[
https://issues.apache.org/jira/browse/FELIX-335?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12520362
]
Karl Pauls commented on FELIX-335:
----------------------------------
With your patch, I see the following exception appear quite often in a heavily
multithreaded project:
java.lang.NullPointerException
at
org.apache.felix.cm.file.FilePersistenceManager.encodePid(FilePersistenceManager.java:140)
at
org.apache.felix.cm.file.FilePersistenceManager.getFile(FilePersistenceManager.java:473)
at
org.apache.felix.cm.file.FilePersistenceManager.store(FilePersistenceManager.java:394)
at
org.apache.felix.cm.impl.ConfigurationImpl.store(ConfigurationImpl.java:340)
at
org.apache.felix.cm.impl.ConfigurationImpl.setBundleLocation(ConfigurationImpl.java:223)
at
org.apache.felix.cm.impl.ConfigurationManager$ManagedServiceUpdate.run(ConfigurationManager.java:840)
at org.apache.felix.cm.impl.UpdateThread.run(UpdateThread.java:89)
> Configuration Admin updates with empty properties and throws
> NullPointerExceptions due to race condition
> --------------------------------------------------------------------------------------------------------
>
> Key: FELIX-335
> URL: https://issues.apache.org/jira/browse/FELIX-335
> Project: Felix
> Issue Type: Bug
> Components: Configuration Admin
> Reporter: Karl Pauls
> Assignee: Felix Meschberger
> Attachments: ConfigAdmin.FELIX-335.patch, empty-properties.patch
>
>
> There appears to be a race condition in the configuration admin that allows
> managed services to get updated with empty properties (but not null).
> Furthermore, due to the same bug NullPointerExceptions appear under heavy
> load.
> Fortunately, I think I was able to isolate the bug. The issue is that
> configurations are persisted when created but before they are initialized
> (i.e., given a configuration). If during this window a new ManagedService
> appears it gets updated with an empty Dictionary. Likewise, if the
> configuration is re-read after such an update attempt it does not have a pid
> which causes the null pointer exceptions.
> The fix seems to be rather simple (unless I'm missing something :-) but took
> me some time to figure out.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
