[
https://issues.apache.org/jira/browse/DERBY-6354?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13845275#comment-13845275
]
Knut Anders Hatlen commented on DERBY-6354:
-------------------------------------------
Thanks for looking into this, Dag. I'm not sure I've understood what the extra
diagnostics code does. It seems to write the stack trace to System.err when
holdsThis == false? Will that stack trace be any different from the one
reported by the SanityManager.ASSERT() call right below it?
I found this bug that was reported against OpenJDK:
https://bugs.openjdk.java.net/browse/JDK-8028564
It sounds like a probable root cause for the NPE we're seeing. The NPE happens
because we get null from ConcurrentHashMap.get() for an entry that we expect to
be there. If there's a bug in the JDK that causes elements to be lost from
ConcurrentHashMaps in certain situations, that NPE sounds exactly like the kind
of symptom one might expect.
I'll keep an eye on the bug report and ask to have the JDK upgraded on the test
rigs when the fix is available in an official build.
> DatabaseMetaDataTest.testGetColumnsReadOnly fails with NPE when releasing page
> ------------------------------------------------------------------------------
>
> Key: DERBY-6354
> URL: https://issues.apache.org/jira/browse/DERBY-6354
> Project: Derby
> Issue Type: Bug
> Components: Store
> Affects Versions: 10.11.0.0
> Reporter: Knut Anders Hatlen
> Attachments: derby-6343-diag.diff, fail-solaris.zip, fail-windows.zip
>
>
> Seen twice in the nightly testing lately:
> http://download.java.net/javadesktop/derby/request_5581975/javadb-task-3729984.html
> http://download.java.net/javadesktop/derby/request_5581997/javadb-task-3730576.html
> Both failures happened with JDK 8-ea-b107 (Windows and Solaris).
--
This message was sent by Atlassian JIRA
(v6.1.4#6159)
