[
https://issues.apache.org/jira/browse/JCR-993?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12509524
]
Dominique Pfister commented on JCR-993:
---------------------------------------
Hi Stefan,
Thank you for reporting this nasty bug in the CachingHierarchyManager (CHM) and
providing a testcase that makes it really easy to reproduce what happens
underneath. I found the problem to be the following: initially the ordering of
elements is as follows:
T1: a b c
You then reorder a before a (which is a no-op) and b before a. This results in:
T2: b a c
At this point, b's path is retrieved and the CHM caches it as the first element
in the list. You finally move c before b:
T3: c b a
The CHM will receive a notification about this reordering, and retrieve the
list of reordered child node entries. Since b did not change its position
compared to T1 (which is the last state that was persisted) the change in b's
index passes unnoticed.
> corrupted paths after moving nodes
> ----------------------------------
>
> Key: JCR-993
> URL: https://issues.apache.org/jira/browse/JCR-993
> Project: Jackrabbit
> Issue Type: Bug
> Affects Versions: 1.3
> Reporter: Stefan Rinner
> Assignee: Dominique Pfister
> Attachments: JackrabbitPathProb.java
>
>
> we just found a bug which corrupts the results of Node.getPath() - it seems
> to be related to older Jackrabbit bugs (e.g. JCR-768) but still happens in
> jackrabbit 1.3 and jackrabbit-1.4-SNAPSHOT
> Basically we have a node with 3 subnodes (a, b, c), we move all of them to
> index 1 - this works fine, unless we call getPath() of the third Node before
> moving it.
> The expected paths after moving would be:
> a: /pages[37]/page/element[3]
> b: /pages[37]/page/element[2]
> c: /pages[37]/page/element
> But we get these paths:
> a: /pages[37]/page/element[3]
> b: /pages[37]/page/element
> c: /pages[37]/page/element
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
