I’d like to propose adjusting the TCK
process for a service release in response to a challenge.
In the recent Batch TCK v2.1.1 service
release:
https://github.com/eclipse-ee4j/batch-tck/releases/tag/2.1.1
we addressed a challenge by modifying a
test, and “loosening” the validation logic.
Now, I think a straightforward reading of
the TCK process
https://jakarta.ee/committees/specification/tckprocess/
suggests this isn’t allowed… you can only
exclude the test or workaround.
If it weren’t for Scott Marlow voicing his
support for the idea,
https://github.com/eclipse-ee4j/batch-tck/issues/49#issuecomment-1145335955
I’d have assumed we couldn’t do this.
The key point was ensuring that, with this
modification, any implementation passing the old test should
also pass the new, modified test.
Abstracting over the details… instead of
verifying that variable “a” was set to value “XYZ” we allowed
for either “a” OR “b” to be set to value “XYZ”.
I am in favor of the change, since it
allows us to avoid throwing away a useful test ( and we did go
ahead and do the change and v2.1.1 release).
But I’m circling back here with this email,
and this PR I drafted on the subject:
https://github.com/jakartaee/jakarta.ee/pull/1496