axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rob Jellinghaus <r...@unrealities.com>
Subject RE: [GUMP] Function Test Failure - Axis
Date Fri, 15 Jun 2001 00:41:15 GMT
At 07:34 PM 6/14/2001 -0400, Sam Ruby wrote:
>Rob Jellinghaus wrote:
>>I tend to prefer this (fix the underlying problem in the test) to:
>>
>>>> What I would like to do is to focus on factor out the
>>>> transport tests from
>>>> the tests which are simply intended to test the end-to-end
>>>> operation of the
>>>> system.
>
>Interesting....  I explictly state in my note (conveniently omitted from
>your quote) that I am not advocating the removal of any tests, but simply
>the addition of ones that would help isolate the problem area.

Yes, and I agree with that.  I didn't mean to misrepresent your position by
omitting anything.

I think we're getting confused about the concept of "test removal".  Are we
talking about removing tests altogether (neither of us wants that), or
about factoring some tests out of the nightly test suite?  I started this
conversation with a complaint about the nightly suite.  You then suggested
factoring out the transport tests.  I took this to mean you were
considering running only the doLocal tests nightly.  If I misinterpreted, I
apologize.

>Knowing that GetQuote("IBM") fails using the file transport brings forward
>the suggestion from you that this test should be omitted.

Actually, I was just going off the prior discussion, which talked about it
failing due to xmltoday.com being offline, not due to any Axis-internal
transport issues.

I don't want the nightly tests to be dependent on any external servers --
only on processes running on the GUMP machine itself.  However, I don't see
any reliability issues with processes on the GUMP machine communicating via
localhost networking (or local filesystem, if the file transport were
reliable), and I think it's important to exercise all those tests nightly.

>Once we had the
>piece of information that the same test succeeds using http/servlet, the
>picture changes a bit, eh?

No, not really.  I still think we should keep the IBM test diked out of the
nightlies, and we should nightly run *all* other tests that can be run
using only Ant-launched-and-controlled processes.

Are we nearing clarity?
Cheers,
Rob


Mime
View raw message