Stuart Henderson <s...@spacehopper.org> writes:

> On 2009/01/14 12:06, Simon Kuhnle wrote:
>> On Sat, Jan 10, 2009 at 11:07:53PM +0100, Jonathan Schleifer wrote:
>> > Am 10.01.2009 um 22:33 schrieb Stuart Henderson:
>> >
>> >> On 2009/01/10 15:35, Jonathan Schleifer wrote:
>> >>> I did several patches, one increasing the timeouts to sane values,
>> >>
>> >> That sounds like a reasonable thing to put in the port.
>> >
>> > http://trac.gajim.org/changeset/10848
>> >
>> > It was reverted because some (for example Asterix) complained that their 
>> > router closes their connection if there's no traffic for 55 sec. IMO, 
>> > there should be a space every 55 sec then and then the increased values 
>> > for the XMPP ping timeout can be used there as well.
>> 
>> Nice to see that the gajim port was updated,
>> but what about the patches Jonathan mentioned?
>
> I was trying to avoid doing things that way, and thought about just
> disabling keepalives, but the NAT timeout on crappy routers is also a
> valid concern... but having looked at the code I think we'll have to
> do something like that as a workaround for now.
>
> but I note the commit log:
>
>           revert [10848] That's not the way to solve the problem you're
>           trying to fix: first keepalive packet is sent too early.
>
> pea, you mentioned taking maintainer, if you want to commit that
> patch to increase the timeouts when you do that, it's ok with me,
> but please reference ticket 4643 and svn commits 10848/10851 in
> comments in the patch file.
>
>

I tested gajim with this patch (http://trac.gajim.org/changeset/10848)
and it's better but i still have random deconnexions with my jabber
servers  (Ichat server and 2 ejabberd )(less deconnexions but there are
still here). The only thing to do is to disable the keep-alive option
and everything runs fine. 
Maybe we should add a message to warn users about this issue.
The previous commit missed to add the py-openssl in RUN-DEPENDS. Without
it, i get a warning about an insecure connexion.
Just for the record: regressions tests are broken 
(http://trac.gajim.org/ticket/4613).
More testing is welcome.

Index: Makefile
===================================================================
RCS file: /cvs/ports/net/gajim/Makefile,v
retrieving revision 1.26
diff -u -p -r1.26 Makefile
--- Makefile	13 Jan 2009 11:40:20 -0000	1.26
+++ Makefile	15 Jan 2009 09:33:17 -0000
@@ -4,12 +4,15 @@ SHARED_ONLY=	Yes
 
 COMMENT=	jabber client written in pygtk
 DISTNAME=	gajim-0.12.1
+PKGNAME=	${DISTNAME}p0
 CATEGORIES=	net x11
 
 MASTER_SITES=	${HOMEPAGE}/downloads/
 
 HOMEPAGE=	http://www.gajim.org/
 
+MAINTAINER=	Pierre-Emmanuel Andre <p...@openbsd.org>
+
 # GPLv3 only
 PERMIT_DISTFILES_CDROM=	Yes
 PERMIT_DISTFILES_FTP=	Yes
@@ -37,7 +40,8 @@ RUN_DEPENDS=	::databases/py-sqlite2 \
 		::devel/desktop-file-utils \
 		::net/py-dns \
 		:dbus-python-*:x11/dbus-python \
-		::x11/py-gtk2
+		::x11/py-gtk2  \
+		::security/py-openssl 
 WANTLIB=	X11 Xcomposite Xdamage Xext Xss Xrender fontconfig freetype \
 		m z atk-1.0.>=1011.0 cairo gio-2.0 gdk-x11-2.0.>=600.0 pcre \
 		gdk_pixbuf-2.0.>=600.0 glib-2.0.>=600.0 gmodule-2.0.>=600.0 \
Index: patches/patch-src_common_config_py
===================================================================
RCS file: patches/patch-src_common_config_py
diff -N patches/patch-src_common_config_py
--- /dev/null	1 Jan 1970 00:00:00 -0000
+++ patches/patch-src_common_config_py	15 Jan 2009 09:33:17 -0000
@@ -0,0 +1,14 @@
+$OpenBSD$
+--- src/common/config.py.orig	Wed Jan 14 21:51:53 2009
++++ src/common/config.py	Wed Jan 14 21:53:05 2009
+@@ -308,8 +308,8 @@ class Config:
+ 			'attached_gpg_keys': [ opt_str, '' ],
+ 			'keep_alives_enabled': [ opt_bool, True],
+ 			# send keepalive every N seconds of inactivity
+-			'keep_alive_every_foo_secs': [ opt_int, 55 ],
+-			'time_for_keep_alive_answer': [ opt_int, 30, _('How many seconds to wait for the answer of keepalive packet before we try to reconnect.') ],
++			'keep_alive_every_foo_secs': [ opt_int, 120 ],
++			'time_for_keep_alive_answer': [ opt_int, 60, _('How many seconds to wait for the answer of keepalive packet before we try to reconnect.') ],
+ 			# try for 2 minutes before giving up (aka. timeout after those seconds)
+ 			'try_connecting_for_foo_secs': [ opt_int, 60 ],
+ 			'http_auth': [opt_str, 'ask'], # yes, no, ask

-- 
Pierre-Emmanuel André <pea at raveland.org>
GPG key: 0x7AE329DC

Reply via email to