From mboxrd@z Thu Jan  1 00:00:00 1970
From: ludo@gnu.org (Ludovic =?UTF-8?Q?Court=C3=A8s?=)
Subject: bug#26976: On Hydra,
	offload crashes while trying to build linux-libre source
Date: Tue, 13 Jun 2017 23:32:06 +0200
Message-ID: <87shj3tvzd.fsf@gnu.org>
References: <87h90japz0.fsf@netris.org> <87h90h966f.fsf@gnu.org>
	<87pof5dnja.fsf@gnu.org> <87mva88oe9.fsf@gnu.org>
	<87k24q86hk.fsf@gnu.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Return-path: <bug-guix-bounces+gcggb-bug-guix=m.gmane.org@gnu.org>
Received: from eggs.gnu.org ([2001:4830:134:3::10]:55658)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <Debian-debbugs@debbugs.gnu.org>) id 1dKtQv-0001tc-58
	for bug-guix@gnu.org; Tue, 13 Jun 2017 17:33:06 -0400
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <Debian-debbugs@debbugs.gnu.org>) id 1dKtQs-0007OB-2x
	for bug-guix@gnu.org; Tue, 13 Jun 2017 17:33:05 -0400
Received: from debbugs.gnu.org ([208.118.235.43]:43627)
	by eggs.gnu.org with esmtps (TLS1.0:RSA_AES_128_CBC_SHA1:16)
	(Exim 4.71) (envelope-from <Debian-debbugs@debbugs.gnu.org>)
	id 1dKtQs-0007O4-0F
	for bug-guix@gnu.org; Tue, 13 Jun 2017 17:33:02 -0400
Sender: "Debbugs-submit" <debbugs-submit-bounces@debbugs.gnu.org>
Resent-Message-ID: <handler.26976.B26976.14973895434279@debbugs.gnu.org>
In-Reply-To: <87k24q86hk.fsf@gnu.org> ("Ludovic
	\=\?utf-8\?Q\?Court\=C3\=A8s\=22'\?\=
	\=\?utf-8\?Q\?s\?\= message of "Mon, 05 Jun 2017 23:33:27 +0200")
List-Id: Bug reports for GNU Guix <bug-guix.gnu.org>
List-Unsubscribe: <https://lists.gnu.org/mailman/options/bug-guix>,
	<mailto:bug-guix-request@gnu.org?subject=unsubscribe>
List-Archive: <http://lists.gnu.org/archive/html/bug-guix/>
List-Post: <mailto:bug-guix@gnu.org>
List-Help: <mailto:bug-guix-request@gnu.org?subject=help>
List-Subscribe: <https://lists.gnu.org/mailman/listinfo/bug-guix>,
	<mailto:bug-guix-request@gnu.org?subject=subscribe>
Errors-To: bug-guix-bounces+gcggb-bug-guix=m.gmane.org@gnu.org
Sender: "bug-Guix" <bug-guix-bounces+gcggb-bug-guix=m.gmane.org@gnu.org>
To: Mark H Weaver <mhw@netris.org>
Cc: Artyom Poptsov <poptsov.artyom@gmail.com>, 26976@debbugs.gnu.org

Hello,

ludo@gnu.org (Ludovic Court=C3=A8s) skribis:

> AIUI, that means there=E2=80=99s one output compression buffer per sessio=
n, and
> it=E2=80=99s not thread-safe (in Guile 2.2 finalizers are called from a s=
eparate
> thread.)
>
> I think the fix, in Guile-SSH, is to associate each libssh object
> (session, channel, etc.) with a mutex, and to protect all uses of the
> libssh object by that mutex.

I=E2=80=99ve pushed a workaround that seems to work (=E2=80=98guix copy=E2=
=80=99 commands that
previously segfaulted on hydra.gnu.org no longer do):

  https://git.savannah.gnu.org/cgit/guix.git/commit/?id=3D8e469b67f95cfe5b9=
5405b503b8ee315fdf8ce66

The idea is simply to explicitly close all the channel ports.  That way,
when their finalizer gets called, it doesn=E2=80=99t invoke libssh code.

(guix build offload) and (guix ssh) already explicitly close all the
channel ports they open, so I found that the culprit is =E2=80=98node-eval=
=E2=80=99 in
Guile-SSH.  The patch changes =E2=80=98node-eval=E2=80=99 to explicitly clo=
se the RREPL
channel upon completion.  (BTW, performance-wise, it may be best to
avoid opening a new channel every time =E2=80=98node-eval=E2=80=99 is calle=
d.)

That may be good enough for Guix, but of course that=E2=80=99s a workaround=
 and
not a proper fix.

I=E2=80=99ll do some more testing and then maybe try switching hydra.gnu.or=
g to
Guile 2.2 again.

Thanks,
Ludo=E2=80=99.