You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@qpid.apache.org by Steve Huston <sh...@riverace.com> on 2010/04/12 19:35:34 UTC
Status of qpid/tests/src/py/qpid_tests/broker_0_10?
Hi folks,
I'm working on some testing for the Windows persistence stuff and ran
across a seeming treasure trove of tests in
qpid/tests/src/py/qpid_tests/broker_0_10/dtx.py. What's the status of
these? I haven't found any reference to these tests being run regularly,
and I have some questions about them. Can anyone help me please?
When a bunch of these tests started failing when run against the Windows
persistence stuff, I fixed some problems in the Windows code, but now
I'm seeing a failure that looks like a test problem itself. I can say
this with relative confidence since the test fail is being triggered by
broker code that isn't getting to the store. In
test_invalid_commit_one_phase_true, the code does, effectively:
Start a session
Select dtx
Begin
Send a message
Prepare
Commit (one_phase=true)
The last commit is illegal and the broker correctly reports this; but
then the test tries to do a rollback but the session was disconnected on
the failed commit, so the rollback also fails which is reported in the
test as a broken socket.
Can someone confirm or deny for me:
1. The broker_0_10 tests' reliability and use?
2. The behavior I'm seeing - should the test be changed to re-establish
the session and do the rollback? Or is the test code correct, and is the
session should still be valid after failing the commit?
Thanks,
-Steve
--
Steve Huston, Riverace Corporation
Total Lifecycle Support for Your Networked Applications
http://www.riverace.com
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:dev-subscribe@qpid.apache.org
RE: Status of qpid/tests/src/py/qpid_tests/broker_0_10?
Posted by Steve Huston <sh...@riverace.com>.
Hi Gordon,
> On 04/12/2010 06:35 PM, Steve Huston wrote:
> > Hi folks,
> >
> > I'm working on some testing for the Windows persistence
> stuff and ran
> > across a seeming treasure trove of tests in
> > qpid/tests/src/py/qpid_tests/broker_0_10/dtx.py. What's the
> status of
> > these? I haven't found any reference to these tests being run
> > regularly, and I have some questions about them. Can anyone help me
> > please?
>
> They should get run during make check as part of python_tests.
Hmmm... I didn't port that aspect of python_tests to Windows. That
explains it.
Now it's running all the tests, and indeed, dtx is clean without the
store loaded. Thanks for checking this out, Gordon!
-Steve
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:dev-subscribe@qpid.apache.org
Re: Status of qpid/tests/src/py/qpid_tests/broker_0_10?
Posted by Gordon Sim <gs...@redhat.com>.
On 04/12/2010 06:35 PM, Steve Huston wrote:
> Hi folks,
>
> I'm working on some testing for the Windows persistence stuff and ran
> across a seeming treasure trove of tests in
> qpid/tests/src/py/qpid_tests/broker_0_10/dtx.py. What's the status of
> these? I haven't found any reference to these tests being run regularly,
> and I have some questions about them. Can anyone help me please?
They should get run during make check as part of python_tests.
> When a bunch of these tests started failing when run against the Windows
> persistence stuff, I fixed some problems in the Windows code, but now
> I'm seeing a failure that looks like a test problem itself. I can say
> this with relative confidence since the test fail is being triggered by
> broker code that isn't getting to the store. In
> test_invalid_commit_one_phase_true, the code does, effectively:
>
> Start a session
> Select dtx
> Begin
> Send a message
> Prepare
> Commit (one_phase=true)
>
> The last commit is illegal and the broker correctly reports this; but
> then the test tries to do a rollback but the session was disconnected on
> the failed commit, so the rollback also fails which is reported in the
> test as a broken socket.
>
> Can someone confirm or deny for me:
>
> 1. The broker_0_10 tests' reliability and use?
> 2. The behavior I'm seeing - should the test be changed to re-establish
> the session and do the rollback? Or is the test code correct, and is the
> session should still be valid after failing the commit?
I *think* the test code is correct as the illegal commit is done on a
different session to that doing the subsequent rollback (tester, rather
than self.session).
---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project: http://qpid.apache.org
Use/Interact: mailto:dev-subscribe@qpid.apache.org