karaf-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Daniel Kulp (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (KARAF-3722) Feature resolver change bundles package export
Date Tue, 12 May 2015 15:14:00 GMT

     [ https://issues.apache.org/jira/browse/KARAF-3722?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Daniel Kulp updated KARAF-3722:
    Attachment: testcase.tar.gz

The fix gnodet did fixes the "display" but doesn't actually fix the problem.   A client that
hits the service will get the NoClassDefFound stuff.

I've attached a somewhat-bigger-than-I-wanted test case.   Build the two projects in it. 
The start a clean karaf and do:

feature:repo-add cxf 3.1.1-SNAPSHOT
feature:install cxf
feature:install cxf-sts

install -s mvn:org.talend.cxf-examples.jaxws-ws-secpol/ws-secpol-common/1.0
install -s mvn:org.talend.cxf-examples.jaxws-ws-secpol/ws-secpol-server/1.0
install -s mvn:org.talend.cxf-examples.jaxws-cxf-sts-advanced/jaxws-cxf-sts-advanced-sts/1.0
install -s mvn:org.talend.cxf-examples.jaxws-cxf-sts-advanced/jaxws-cxf-sts-advanced-wsp/1.0
install -s mvn:org.talend.cxf-examples.jaxws-ws-secpol/ws-secpol-client/1.0

That should get a WORKING sample up and running.  You should see something like "server responded
with: Hello dkulp" on the console.  A "log:display" should see some soap messages going back
and forth and such.

Now, stop the last bundle (ws-secpol-client) and then do:
feature:install spring-web/3.2.11.RELEASE_1
That seems to trigger a large refresh since that adds spring and thus CXF gets refreshed to
include spring (I think).    Now start the ws-secpol-client bundle again.   It won't get the
above line.  A log:display shows some stack traces.  The important one being:

Caused by: java.lang.ClassNotFoundException: org.opensaml.saml.common.SAMLVersion not found
by org.apache.cxf.cxf-rt-ws-security [93]
	at org.apache.felix.framework.BundleWiringImpl.findClassOrResourceByDelegation(BundleWiringImpl.java:1558)[org.apache.felix.framework-5.0.0.jar:]
	at org.apache.felix.framework.BundleWiringImpl.access$400(BundleWiringImpl.java:79)[org.apache.felix.framework-5.0.0.jar:]
	at org.apache.felix.framework.BundleWiringImpl$BundleClassLoader.loadClass(BundleWiringImpl.java:1998)[org.apache.felix.framework-5.0.0.jar:]
	at java.lang.ClassLoader.loadClass(ClassLoader.java:358)[:1.7.0_71]

which shows that the bundle classloader for the cxf-rt-ws-security does not have access to
the OpenSAML stuff that it needs.

> Feature resolver change bundles package export
> ----------------------------------------------
>                 Key: KARAF-3722
>                 URL: https://issues.apache.org/jira/browse/KARAF-3722
>             Project: Karaf
>          Issue Type: Bug
>          Components: karaf-feature
>    Affects Versions: 4.0.0.M3
>            Reporter: Jean-Baptiste Onofré
>            Priority: Blocker
>             Fix For: 4.0.0.M3
>         Attachments: testcase.tar.gz
> With latest 4.0.0-SNAPSHOT, when we code:
> {code}
> feature:repo-add cxf 3.1.1-SNAPSHOT
> feature:install cxf
> {code}
> and check the exports of the CXF security bundle (with bundle ID 93 on my box), we have:
> {code}
> imports -b 93
> org.opensaml.saml.common                       | [3.1.0,4.0.0) | resolved | 93 | org.apache.cxf.cxf-rt-ws-security
> org.opensaml.saml.saml1.core                   | [3.1.0,4.0.0) | resolved | 93 | org.apache.cxf.cxf-rt-ws-security
> org.opensaml.saml.saml2.core                   | [3.1.0,4.0.0) | resolved | 93 | org.apache.cxf.cxf-rt-ws-security
> …..
> net.sf.ehcache                                 | [2.5.0,3.0.0) | resolved | 93 | org.apache.cxf.cxf-rt-ws-security
> net.sf.ehcache.config                          | [2.5.0,3.0.0) | resolved | 93 | org.apache.cxf.cxf-rt-ws-security
> {code}
> Now, we install the cxf-sts feature and do the same thing:
> {code}
> feature:install cxf-sts
> imports -b 93
> {code}
> the optional imports are no longer resolver (causing NoClassDefFound cxf-rt-ws-security
for instance).
> It looks like the resolver change the export/resolution.
> It's an important regression IMHO that we have to address.

This message was sent by Atlassian JIRA

View raw message