[infinispan-issues] [JBoss JIRA] Resolved: (ISPN-1175) XA transaction rollback failure throws the wrong exception and fails to cleanup the transaction table

[Mircea Markus (JIRA)]

     [ https://issues.jboss.org/browse/ISPN-1175?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Mircea Markus resolved ISPN-1175.
---------------------------------

    Resolution: Duplicate Issue


This is a duplicate of https://issues.jboss.org/browse/ISPN-1036 and should be solved when integrating https://github.com/infinispan/infinispan/pull/403

> XA transaction rollback failure throws the wrong exception and fails to cleanup the transaction table
> -----------------------------------------------------------------------------------------------------
>
>                 Key: ISPN-1175
>                 URL: https://issues.jboss.org/browse/ISPN-1175
>             Project: Infinispan
>          Issue Type: Bug
>          Components: Transactions
>    Affects Versions: 5.0.0.CR5
>            Reporter: Sanne Grinovero
>            Assignee: Mircea Markus
>             Fix For: 5.0.0.CR7
>
>
> I'm getting an NPE, as tx is null at org.infinispan.transaction.TransactionTable.failureCompletingTransaction(Transaction tx)
> being invoked from org.infinispan.transaction.TransactionCoordinator.rollback(LocalTransaction)
> which has this code:
> {code}LocalTxInvocationContext ctx = icc.createTxInvocationContext();
>       ctx.setLocalTransaction(localTransaction);
>       try {
>          invoker.invoke(ctx, rollbackCommand);
>          txTable.removeLocalTransaction(localTransaction);
>       } catch (Throwable e) {
>          txTable.failureCompletingTransaction(ctx.getTransaction());
>          log.errorRollingBack(e);
>          throw new XAException(XAException.XA_HEURHAZ);
>       } finally {
>          icc.suspend();
>       }{code}
> you cant see that a localtransaction is set on the context, but then ctx.getTransaction() is passed to the failing method. This variable was unset.
> so basically I debugged it already but I'm not sure how this should be solved.
> This is currently reproducible by running org.infinispan.lock.APITest (as it fails for some other reason)

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira