Community
Participate
Working Groups
We are using some ch.qos.logback.* orbit bundles taken from this recommended download location http://download.eclipse.org/tools/orbit/downloads/drops/R20100519200754/ and we are experiencing some warnings during shutdown of our application which is using these bundles: !ENTRY org.eclipse.osgi 2 0 2011-03-14 10:34:14.043 !MESSAGE Info: cycle(s) found while stopping bundles: [[ch.qos.logback.classic,org.slf4j.api,ch.qos.logback.slf4j,ch.qos.logback.core]]. I think bug 339871 and bug 342802 are caused due to a problem in these bundles.
According to bug 115720 the message should only be logged when running in debug mode "-debug" option set. According to bug 275676 comment 2 it also does not do any harm. The bundles don't have activators and so the shutdown order is irrelevant. Also note that it's not a warning but an "INFO" message. I previously reported this with Equinox. See bug 314289.
Marking as dup. *** This bug has been marked as a duplicate of bug 314289 ***
I thought this bug was opened to get rid of the cycle because that causes other build issues?
*** Bug 339871 has been marked as a duplicate of this bug. ***
(In reply to comment #3) > I thought this bug was opened to get rid of the cycle because that causes other > build issues? I'm not sure how to resolve the cycle. It's "by design". Bundle slf4j.api provides package org.slf4j Bundle slf4j.api expects a fragment to provide package org.slf4j.impl (some class in org.slf4j will load a class from org.slf4j.impl) Bundle logback.* requires package org.slf4j Fragment logback.slf4j (host slf4j.api) provides package org.slf4j.impl based on logback.* The cycle is introduced by the fragment but this is the inherited design of how SLF4J expects a single native logger implementation to be contributed (Class#forName).
(In reply to comment #5) > (In reply to comment #3) > > I thought this bug was opened to get rid of the cycle because that causes other > > build issues? > > I'm not sure how to resolve the cycle. It's "by design". > I understand that. Then this should be closed as wontfix instead of as a dup of the framework bug about the logging of the cycle. Anyway, it is just the semantics of what you meant when you closed the bug. ;-)
Done. :)
