Hi Kohsuke,
Thanks for your suggestions. That sounds fine to me. I'll work on making
that change. I was looking at the version-checking code... the harness
already accepts 'applicable versions' for the test and for the indvidual
clients. But no comparisons are ever done. That's easy to add.
I'm not clear where to pull the version info for the code being tested
(jaxws/wsit). How do you derive that with JAXB? Is it passed in on
test command line or from a file? Something in META-INF?
thanks,
Ken
Kohsuke Kawaguchi wrote:
> Ken Hofsass wrote:
>
>> Counting it as a failure kind of defeats the purpose of skipping the
>> test. The tests already fail or error. I want to make sure that the
>> WSIT/JAXWS developers who are required to run these tests before
>> check-in can assume all tests that get run are expected to pass...
>> which has not been the case thus far... I think that expecting a
>> developer to go look at hudson to make sure he/she is getting the
>> same number & kind of failures/errors would likely cause more
>> problems than skipping tests that are known to be broken.
>
>
> OK, in that case I propose my plan B, which is to exclude the test
> only from the current version. We already support version markers, right?
>
>> I think it would be really handy if we could separate out the
>> count/list of skipped tests... either have JUnit keep track of them
>> (somehow) or add code to the harness itself to track & record them.
>> Does something like that sound acceptable?
>
>
> Since we build on JUnit for test execution, we really can't do
> anything that goes against JUnit's notion of tests. JUnit does not
> have "not-run" status for a test --- a test either succeeds or fails,
> so I don't think we can do things like that.
>
>