On 12 December 2016 at 16:22, Bill Burke <bburke(a)redhat.com> wrote:
On 12/12/16 8:19 AM, Stian Thorgersen wrote:
> Let's put this on hold for after 7.1 GA and make it a priority after
that.
We can't have tomcat and jetty adapters failing and regressions
happening. There's a lot of people that depend on them. Our customer
base comes from a successful community project. You don't start
undercutting community just because its not supported in product. I'm
adding them back to the build because its the right thing to do. I
shouldn't have even asked and just did it, but I just found it quite
annoying that they were removed as I knew they were removed on purpose.
You completely miss understood me. No one removed them on purpose. AFAIK
these tests have never actually ran on Travis.
No one has ever said that we are undercutting the community and that Tomcat
and Jetty is not still important. I've got absolutely no clue why you think
anyone has suggested that?!
> We are going to use Arquillian bases tests and removing the old tests as
> they have less value from a QE perspective.
Negative. You are just 100% wrong on this. I've said this over and
over and you just don't listen and its getting really annoying. There
are many tests which it doesn't matter if they run in the container or
not. Many tests it is not possible to even run as its impossible to set
up the conditions you want to happen. Many tests are functional SPI
tests and it doesn't matter at all if they run inside the app server or
not and its much easier to setup and debug
All tests matter if they run in the container! Bad imports, wrong module
definitons and bad/different versions of dependencies can all cause tests
to pass in the embedded/fake container while not working properly when
running within WildFly/EAP.
All tests that rely on the DB layer also needs to test different DB
vendors! Setting up and configuring CI is already complicated for the real
containers and that effort would have to be duplicated for a fake
environment as datasources are not available.
.
> This has been discussed many
> times now and we have agreed that this will be done a long time ago so
> please let's not reiterate this discussion again and again. If dev and QE
> are going to collaborate on the tests we need to share the approach and
QE
> has to have full functional tests.
Discussed many times and I never agreed to anything. And kept telling
you over and over that this mandate is uneccessary, hurtful to
development, and is a lot of wasted effort, but you don't want to listen.
See above
> We need to resolve this issue though so we can get access to the
> KeycloakSession from within certain tests. Ability to run adapter tests
> from within the IDE is also a nice to have, but that may be harder to
> achieve (and the effort may not be worth it, at least not for all
adapters).
There's a difference between unit tests, functional tests, and
integration tests that you are missing. Everything is not an
integration test. Its unacceptable for adapter tests to not be able to
run in the IDE. If you want to have additional non-IDE runnable
integration tests, that is fine, but to be able to develop new features
and debug most problems, running in the IDE is a must and you're just
being stubborn to state otherwise. Its quite annoying.
I agree that adapter tests should run within the IDE, but I disagree with
the approach you are suggesting. Adapter tests are even more important to
have a real environment IMO.
Bill
_______________________________________________
keycloak-dev mailing list
keycloak-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/keycloak-dev