The module-infos would not be vendor-specific. They would be specification defined API: The module name is API as are the packages exported by the module. This module name would be required by other modules which want to use the API packages. The specification would also need to define how the API connects to implementations of the API via the module's provides/uses (ServiceLoader) mechanism since applications themselves should not have to couple to a specific implementation of the API.
which an implementation of the API must "provides":
provides jakarta.ws.rs.client.ClientBuilder with com.acme.client.ClientBuilderImpl
So this information is part of the specification that API users and API implementors must rely upon when using the API in module mode. But this is only relevant when running in module mode, which does not include platform runtimes at this time. So from a platform TCK point of view, it does not make sense to test the module-info information. But some build-time testing would be useful to ensure the module-info class in the generated API jar conforms to the specification.
--
BJ Hargrave Senior Technical Staff Member, IBM // office: +1 386 848 1781 OSGi Fellow and OSGi Specification Project lead // mobile: +1 386 848 3788 hargrave@xxxxxxxxxx
----- Original message ----- From: "David Blevins" <dblevins@xxxxxxxxxxxxx> Sent by: "jakartaee-platform-dev" <jakartaee-platform-dev-bounces@xxxxxxxxxxx> To: "jakartaee-platform developer discussions" <jakartaee-platform-dev@xxxxxxxxxxx> Cc: Subject: [EXTERNAL] Re: [jakartaee-platform-dev] module-info tests Date: Fri, Oct 8, 2021 13:39
Catching up on this thread and trying to wrap my head around the what exactly it is we're proposing.
On the face of it it sounds:
- we want to add module-infos to the API jars produced out of the Jakarta spec projects
- such module-infos will be considered vendor-specific, non-portable, and not be part of a standard
- there will be no requirements or tests for them affecting certification even if an implementation does support JPMS
- there is no action that needs to be taken by anyone else who produces API jars
I'm fairly certain much of the above is wrong, but this will help me contrast/understand a bit better what is being proposed.
A quick question: by "validating on the spec api jars", do you mean manually eyeballing it? It sounds good. However, we need to agree on where the module-info class should end up.
Please see the other thread on the desired location of module-info.class.
But what has been argued by advocates of adding the module-info is that there are runtimes supporting JPMS and Jakarta API jars that cannot make use of tools like jlink because that does not work with automatic modules. To me a compromise is to add the module-info.class, validate it in the spec api jars, but don't test for it via signature or TCK tests.
I am with Tom on this. If we just superfluously verify such a file exists in the api jar, we don't mandate it will be used. In other words, JPMS is not forced. I don't see the point of adding a tck or even adding this class at all. We either drop it or force it. For forcing it, it will need more discussion as we have to wait and see how useful it is. If some specs wants to add it so that some impls can experiment with it, it is fine. However, don't add tcks before we force JPMS.
That is essentially my point. At runtime we have no requirement to run in JPMS for the containers. Containers that do not run in JPMS should not be forced to provide module-infos in their implementation at runtime. It would provide no value to them nor the applications and I argue it could limit some possible innovation at the runtime implementation level.
Here is were my inexperience with the TCK shows. Does the TCK signature tests run outside of the container directly against some set of JARs provided by the implementation?
If your point is just about that we should only test the API jars from the specification project release for the module-info, and not in general during compatibility testing because we don't have a requirement for JPMS in the containers, that is valid, and one I probably agree with.
Implementations are not required, but if they do, then how do you certify? Right now the TCK signature tests look to the jars/content provided by the implementation under test. If they have their own versions of the API jars, they need to pass the same requirements as the specification project producing the API jars.
I did not intent to suggest that apps are only allowed to be compiled against the API JARs from Jakarta projects. I was asking if implementations are required to provide to such JARs for development purposes? Don't get me wrong, implementations should be allowed to provide such JARs to allow them to provide various developer experiences as they see fit. Open Liberty certainly does provide such JARs for developers to compile against also. I do think such JARs should conform to the module-info requirements from Jakarta. But I don't think the specification requires an implementation to provide such JARs for the developer to compile against.
If implementations are not required to provide API JARs for compilation/development purposes then I do not see the point of the TCK testing for module-info classes against the implementation. On the other hand the Jakarta build of the API JARs should contain a test that validates they are providing the correct things.
_______________________________________________ jakartaee-platform-dev mailing list jakartaee-platform-dev@xxxxxxxxxxx To unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/jakartaee-platform-dev
