It would be incredibly helpful, if you could attach a stacktrace taken at the
moment when the build is not responding anymore. Otherwise I can probably only
guess what's the problem. You can use jstack
(http://java.sun.com/javase/6/docs/technotes/tools/share/jstack.html) fot that.

Also, please append any logs. For instance: don't you have any log entries like:
"Connected to XMPP on "
or
"Trying to reconnect"
?

i'm trying to reproduce the problem to get you a stacktrace.
this doesn't seem to happen if i don't use the suffix feature.

i have some
Aug 24, 2009 10:43:08 AM hudson.plugins.jabber.im.transport.JabberIMConnection
connectAll
INFO: Connected to XMPP on im.f4-toys.com:5222
Aug 24, 2009 10:43:07 AM
hudson.plugins.jabber.im.transport.JabberIMConnection$Connector run
INFO: Connect thread interrupted

in the logs, and some working notifications.

while i had the problems, i could not find anything in the logs related to
jabber. i'll update as soon as i reproduce (or close otherwise)

I've got an idea how the build could hang, but that involves a blocking firewall
and a reconnection attempt from the plugin to the Jabber server.

the problem does not appear anymore, despite going back to the options which
generated the problem at first.

i'll reopen with stack and logs if i happen to reproduce in the future.

thanks for the help anyway

(our jabber server is not behind a firewall so this is probably not the problem i
encountered)

closing.

ok it just happened again

logs (from the logs file, filtered with jabber keyword)

Aug 27, 2009 11:20:00 AM
hudson.plugins.jabber.im.transport.JabberIMConnection$Connector run
INFO: Connect thread interrupted
Aug 27, 2009 11:20:00 AM hudson.plugins.jabber.im.transport.JabberIMConnection
connectAll
INFO: Connected to XMPP on im.f4-toys.com:5222
...
Aug 27, 2009 11:54:26 AM hudson.plugins.jabber.im.IMPublisher perform
INFO: Notifying suspects
Aug 27, 2009 11:54:26 AM hudson.plugins.jabber.im.IMPublisher
calculateSuspectsTargets
INFO: Default Suffix: @f4-group.com
Aug 27, 2009 11:54:26 AM hudson.plugins.jabber.im.IMPublisher
calculateSuspectsTargets
INFO: Possible target: dmo dmo
Aug 27, 2009 11:54:26 AM hudson.plugins.jabber.im.IMPublisher
calculateSuspectsTargets
INFO: Possible target: dmo dmo
Aug 27, 2009 11:54:26 AM hudson.plugins.jabber.im.IMPublisher
calculateSuspectsTargets
INFO: Possible target: yjo yjo
...
Aug 27, 2009 12:50:31 PM hudson.plugins.jabber.im.IMPublisher perform
INFO: Notifying suspects
Aug 27, 2009 12:50:31 PM hudson.plugins.jabber.im.IMPublisher
calculateSuspectsTargets
INFO: Default Suffix: @f4-group.com
Aug 27, 2009 12:50:31 PM hudson.plugins.jabber.im.IMPublisher
calculateSuspectsTargets
INFO: Possible target: dmo dmo
Aug 27, 2009 12:50:31 PM hudson.plugins.jabber.im.IMPublisher
calculateSuspectsTargets
INFO: Possible target: yjo yjo
Aug 27, 2009 12:50:31 PM hudson.plugins.jabber.im.IMPublisher
calculateSuspectsTargets
INFO: Possible target: yjo yjo
Aug 27, 2009 12:50:31 PM hudson.plugins.jabber.im.IMPublisher
calculateSuspectsTargets
INFO: Possible target: yjo yjo

however i have this in logs :
Aug 27, 2009 11:21:20 AM org.jinterop.dcom.core.PingObject read
SEVERE: Some error ! Simple ping failed , hresult: 1912
like a thousand times

in the build console logs i have :
Recording test results
Sending notification to suspect: yjo@f4-group.com
Sending notification to suspect: dmo@f4-group.com

and that's the last lines.

stack traces incoming

right now, i have 2 builds stuck in the same way.
i did two stack trace, one with all my running build, one with hudson in shuting
down mode and the 2 stuck builds (which i can't stop from hudson)

As far as I've seen org.jinterop.dcom.core.PingObject has nothing to do with the
Jabber plugin.

Please remember to attach the stacktraces.

thanks

Created an attachment (id=874)
stack with full builds hudson

Created an attachment (id=875)
stack with only stuck builds on hudson

just to confirm :
this line : SEVERE: Some error ! Simple ping failed , hresult: 1912 doesn't appear
anymore, but the builds still get stuck. (I restarted hudson)

seems like this is linked to the Default Suffix feature. (because without a
specified suffix, it works well)

both users in the previous log have a Jabber ID specified.

on a sidenote, i did not receive mails for those stuck builds, so it might be the
mail plugin that's failing. it is possible the suffix feature impacts the mail
sending ? (or it might simply be that mails are sent after the jabber plugin
returns)

I assume that the previous build in the cases where Hudson did lock up was an
aborted (or 'not_build') build?

In that case, it is fixed in 0.10

yup, the stuck builds were indeed aborted
i updated, i'll close tomorrow if it's ok

thanks for the quick fix !

it works ! thanks