From mboxrd@z Thu Jan  1 00:00:00 1970
Received: (from majordomo@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id EAA07066; Wed, 10 Jul 2002 04:21:17 +0200 (MET DST)
X-Authentication-Warning: pauillac.inria.fr: majordomo set sender to owner-caml-list@pauillac.inria.fr using -f
Received: from nez-perce.inria.fr (nez-perce.inria.fr [192.93.2.78]) by pauillac.inria.fr (8.7.6/8.7.3) with ESMTP id EAA07136 for <caml-list@pauillac.inria.fr>; Wed, 10 Jul 2002 04:21:16 +0200 (MET DST)
Received: from gato.phys.clemson.edu (gato.phys.clemson.edu [130.127.188.133])
	by nez-perce.inria.fr (8.11.1/8.11.1) with ESMTP id g6A2LEH26836;
	Wed, 10 Jul 2002 04:21:15 +0200 (MET DST)
Received: from adsl-78-167-177.gsp.bellsouth.net ([216.78.167.177] helo=elefante.local)
	by gato.phys.clemson.edu with asmtp (Exim 3.35 #1 (Debian))
	id 17S76N-0006iB-00; Tue, 09 Jul 2002 22:21:11 -0400
Received: from oso.local ([192.168.1.3])
	by elefante.local with esmtp (Exim 3.35 #1 (Debian))
	id 17S76o-0001m1-00; Tue, 09 Jul 2002 22:21:38 -0400
Received: from fernando by oso.local with local (Exim 3.35 #1 (Debian))
	id 17S76u-0005Ll-00; Tue, 09 Jul 2002 22:21:44 -0400
Date: Tue, 9 Jul 2002 22:21:44 -0400
To: caml-list@inria.fr
Cc: xavier.leroy@inria.fr
Subject: [Caml-list] more bigarray problems
Message-ID: <20020710022144.GA20544@oso.local>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.3.28i
From: Fernando Alegre <fernando@cc.gatech.edu>
Sender: owner-caml-list@pauillac.inria.fr
Precedence: bulk


After I thought we had tamed bigarrays, it seems we did not. Certain
operations such as slices and reshaping allocate new bigarray structs. They
use the stock bigarray_alloc instead of the custom one, of course, and so
they bypass the finalizers. We managed to make OCaml segfault for this!

I don't see any way to work around that problem other than having support for
external finalizers in bigarray_alloc, and having proxies and slices
remember them. Even with a perfectly working Gc.finalise, we would encounter
the same problem, as the finalizers will get forgotten after a reshape, not
a very nice feature...

I think, given this input, lack of finalization support in bigarrays should be
promoted from just a feature wish to a bug, since there seems to be no way
to work around it.

Thanks,
       Fernando

-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners