[ejabberd] Fix or workaround for EJAB-731 - shared roster fails to show new user accts

Martin Langhoff martin.langhoff at gmail.com
Sat Dec 20 16:20:00 MSK 2008


On Sat, Dec 20, 2008 at 10:01 AM, Badlop <badlop at gmail.com> wrote:
>> - is restarting ejabberd graceful to the state of the clients?
>
> When ejabberd is restarted, the TCP connections and processes that
> manage users' sessions are stopped, right? The clients are kicked out
> of the server and need to relogin. That ensures they get the updated
> roster.

So it's not trasparent for the end user -- I just tested it and at
least the Sugar client code doesn't reconnect transparently. We can't
go around restarting ejabberd freely :-/

>> - is there an altrnative workaround, something to force ejabberd to
>> re-scan its list of users without a restart?
>
> Yes, check
> https://support.process-one.net/browse/EJAB-731
> That new patch for ejabberd 2.0.2 fixes this and other problems
> related to Shared Roster. There is a comment at the bottom of the page
> with explanation of the changes.
>
> If you try the patch, please tell me if all seems to work correctly.

Right - following that discussion, you are saying that the patch that
fixed EJAB-71 should fix it, correct? But we are seeing the problem on
2.0.1, and that release already has that patch. Trying to apply that
patch, I get a "patch already applied, revert?".

Which is in line with what Guillaume stated in the bug. We _are_
registering inline, and yet the online roster isn't getting the
updates.

What can I do to debug this?

cheers,



m
-- 
 martin.langhoff at gmail.com
 martin at laptop.org -- School Server Architect
 - ask interesting questions
 - don't get distracted with shiny stuff  - working code first
 - http://wiki.laptop.org/go/User:Martinlanghoff


More information about the ejabberd mailing list