[
https://issues.apache.org/jira/browse/GERONIMO-4576?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Christian Schneider updated GERONIMO-4576:
--
Attachment: GERONIMO-4576-1.patch
I attached a patch that fixes this issue. The problem was that calling
setRollbackOnly() created a quite uninformative exception and set it using
markRollbackCause(e). Then at a later point the real exception could not be set
anymore as markRollbackCause only sets the exception if the current one is not
set.
My patch avoids the generic exception and only calls markRollbackCause if a
real exception was given.
> Make persistence exceptions more visible to client
> --
>
> Key: GERONIMO-4576
> URL: https://issues.apache.org/jira/browse/GERONIMO-4576
> Project: Geronimo
> Issue Type: Improvement
> Security Level: public(Regular issues)
> Components: persistence
>Affects Versions: 2.2
> Environment: Linux, Windows
>Reporter: Joe Bohn
>Priority: Minor
> Fix For: Wish List
>
> Attachments: GERONIMO-4576-1.patch
>
>
> See http://issues.apache.org/jira/browse/GERONIMO-3907 for details of the
> original problem. That core problem was resolved. However, upon resolution
> it was mentioned that it would be beneficial to report more specific failure
> information back to the client. From GERONIMO-3907:
> Ralf Baumhof - 06/May/08 06:17 AM
> Today if have tested the new Geronimo release 2.1.1 (published on
> 28.04.2008). The problem is now fixed. If the server gets an error on trying
> a commit, this error is now thrown to the web bean.
> Exception text:
> javax.ejb.EJBTransactionRolledbackException: Transaction was rolled back,
> presumably because setRollbackOnly was called during a synchronization:
> Unable to commit: transaction marked for rollback Root Cause:
> javax.transaction.TransactionRolledbackException : Transaction was rolled
> back, presumably because setRollbackOnly was called during a synchronization:
> Unable to commit: transaction marked for rollback
> Unfortunately there is no proper root cause attached to the exception. So the
> cause can only be seen in the server console, but can not be reported to the
> user. It would be very nice if you could change this in a later release.
> Thanks for your help.
> Vincent MATHON - 06/Nov/08 02:03 AM
> I agree that exceptions on the server side should not be thrown to the client
> side since such exceptions types might not be known by the client. However,
> for unit testing purpose, it is useful to attach the root cause to the
> RollBackException. I have modified a few lines in
> org.apache.geronimo.transaction.manager.TransactionImpl.java to attach the
> root cause in case of RollBackException and it works for my unit testing
> purpose (I have not enough background on the java transaction architecture
> topic to submit a patch for production). It would be great to define a
> configuration parameter that permits to provide the root cause to the client
> and keep the current behaviour that does not by default.
> Geoff Callender - 22/Dec/08 03:36 AM
> It's useful for more than unit testing - it's essential to be able to inform
> the client what's wrong with the request. I've added some examples of this to
> https://issues.apache.org/jira/browse/OPENEJB-782 .
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
