-METHOD.
> 
> This would leave three commits-to-cmucl outstanding, as far as I can
> see:
> 
> * change *LOAD-TRUENAME* to *LOAD-PATHNAME* (this would have the effect
>   that it might be possible to engineer a build with logical pathnames
>   or somesuch, so that users can put the sbcl sources in a place
>   findable by the system, though it wouldn't yet fix the problem that
>   they're unreadable by the running lisp);
> * patch DEFGENERIC so that it doesn't allow forbidden declarations;
> * fix the :ARGUMENTS argument to DEFINE-METHOD-COMBINATION.
> 
> [ I /think/ we have already committed the optimization related to the
>   "structure slots are never unbound"; if not, that would be a fourth
>   one outstanding ]

ported by NJF in 0.7.8.14
 
> The :ARGUMENTS fix is I think out of scope for 0.7.9, in that it's still
> in flux for cmucl (Gerd's got a new version which seems significantly
> better than his old version, which in turn is better than the
> non-functional version :); which of these others am I allowed to commit?

My preference would be not to change code at this point unless we're
fixing bugs, and even then to be conservative. (Except comments should
be OK, and, since the subject is PCL maintenance after all, deleting
clearly-unused code should also be OK.) That standard would probably
accommodate the ERROR-to-PROGRAM-ERROR changes. Arguably it would
allow the DEFGENERIC fix -- use your judgment as to how likely it is
to help people by catching programming errors rather than hurt people
by introducing bugs. I don't remember the :ARGUMENTS problem, so I'm
not sure about that.

-- 
William Harold Newman <wil...@ai...>
overlooking factors of two in offsets and factors of three in
performance since 1982
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C  B9 25 FB EE E0 C3 E5 7C