From edi at agharta.de  Thu Oct 11 07:03:17 2007
From: edi at agharta.de (Edi Weitz)
Date: Thu, 11 Oct 2007 09:03:17 +0200
Subject: [flexi-streams-devel] New release 0.13.1
Message-ID: <uzlyqgmkq.fsf@agharta.de>

ChangeLog:

  Version 0.13.1
  2007-10-11
  Small fix for AllegroCL's "modern" mode

Download:

  http://weitz.de/files/flexi-streams.tar.gz


From jcornez at alum.mit.edu  Thu Oct 25 16:29:53 2007
From: jcornez at alum.mit.edu (Jason S. Cornez)
Date: Thu, 25 Oct 2007 18:29:53 +0200
Subject: [flexi-streams-devel] [Fwd: Problem with TLS connection on
	Franz/ACL8.1]
Message-ID: <4720C481.9050707@alum.mit.edu>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

This is a report I also sent to cl-xmpp-devel as that is where I
discovered it.  It does however seem to be more related to Flexi-Streams
on ACL8.1.

Thanks in advance for any ideas you might have about how to resolve this
issue.

Best,
- -Jason

- -------- Original Message --------
Subject: [cl-xmpp-devel] Problem with TLS connection on Franz/ACL8.1
Date: Thu, 25 Oct 2007 18:16:29 +0200
From: Jason S. Cornez <jcornez at alum.mit.edu>
To: cl-xmpp-devel at common-lisp.net

Hi I know this is a small list and that this problem is likely not to do
specifically with CL-XMPP code, but with one of the libraries it uses.
So if you can help by even suggesting where better I might ask this, it
would be appreciated.  Thank you.

What follows is run using the latest cl-xmpp and cxml/closure-commons
from CVS along with other libraries from asdf-install tarballs.  I use
ACL8.1 on linux-x86.32.  The net result is that the lisp crashes and
dies with a segmentation violation.  Obviously that in itself is a Franz
issue, but this happens because (figure-encoding...) calls (read-byte
...) on a flexi-stream.  And I figure this doesn't make much sense.

I've traced figure-encoding to see that this is clear.  Perhaps there
should be a specialized method of figure-encoding for flexi-streams?

Again, if you can help me to direct this query to a better place, please
let me know.

Best,
- -Jason


CL-USER(6): (trace runes::figure-encoding)
(RUNES::FIGURE-ENCODING)
CL-USER(7): (xmpp:connect-tls :hostname "localhost")
<?xml version='1.0' ?><stream:stream to='localhost'
xmlns='jabber:client' xmlns:stream='http://etherx.jabber.org/streams'
version='1.0'>
 0[2]: (RUNES::FIGURE-ENCODING #<MULTIVALENT stream socket connected
from localhost/56698 to localhost/5222 @ #x72766952>)
 0* #<STANDARD-METHOD RUNES::FIGURE-ENCODING (STREAM)>
 0[2]: returned :UTF-8 (60 115)
<starttls xmlns='urn:ietf:params:xml:ns:xmpp-tls'/>
<?xml version='1.0' ?><stream:stream to='localhost'
xmlns='jabber:client' xmlns:stream='http://etherx.jabber.org/streams'
version='1.0'>
 0[2]: (RUNES::FIGURE-ENCODING #<FLEX::FLEXI-LATIN-1-IO-STREAM @
#x722197ca>)
 0* #<STANDARD-METHOD RUNES::FIGURE-ENCODING (STREAM)>
_______________________________________________
cl-xmpp-devel mailing list
cl-xmpp-devel at common-lisp.net
http://common-lisp.net/cgi-bin/mailman/listinfo/cl-xmpp-devel
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.7 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHIMSBQlm6HDTMLyMRAulqAJ9k3pDbNKbvwFYc/6ZAuywkze2IyACgmr2t
qgpe65voZq71z9OqocyY4eg=
=e96r
-----END PGP SIGNATURE-----


From edi at agharta.de  Thu Oct 25 16:36:12 2007
From: edi at agharta.de (Edi Weitz)
Date: Thu, 25 Oct 2007 18:36:12 +0200
Subject: [flexi-streams-devel] [Fwd: Problem with TLS connection on
	Franz/ACL8.1]
In-Reply-To: <4720C481.9050707@alum.mit.edu> (Jason S. Cornez's message of
	"Thu, 25 Oct 2007 18:29:53 +0200")
References: <4720C481.9050707@alum.mit.edu>
Message-ID: <ulk9rtamb.fsf@agharta.de>

On Thu, 25 Oct 2007 18:29:53 +0200, "Jason S. Cornez" <jcornez at alum.mit.edu> wrote:

> This is a report I also sent to cl-xmpp-devel as that is where I
> discovered it.  It does however seem to be more related to
> Flexi-Streams on ACL8.1.
>
> Thanks in advance for any ideas you might have about how to resolve
> this issue.

It would be helpful if you could provide a self-contained test case
which I can use to reproduce the problem - preferably without the need
to load other libraries.

Cheers,
Edi.


From jcornez at alum.mit.edu  Thu Oct 25 16:41:48 2007
From: jcornez at alum.mit.edu (Jason S. Cornez)
Date: Thu, 25 Oct 2007 18:41:48 +0200
Subject: [flexi-streams-devel] [Fwd: Problem with TLS connection
	on	Franz/ACL8.1]
In-Reply-To: <ulk9rtamb.fsf@agharta.de>
References: <4720C481.9050707@alum.mit.edu> <ulk9rtamb.fsf@agharta.de>
Message-ID: <4720C74C.6010507@alum.mit.edu>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Yes, perfectly understood about the test case.  At this point I was just
trying to see if this problem sounded familiar to anyone.  If I can
generate an isolated test case, I certainly will.

In the meantime, David Lichteblau wrote me the following that may help:
- ----
CL+SSL is known to be broken on Allegro with a failure mode like this.
I can only guess why, but since Allegro already includes SSL support,
perhaps there is some kind of conflict between the two FFI uses.

I think the workaround Edi used in his libraries was to do
  #+allegro (excl:make-ssl-client-stream ...)
  #-allegro (cl+ssl:make-ssl-client-stream ...)
but if the problem stays unsolved, it would be better to move this kind
of hack into CL+SSL itself, so that it wraps the built-in streams on
Allegro instead of doing things itself.

Unfortunately I don't really have the time for CL+SSL development at the
moment, but patches are always welcome, of course.
- ----

I'll see what I can do about producing some code that gets past this.

- -Jason

Edi Weitz wrote:
> It would be helpful if you could provide a self-contained test case
> which I can use to reproduce the problem - preferably without the need
> to load other libraries.
> 
> Cheers,
> Edi.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.7 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHIMdMQlm6HDTMLyMRAkCkAKCLPo19P8mnXTVp1eIxzj8ceoYUsgCgqDbi
rgCzNA35fN3V6OXnQ9/14Os=
=aes5
-----END PGP SIGNATURE-----


From jcornez at alum.mit.edu  Thu Oct 25 17:23:25 2007
From: jcornez at alum.mit.edu (Jason S. Cornez)
Date: Thu, 25 Oct 2007 19:23:25 +0200
Subject: [flexi-streams-devel] Re: [cl-xmpp-devel] Problem with TLS
	connection on Franz/ACL8.1
In-Reply-To: <20071025163158.GA31888@radon>
References: <4720C15D.6010902@alum.mit.edu> <20071025163158.GA31888@radon>
Message-ID: <4720D10D.2000600@alum.mit.edu>

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Many thanks.  Here is a patch which works for me to get past this
particular problem.

===================
- --- streams.lisp        2007-07-07 18:47:57.000000000 +0200
+++ /home/jcornez/Library/world/site/cl+ssl-2007-07-07/streams.lisp
2007-10-25 19:05:27.000000000 +0200
@@ -156,6 +156,25 @@

 ;;; interface functions
 ;;;
+#+allegro
+(defun make-ssl-client-stream
+    (socket &key certificate key (method 'ssl-v23-method) external-format
+                 close-callback)
+  (declare (ignore external-format close-callback))
+  (let ((acl-method (case method
+                      ((ssl-v23-method
+                        ssl-v3-method
+                        ssl-v2-client-method
+                        ssl-v23-client-method
+                        ssl-v3-client-method)    :sslv23)
+                      ((ssl-TLSv1-method
+                        ssl-TLSv1-client-method) :tlsv1))))
+    (socket:make-ssl-client-stream socket
+                                   :certificate certificate
+                                   :key         key
+                                   :method      acl-method)))
+
+#-allegro
 (defun make-ssl-client-stream
     (socket &key certificate key (method 'ssl-v23-method) external-format
                  close-callback)
===================

- -Jason

David Lichteblau wrote:
> CL+SSL is known to be broken on Allegro with a failure mode like this.
> I can only guess why, but since Allegro already includes SSL support,
> perhaps there is some kind of conflict between the two FFI uses.
> 
> I think the workaround Edi used in his libraries was to do
>   #+allegro (excl:make-ssl-client-stream ...)
>   #-allegro (cl+ssl:make-ssl-client-stream ...)
> but if the problem stays unsolved, it would be better to move this kind
> of hack into CL+SSL itself, so that it wraps the built-in streams on
> Allegro instead of doing things itself.
> 
> Unfortunately I don't really have the time for CL+SSL development at the
> moment, but patches are always welcome, of course.
> 
> 
> d.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.7 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHINENQlm6HDTMLyMRAiSTAJ46LaVUnJDueUxa3E6DtoI+TsKIaQCgvAtq
cGkrHllMGVHCWONoJBg1pFA=
=/0Nx
-----END PGP SIGNATURE-----