You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@manifoldcf.apache.org by "Karl Wright (JIRA)" <ji...@apache.org> on 2014/07/17 13:40:04 UTC

[jira] [Commented] (CONNECTORS-998) FileSystem connector derby tests lock up when non-empty version string is used for directories

    [ https://issues.apache.org/jira/browse/CONNECTORS-998?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14064841#comment-14064841 ] 

Karl Wright commented on CONNECTORS-998:
----------------------------------------

To reproduce the problem using the branch I have created:

(1) svn co https://svn.apache.org/repos/asf/manifoldcf/branches/CONNECTORS-998
(2) cd CONNECTORS-998
(3) ant make-core-deps build
(4) cd connectors/filesystem
(5) ant run-IT-derby

In a separate window:
(1) cd CONNECTORS-998/connectors/filesystem/test-derby-output
(2) tail -f manifoldcf.log | grep "TEST FAILED"

When you see the failure, the test has definitely failed, and you can look at the manifoldcf.log file for everything that took place.
Note that the test in question is APISanityDerbyIT; ant may run other tests first on your system.  If that happens, you may want to remove other tests with names ending in DerbyIT.



> FileSystem connector derby tests lock up when non-empty version string is used for directories
> ----------------------------------------------------------------------------------------------
>
>                 Key: CONNECTORS-998
>                 URL: https://issues.apache.org/jira/browse/CONNECTORS-998
>             Project: ManifoldCF
>          Issue Type: Bug
>          Components: Framework core
>    Affects Versions: ManifoldCF 1.7
>            Reporter: Karl Wright
>            Assignee: Karl Wright
>             Fix For: ManifoldCF 1.7
>
>         Attachments: edited.manifoldcf.log
>
>
> The reason for the lock-up seems to be multiple threads trying to change the same records in the JobQueue table.  One thread is adding references, and another thread is marking the record as "complete".  The starting record status field seems to be the same, despite transactions and FOR UPDATE clauses on both queries.
> Since this requires incorrect file connector code, I will be creating a branch to explore it in more detail, and perhaps open a Derby ticket.  I also updated the Derby version to the current release in hopes that this may have fixed the problem.
>  



--
This message was sent by Atlassian JIRA
(v6.2#6252)