>>>>> "Charles" == Charles Earl <cs.uchicago.edu at earl> writes:

 Charles> Just curious about the following result I obtained under ACL4.3 o
n a
 Charles> sparcstation running solaris

 Charles> 	USER(22): (setq a '(b c))
 Charles> 	(B C)
 Charles> 	USER(23): (setq d '(e f))
 Charles> 	(E F)
 Charles> 	USER(24): (cdr a . d)

 Charles> 	(C)

 Charles> It seems to me that it should return an error, which is the case 
under
 Charles> CMUCL

 Charles> 	* (setq a '(b c))
 Warning> Declaring A special.
 
 Charles> 	(B C)
 Charles> 	* (setq d '(e f))
 Warning> Declaring D special.
 
 Charles> 	(E F)
 Charles> 	* a
 
 Charles> 	(B C)
 Charles> 	* (cdr a . d)
 
 Charles> 	Type-error in KERNEL::OBJECT-NOT-LIST-ERROR-HANDLER:  D is not
 of type LIST
	 
 Charles> MCL3.9, which returns
 Charles> 	Error value D is not of expected type LIST

 Charles> ACL3.0.2 returns an error

 Error> Argument D given to RPLACA or RPLACD should have been a cons in
 ACL> SC-DISPATCH

 Charles> And Harlequin FreeLisp gives
 Error> Non-list argument D for endp

 Charles> What's the right thing to do? Does it matter?

Not only your examples show a flaw (IMHO) in Allegro but this flaw
also occurs in many other (simpler) situations:

USER(30): (+ 1 2 . 3)
3
USER(31): (* 2 . 4)
2

I searched the CommonLisp Hyperspec and there's nothing that forces
a form to be a proper list, but there isn't nothing that suggest that
it can be a non-proper list.

António Leitão