[
https://issues.apache.org/jira/browse/JCR-3617?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13697642#comment-13697642
]
Marcel Reutegger commented on JCR-3617:
---------------------------------------
Committed a preliminary fix in revision: 1498850.
This does not fix the root cause, but rather detects the inconsistent state and
works around it after it happened.
I'll further investigate how the CachingHierarchyManager gets into the
inconsistent state.
> Inconsistent CachingHierarchyManager under concurrent access
> ------------------------------------------------------------
>
> Key: JCR-3617
> URL: https://issues.apache.org/jira/browse/JCR-3617
> Project: Jackrabbit Content Repository
> Issue Type: Bug
> Components: jackrabbit-core
> Affects Versions: 2.2, 2.4, 2.6
> Reporter: Marcel Reutegger
> Assignee: Marcel Reutegger
> Priority: Minor
>
> This is a bit difficult to reproduce and so far I'm not able to provide a
> standalone test case for this issue. However, the following happens on the
> application level: a sub-tree is replaced with a modified version of the
> sub-tree while event listeners track those changes and try to get items for
> the given event paths. In some cases the repository throws an exception when
> Session.getItem() is called similar to what was reported in JCR-3368.
> It is important to note that the replaced subtree has some special
> characteristics. The root node of the sub-tree is re-created with the same
> UUID, while descendant nodes may be replaced with different UUIDs, but still
> have the same name.
> There seems to be a short time window where the modifying Session saves this
> kind of change, which gets propagated up to the ItemState layers and into the
> CachingHierarchyManager and the listening Session (owning the
> CachingHierarchyManager) access modified items at the same time through the
> CachingHierarchyManager. In some cases this seems to create an inconsistent
> state in the CachingHierarchyManager where a path is mapped to the old UUID
> (and vice versa) even though the replace already happened.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
