> Sorry I broke your app, I was kind of worried when I realized the xa rar
> had never been assembled, but noone said "no".  Is there some usable or
> automatic way to mark in cvs all the changes that go together like the ones
> I made?

No worries.  I have a version of JBoss from thursday which I can use until
this is fixed.  You are moving in the right direction.  Things like this
will happen.  Lets just try to minimize the frequency and keep on top of
such problems when the occur.

> Can you provide any more details of what's going on?  Did you get the
> Oracle XA driver to work?  Are you using mdb? Is 2pc happening?

I reverted to the XADataSourceLoader using the
org.jboss.pool.jdbc.xa.wrapper.XADataSourceImpl currently.  I tried the
oracle classes early on and had huge problems with them.  These worked fine,
so I just kept using them.

I am using lots of stuff, spanning multipule vms:

  o stateless sessions
  o stateful sessions
  o cmp entities (1.1)
  o jmx services
  o standard jms
  o mdb
  o jms ra
  o jbossmq
  * and more

> One difference between the XADataSourceLoader connection management and the
> xa-rar management is that the old way routed all work on one transaction to
> the same managed connection, no matter how many beans were in the
> transaction.  The jca stuff does not do this, whenever you need to do some
> work it gets a random connection, calls start (on the XAResource), does the
> work, calls end.  This for instance breaks the XADataSourceImpl which has a

Yikes.

> check that when you call start, and the connection has done work on some
> transaction, you need to be supplying that same xid.  The last time I
> looked I couldn't find the spec saying we can be calling start(xid) with
> any xid, but I did find a part saying the tm can call commit on any
> xaresource, not necessarily one any work was done on.
>
> I found this problem trying the banktest with the xa-rar using
> XADataSourceImpl as the "xa" driver.  The problem occurred in
> testMultiThread and TestMultiTest2 after some time, but for me pretty
> consistently in each run.  What happens when you run banktest with your
> Oracle driver?  Maybe it has a similar problem.  Dr. Jung was proposing a

I can revisit the Oracle ds impl again... but we should fix the
XADataSourceImpl, as many folks (like myself) will be using it.  You can bet
on that.

--jason


_______________________________________________
Jboss-development mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/jboss-development

Reply via email to