bug#27016: possible bug in `defsetf'

[Stefan Monnier]

> (unless t
>   (defalias 'foobar 'cons)
>   (defsetf foobar setcar))            ; nil 
> (foobar pair)                         ; same as above
> (setf (foobar pair) 0)                        ; 0       !

I can confirm that it's a bug.  It's perfectly correct for the defsetf
to be macroexpanded, but not to be evaluated.
Not sure how best to get that behavior.  The naive/straightforward way
might be to introduce some sort of eval-and-compile-when-at-toplevel,
but there's probably a better way.  The first thing to do is to look at
how it was done in Emacs-23.


        Stefan