We talked briefly in the Platform TCK call [1] this week about where Platform TCK tests are stored versus where Standalone TCK tests are stored in the Platform TCK repo.
Currently, we have a source folder structure [2] that mixes all of the Platform/Standalone TCK tests together. So for example, the Standalone Servlet test sources are all under the `servlet` folder [3].
Note that we need to address [4] by adding now missing Servlet TCK tests that meet the Platform requirements for Servlet.
To help the discussion, here are some things to note about the Standalone Servlet TCK and its relationship with the Platform TCK tests for Servlet.
1). The Standalone Servlet TCK might be moved to the Servlet SPEC git repository.
2). The Platform TCK tests for validating the Platform requirements for Servlet could extend/depend on the Standalone Servlet TCK tests.
3). Any Platform TCK tests that depend on Standalone Servlet TCK tests may be broken by service releases for the Standalone Servlet TCK in the future which would create a situation where the Platform TCK would need to release a service release for working with the Servlet TCK service release. :)
4). Regardless of where Standalone TCK source is maintained, we need a plan for how Standalone TCK source can be frozen when their relevant Specification is released as final.
Questions:
Should the Platform TCK use separate artifacts for producing Standalone TCK tests versus Platform TCK tests? If yes, what change should we make to the test directory structure?
Some possible options:
A. Separation of different Platform TCK test buckets:
servlet/platform/
servlet/standalone/
jpa/platform/
jpa/standalone/
B. Single group for Platform TCK tests (via new root platform folder), other test folders are mostly for standalone TCKs and won't change much from what we have today in the refactor branch:
servlet/
jpa/
platform/
C. Something else?
Scott
