On Tue, 21 Aug 2012 23:53:39 +0200
Christoph Anton Mitterer <[email protected]> wrote:

[...]
> I'm running at sid... but still experience all the problems I describe
> e.g. here: https://support.process-one.net/browse/EJAB-1565

Uh, sorry but I'm now somewhat lost now.

First, we don't have a bug in our BTS about this upstream issue, so
definitely uploading of 2.1.10-3 wasn't supposed to fix it (as I did
not even know this issue exists).

Second, the bug is still open, so upstream's 2.1.11 (which we
failed to push into Debian in time before the freeze) wasn't expected to
fix it either.

Third, the s2s issue has (supposedly) nothing to do with clustered
setup: s2s is about different XMPP servers (I mean, controlled by
different authorities) speak (XMPP) to each other to allow their users
communicate.  This (supposedly) has nothing to do with Erlang
clustering which is implemented using Erlang's own native RPC mechanism.

> This includes quite often failing s2s connections.

The original bug report [1] mentions obvious (for
Erlang-literate people) Erlang-level runtime failure to locate a
function, which looked like this:

> ** Reason for termination = 
> ** {undef,[{regexp,sh_to_awk,["shore"],[]},
>            {ejabberd_s2s_in,match_labels,2,
>                             [{file,"ejabberd_s2s_in.erl"},
> {line,834}]},

That "undef ..." clause means that Erlang failed to locate and load the
definition of the regexp:sh_to_awk/1 function (that is, a function
named "sh_to_awk" located in a module named "regexp").

My fix to this bug was to provide a custom implementation which behaves
(almost) identical to the now gone regexp:sh_to_awk/1.

Now that I have shed some light onto this issue (I hope), I have the
next question: do you observe any crash reports, like the one cited
above, in your logs?  I'm asking because it's now really clear why do
you have s2s issues: is this because my fix doesn't work or because
you're having clustering issues?  Or may be the reason is something
completely different?

Two more things.

1) You seem to be affected by that EJAB-1565 issue; if so, please file
   a bug in the Debian BTS against 2.1.10 with the "upstream" tag, so
   we know the issue exists that affects at least one of our users.

2) We have 2.1.11 package ready and could possibly upload it into
   experimental if you feel like experimenting with it and seeing if it
   can resolve some of your problems.  So if you feel like such an
   upload could be handy, let us know.

1. http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=670307#5


-- 
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]

Reply via email to