Doesn't look like Kathy's email went through to the harness alias.
Forwarding... -- Ken
attached mail follows:
Ken-
Thought about this.
Ken Hofsass wrote:
>
> Caveats yet to be corrected:
> 1. server side is still running wsimport, even though it's a provider
> endpoint - can be automatically detected and skipped.
With a Dispatch client, it is not necessary to have a provider endpoint.
Tests for Dispatch will need the flexibility to run against WSDL
Endpoints as well.
>
>
> 2. client side is still running wsimport even thought it's a
> dispatch... thinking of adding an boolean 'dispatch' attribute to the
> <client> element in the descriptor. Since if client talking to a
> WSDL-less provider, wsimport would obviously fail. Thoughts?
That's ok, however, it brings up another issue regarding Dispatch
clients using JAXB data binded Objects. The Dispatch test does need a
way to use
xjc to generate these objects before the tests. In some of the Dispatch
unit tests with JAXB Objects, xjc is used while in other unit tests
wsimport is the
tool used. Developers might use either tool so it would be good if the
test harness flexibility provided for use of either.
cheers-
Kathy
>
>
> thanks,
> Ken
>
> Ken Hofsass wrote:
>
>> Hey there folks,
>>
>> I'm actually making good progress on this... well, I think so anyhow,
>> it will be clear later this afternoon. :-) I've checked in my
>> changes made so far to the ws-test-harness workspace... and a couple
>> changes to test.bsh in Bhakti's dispatch test. Most notably, the
>> endpointAddress isn't exported to the script yet... once that's in,
>> we'll see if it all works.
>>
>> Ken
>>
>> kathy walsh wrote:
>>
>>> thanks-
>>> I did see that bahkti checked in a non-working dispatch test. Had it
>>> in mind to
>>> take a look at it.
>>> Thanks to you and bhakti for working on this-
>>> Kathy
>>>
>>> Ken Hofsass wrote:
>>>
>>>> Hi Kathy, fyi, I'm working on it (with Bhakti too). Not there yet.
>>>> Bhakti's already converted a dispatch test (which of course blows
>>>> up)... she had one in the WS-RM tests. If you want to look at it,
>>>> it's in wsit/test/e2e/testcases/wsrm/dispatch.
>>>>
>>>> Ken
>>>>
>>>> kathy walsh wrote:
>>>>
>>>>> Ken-
>>>>> Can you add the endpointAddress variable to the test harness?
>>>>>
>>>>> Also, can you check to make sure the harness can control the
>>>>> communication
>>>>> channel.
>>>>>
>>>>> I can work with you to write a dispatch test, but still I am not
>>>>> that familiar
>>>>> with the test harness. I will consult the user guide again-
>>>>>
>>>>> Thanks,
>>>>> Kathy
>>>>>
>>>>> -------- Original Message --------
>>>>> Subject: Re: Dispatch and new test harness
>>>>> Date: Mon, 07 Aug 2006 14:03:38 -0700
>>>>> From: Kohsuke Kawaguchi <Kohsuke.Kawaguchi_at_Sun.COM>
>>>>> Reply-To: dev_at_jax-ws.dev.java.net
>>>>> To: Kathleen.Walsh_at_Sun.COM
>>>>> CC: dev_at_jax-ws.dev.java.net, dev_at_ws-test-harness.dev.java.net,
>>>>> Ken Hofsass <Kenneth.Hofsass_at_Sun.COM>
>>>>> References: <44D0DB5A.5030506_at_sun.com>
>>>>>
>>>>>
>>>>>
>>>>> kathy walsh wrote:
>>>>>
>>>>>> Kohsuke,
>>>>>>
>>>>>> I haven't looked much at the test harness yet, and am wondering
>>>>>> if you
>>>>>> can implement a dispatch test sample and install it in the
>>>>>> workspace.
>>>>>>
>>>>>> With the caveat that I haven't examined the harness closely,
>>>>>> I have a bit of concern as to whether support for dispatch clients
>>>>>> is currently present.
>>>>>>
>>>>>> Dispatch clients may not have access to wsdl, and standardly do not
>>>>>> use jax-ws tools. Can you send some attention to dispatch and
>>>>>> the test harness and reassure me that dispatch capability is
>>>>>> present?
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> Thanks for bringing this up, and sorry for my delay in the response!
>>>>>
>>>>> We certainly don't have a dispatch example today, but I believe
>>>>> it's relatively straight-forward to support it. The only aspect
>>>>> that we need to pay attention is that the harness needs to 'own'
>>>>> the communication channel between the client and the server, and
>>>>> hidden from the test code. This is so that the test data can be
>>>>> reused across multiple containers and transports.
>>>>>
>>>>> The dispatch is bit tricky with this respect, because the test
>>>>> code needs to set the endpoint address, which is otherwise hidden
>>>>> by the harness. And my current thinking is that we just export
>>>>> variable 'endpointAddress' into the test code (for the test code
>>>>> to create a dispatch), and I think this still allow the harness to
>>>>> control the communication channel.
>>>>>
>>>>> Please talk to Ken. I believe he can change the harness to export
>>>>> this new variable, and work with you to get one sample from the
>>>>> current JAX-WS RI converted into the new harness.
>>>>>
>>>>> It will be a good exercise for the test harness team as well.
>>>>>
>>>
>>