Operating System: Mac OS
Platform: Macintosh

Slightly more detailed exception,

Sep 10, 2015 5:40:19 PM org.apache.karaf.main.Main launch
INFO: Installing and starting initial bundles
Sep 10, 2015 5:40:19 PM org.apache.karaf.main.Main main
SEVERE: Could not launch framework
java.lang.RuntimeException: Error installing bundle listed in startup.properties with url: wrap:mvn:javax.servlet.jsp/jsp-api/2.1 and startlevel: 30
at org.apache.karaf.main.Main.installAndStartBundles(Main.java:392)
at org.apache.karaf.main.Main.launch(Main.java:245)
at org.apache.karaf.main.Main.main(Main.java:167)
Caused by: java.lang.RuntimeException: Could not resolve wrap:mvn:javax.servlet.jsp/jsp-api/2.1
at org.apache.karaf.main.util.SimpleMavenResolver.resolve(SimpleMavenResolver.java:59)
at org.apache.karaf.main.Main.installAndStartBundles(Main.java:384)
... 2 more

This seems to only happen on the controller distribution

Hi Moiz,

I still see reports of this occurring in master; do you know if a root cause was ever found for this? Any additional information you can provide is greatly helpful!

Thanks!
Ryan

I'm just hitting something similar, with the infrautils SampleIntegrationTest which worked last week suddenly failing (both locally as well as on Jenkins) with this in infrautils/itestutils/it/target/surefire-reports/org.opendaylight.infrautils.itestutils.it.SampleIntegrationTest-output.txt:

Error installing bundle listed in startup.properties with url: mvn:org.osgi/org.osgi.service.event/1.3.1 and startlevel: 7
Exception in thread "main" java.lang.NullPointerException
at org.apache.karaf.main.Main.destroy(Main.java:634)
at org.apache.karaf.main.Main.main(Main.java:183)
[main] INFO org.ops4j.pax.exam.spi.reactors.ReactorManager - suite finished

and indeed under infrautils/itestutils/it/target/paxexam/.../system/org there is no osgi/ sub-directory - should there be?

Did we just break this by Or did we just break this when merging https://git.opendaylight.org/gerrit/#/c/58053/ ?

see also https://issues.apache.org/jira/browse/KARAF-3374 for the Karaf issue which may in the future provide a better error message than the slightly confusing NullPointerException (but which won't fix the root cause which seems to be in our assembly).

> under system/org there is no osgi/ sub-directory - should there be?

The real question is what needs org.osgi.service.event, and why that does not transitively pull it along... We can see controller.blueprint/pom.xml having a <dependency> to it, so is it missing in a controller feature?

But as just raised in https://lists.opendaylight.org/pipermail/odlparent-dev/2017-June/001051.html, this currently happens even in odlparent and infrautils without controller in the picture:

odlparent/karaf/opendaylight-karaf4-empty/target/assembly/bin/karaf
Apache Karaf starting up. Press Enter to open the shell now...
Error installing bundle listed in startup.properties with url:
mvn:org.osgi/org.osgi.service.event/1.3.1 and startlevel: 7
Exception in thread "main" java.lang.NullPointerException
at org.apache.karaf.main.Main.destroy(Main.java:634)
at org.apache.karaf.main.Main.main(Main.java:183)

https://git.opendaylight.org/gerrit/#/c/58362/ fixes this (above), but I'm not 100% it's the right (or the best) fix .. I'm not clear what actually needs org.osgi.service.event in an empty basic Karaf, and why the Karaf core features that we depend on do not transitively pull this along - something seems off, that we have to do the kind of small work-around I'm proposing in c/58362.

The fix should work. Controller's blueprint extension cannot just pull it in, as pax-logging has an optional dependency on it – leading to a complete container refresh when it pulled in.

On the console this manifests by karaf printing the banner the second time.

> The fix should work.

Confirming; FYI, I've now locally verified that merging c/58362 would fix both odlparent/karaf/opendaylight-karaf4-empty/target/assembly/bin/karaf as well as e.g. infrautils/itestutils/it (SampleIntegrationTest).

> pax-logging has an optional dependency on it

Oh; didn't know. So... shouldn't it (org.osgi.service.event) then have been listed in some upstream Karaf core feature, in an ideal world, so that we would never have had this problem? (Open a bug / propose a change to Apache Karaf - or forget about it?)

Closing as RESOLVED FIXED, as c/58362 got merged.