[
Date Prev][
Date Next][
Thread Prev][
Thread Next][
Date Index][
Thread Index]
[
List Home]
Re: [microprofile-wg] MP Telemetry TCK Challenge Discussion
|
Hi Jason,
this is a valid challenge in my opinion
and we need to clarify this in the specific case of MP Telemetry
and for MicroProfile umbrella and component specs in general.
This issues was a reported Release
Review finding for MP Telemetry 1.1 already and the reason for me
to vote against the release on behalf of iJUG:
I put the topic on the MP Telemetry
call agenda for next Monday, it would be nice if you and/or your
team can join us there.
To prevent things like this in the future and clarify optional TCK
test design, architectural constraints or organisational aspects
we also could discuss this on the next MP Technical call if you
like - besides here offline of course.
I put some comments in two issues. The
short version:
The current TCK setup requires to
manually deactivate tests and remove dependencies, meaning
manually editing the TCK to be able to run it successfully on a
valid implementation - this must be changed, because this is an
architecture constraint and default behaviour violation!
Best,
Jan
Am 08.11.23 um 22:05 schrieb Jason Lee
via microprofile-wg:
Yesterday, I (with my red fedora firmly in place) opened a
challenge against one set of tests in the MP Telemetry TCK: https://github.com/eclipse/microprofile-telemetry/issues/137
The initial challenge was filed because the test itself is
testing functionality that depends on Jakarta EE specs
(Concurrency and Servlet) that are not listed as required by the
platform spec
(https://download.eclipse.org/microprofile/microprofile-6.1/microprofile-spec-6.1.html#microprofile6.1).
It is our understanding that sub specs can not expand that list
of required specs, so the dependency of these tests (and the
functionality their presence implies) is impermissible. After
more discussion, we realized -- unless we've missed something --
that the tests also cover functionality that is specified in
neither the MP Telemetry nor the OpenTelemetry specifications,
so they're testing non-specified functionality, making them
doubly inappropriate. We contend, then, that these tests should
be removed from the TCK.
Emily responded that these tests are optional, so we can simply
ignore the tests and still certify, and that's certainly our
plan. However, optional or not, these tests -- as a matter of
"legality" -- ought not be there, optional or not, for
implementers and integrators to have to deal with. Furthermore,
the assertion that the tests are optional is also somewhat
suspect, based on the language in the TCK itself
(https://github.com/eclipse/microprofile-telemetry/blob/main/tracing/tck/README.adoc#declaring-the-tests-to-run):
Although support for JAX-RS server async
programming models is not optional, these tests depend on
Jakarta Concurrency because they use ManagedExecutorService
.
If you are testing in an environment which does
not provide Jakarta Concurrency, you should exclude the optional-jaxrs-tests
group.
Since WildFly does provide Jakarta Concurrency, it
would seem that these test are not, in fact, optional.
(Our OpenTelemetry integration is done using a shared library --
smallrye-opentelemtry -- that can not add the Jakarta
dependencies need to properly implement support for the
functionality under test. Addressing that would require a fair
bit of re-work that we'd prefer to avoid given the optional
nature of the requirement).
What we, the WildFly team, would like from this group is some
clarity on these tests, specifically whether or not their
presence, even if optional, is permissible given the concerns
above.
Thanks!
_______________________________________________
microprofile-wg mailing list
microprofile-wg@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://www.eclipse.org/mailman/listinfo/microprofile-wg