axis-java-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Russell Butek" <bu...@us.ibm.com>
Subject WSDL tests as unit tests? (was: Re: Crisping up the interface (was: Undo and chains))
Date Wed, 02 Jan 2002 14:12:22 GMT
Sam wrote:

> An example: I think that the SequenceTest.wsdl as currently written is an
> excellent functional test case.  Given its existence, perhaps it would be
> just as effective if the majority of the remainder of the wsdl tests were
> written as unit tests (i.e., didn't require a server).

Making WSDL tests unit tests brings up a philosophical discussion.  Do we
want to make sure the generated bindings look like we expect them to look
(no server) or do we make sure the generated bindings function (need a
server)?

I worked on an emitter project (idl2xxx) where the IDL tests did not
require a server.  We compared the generated files to a master set of what
we expected the generated files to look like.  Every time the IDL-to-XXX
language bindings changed we had to change the tests.  This was OK because
the specs were fairly solid and we rarely had to change the tests.

Since JAX-RPC is still in flux, the generated files are always changing.
We don't have a master set of bindings because at this point we'd be
changing it all the time.  So we're not testing for looks, we're testing
for function.  Personally, testing for function is a better test.  Some of
the WSDL tests have pointed out runtime problems.



Russell Butek

butek@us.ibm.com


Mime
View raw message