[clisp-list] ext:letf with gethash and symbol-value

[Hoehle, Joerg-C. <Joe...@t-...>]

Hi,

several times I've been bitten by EXT:LETF, which generally sounds like =
a useful macro.
It does something like (unwind-protect (setf place value) body (setf =
place original-value))

However, it does not correctly restore the original state when used =
with GETHASH or SYMBOL-VALUE. Both places are set to NIL at exit, =
whereas they might have been unset/unbound before.

[27]> (let ((h(make-hash-table)))(ext:letf (((gethash 'a h) 'b))1)h)
#S(HASH-TABLE :TEST FASTHASH-EQL (A . NIL)) ; expected empty hash table

Furthermore, (ext:letf (((symbol-value 'a) 3)) a) errors out when a is =
not bound prior to entering the form.

A consistent integration of LETF with places would restore the boundp =
state.
Hmm, the same argument would apply to struct slots...
The current behaviour is very close to a bug, wouldn't you think?

Regards,
	J=F6rg H=F6hle.