Hi Eitaro,

On Sun, Jan 11, 2015 at 6:30 PM, Eitaro Fukamachi <e.arrows@gmail.com> wrote:

This is a reproducible script.

----
(defpackage unexport-test
(:export :a))

(eval-when (:compile-toplevel :load-toplevel :execute)
(when (find-package :unexport-test)
(do-symbols (symbol :unexport-test)
(unexport symbol :unexport-test))))

(defpackage unexport-test
(:export :a))

(prin1 (nth-value 1 (intern (string :a) :unexport-test)))

If cl-colors depends on this behaviour, it's depending on behaviour the cl spec explicitly named "undefined". From the spec:

If the new definition is at variance with the current state of that package, the consequences are undefined; an implementation might choose to modify the existing package to reflect the new definition. (see http://www.ai.mit.edu/projects/iiip/doc/CommonLISP/HyperSpec/Body/mac_defpackage.html, third sentence in the Description section)

the important bit is "an implementation might choose to modify"; ABCL currently is an implementation that doesn't modify. Essentially this means you can - in ABCL's current implementation - just leave out the second DEFPACKAGE form...

Of course, if you say this is pretty rare when compared to other implementations, I think we should seriously consider accepting a patch to change the behaviour. Do you know what the *exact* behaviours of SBCL and CCL are? Maybe it's not too hard to implement one of the two. Patch submission always very much welcome as well, of course!

Regards,

Bye,

Erik.

http://efficito.com -- Hosted accounting and ERP.

Robust and Flexible. No vendor lock-in.