[Sbcl-devel] Re: [Sbcl-commits] CVS: sbcl/src/pcl env.lisp,1.9,1.10

[<ger...@t-...>]

Christophe Rhodes <cs...@ca...> writes:

> I have a slight problem with this that can probably be worked around;
> essentially, I'm worried about enforcing declarations-are-assertions.
> But that's mostly orthogonal to allowing SETF SLOT-VALUE itself, I
> guess.  CLISP, too, allows (SETF SLOT-VALUE) on :READ-ONLY T slots.

One reason more to do the same then, I guess, if three implementations
already follow that path.

> I should say that I have a problem with the current implementation of
> MAKE-LOAD-FORM-SAVING-SLOTS, too, in that there's no guarantee that
> (SETF SLOT-VALUE) actually sets the value of a slot, in the presence
> of arbitrary methods (maybe some user has installed an assertion in an
> :around method, which our implementation of MLFSS violates, or
> something).  I don't know how reasonable it would be to implement
> MAKE-LOAD-FORM-SAVING-SLOTS in terms of lower-level operators?
> (e.g. CLOS-SLOTS-SET for object slots, %INSTANCE-SET for structure
> slots?)

CLHS seems to say that the init forms can consist of (SETF
SLOT-VALUE)s, so I guess it's OK if they do.  ACL does that too, it
seems.  But I dunno, IANALL ;).