[#MULE-2011] Reimplement org.mule.test.integration.providers.jms to use real configs

Mule

Reimplement org.mule.test.integration.providers.jms to use real configs

Created: 09/Jul/07 10:10 PM Updated: 16/Jan/08 05:53 AM

Component/s:

Transport: JMS

Affects Version/s:

1.4.1

Fix Version/s:

None

Time Tracking:

Not Specified

Issue Links:

Block

This issue blocks:

MULE-2065 Intermittent JMS Failures

~~MULE-1846~~ JMS transport fails to pass tests in 2.0

This issue is blocked by:

~~MULE-2005~~ Add support for JBoss TS (Arjuna)

Duplicate

This issue is duplicated by:

~~MULE-2071~~ Intermittent ActiveMQ/JMS Failure

Related

This issue relates to:

~~MULE-2006~~ XaTransaction commit/rollback does not propertly dissassociate the current thread of execution from transaction

Labels:
User impact:	Low

Description

« Hide

~~MULE-2006~~ fix changes (or rather, corrects) XA transaction semantics. The problem with these tests (they are really old) is there's a set of assertions in an abstract superclass which are inherited by all subclasses. While they may be valid for single-resource transactions (non-XA), those assumptions are incorrect for XA case. E.g. TX statuses returned are different for XA, up to a point where there's no physical transaction for the current thread.

Additionally, converting them to real configs would remove the burden of maintaining these tests, and not many people could really tell what was going on in that complex hierarchy in a multithreaded environment..

The downside is there will be a need to convert the configs to Mule 2.x after, but I still consider the task worthwhile.

All

Comments

Work Log

Change History

Transitions

FishEye

Sort Order:

[ Permlink | « Hide ]

Anatoli Kuzmin added a comment - 21/Aug/07 09:23 AM

Single tx
1. NONE - OK
2. ALWAYS_BEGIN - OK
3. BEGIN_OR_JOIN - as ALWAYS_BEGIN
Transaction doesn't sharing between descriptors,
4. ALWAYS_JOIN - Exception
See point 3, as a result in this case we get a exception
5. JOIN_IF_POSSIBLE - as NONE
see point 3.

XA tx (Arjuna TM)
1. NONE - Not implemented
need resume and suspend
2. ALWAYS_BEGIN - OK
3. BEGIN_OR_JOIN - OK
4. ALWAYS_JOIN - Exception
In the current strategy started receiver tried to begin transaction, and got Exception "A transaction is not available for this session, but transaction action is "Always Join" "
5. JOIN_IF_POSSIBLE - as NONE
not implemented at all

[ Permlink | « Hide ]

Andreas Guenther added a comment - 02/Oct/07 10:33 AM

Anatoli,

If I remember correctly, this and related issues are still up for discussion on either trunk (1.4.4) or 2.x. Once a decision has been made, please set fix version on this and related issues.

Thanks,
-Andreas

[ Permlink | « Hide ]

Andrew Cooke added a comment - 19/Oct/07 02:46 PM

Hi, I am trying to understand the status of JMS in 2.x. To what extent is this issue complete? Is it still being actively worked on? Thanks!

[ Permlink | « Hide ]

Anatoli Kuzmin added a comment - 26/Oct/07 06:18 AM

added test cases for topics
http://fisheye.codehaus.org/changelog/mule/?cs=9384
http://fisheye.codehaus.org/changelog/mule/?cs=9392

[ Permlink | « Hide ]

Travis Carlson added a comment - 16/Jan/08 05:38 AM

Hi, I'm trying to understand why this issue is still open.

From what I see, the tests in org.mule.test.integration.providers.jms.functional are all based on XML configs now, while the tests in org.mule.test.integration.providers.jms.activemq are still programmatic. Do the tests in o.m.t.functional supercede the tests in o.m.t.activemq (in which case o.m.t.activemq should be deleted)? Or is the plan to convert the remaining tests in o.m.t.activemq to XML configs and that's why the issue is still open?

Not trying to push anyone, just need some clarification on what's still pending for this issue, thanks.

[ Permlink | « Hide ]

Travis Carlson added a comment - 16/Jan/08 05:40 AM

Just for clarification, the tests in question are now in transports/jms, not tests/integration (despite the package name).

[ Permlink | « Hide ]

Anatoli Kuzmin added a comment - 16/Jan/08 05:53 AM

see http://www.nabble.com/Blocked-JMS-tests-in-2.x-to13301929.html#a13301929
The situation has not changed. we had a lot of work in other sub-projects
There is a need to spend a week to implement resume/suspend functionality and write tests
Also during this week I'm planing to resolve ~~MULE-2687~~