Build the systemconf library on all platforms

[gnu-andrew]

Currently, the makefile excludes building libsystemconf on any platform but Linux, but the JNI code in SystemConfiguration.java then unconditionally tries to load the library.

We could make this conditional on running on Linux, but we still have the problem of a native method in the class that doesn't resolve. The best solution seems to be to build the library, but provide a simple JNI_FALSE return for non-Linux systems. This also leaves the door open for an implementation on these platforms in future.

[martinuy]

One question, do our non-Linux builds (which I assume it's Windows) already include the RPM patches that, for example, introduce the SystemConfigurator.java?

[gnu-andrew]

One question, do our non-Linux builds (which I assume it's Windows) already include the RPM patches that, for example, introduce the SystemConfigurator.java?

I don't know. @akashche?

My aim with this patch is to give us a clean set of PR tests. The alternative would be to turn off Windows & Mac builds, but knowing we don't break them seems useful.

Incidentally, it's just one patch now: https://src.fedoraproject.org/rpms/java-17-openjdk/c/756a991906919de0d448abf84e9a66cf96dc6afd?branch=rawhide

[judovana]

Hello. thanx a lot. Sorry for troubeling you with this.
I had jsut enabled basic upstream qa upon fips brnaches, and that includes windows buildability.

As for downstream, I really do not know. Windows rh builds are based on srpm, but then half of the patches is gone. @stooke woud know more now.

[franferrax]

Hi @gnu-andrew, it looks good to me.

I'm wondering about how much is the benefit of including a build-stage + test-suite running on platforms whose official build don't (or shouldn't) include changes from the FIPS branches we are testing, but it doesn't harm.

[gnu-andrew]

Hi @gnu-andrew, it looks good to me.

I'm wondering about how much is the benefit of including a build-stage + test-suite running on platforms whose official build don't (or shouldn't) include changes from the FIPS branches we are testing, but it doesn't harm.

Well, you've kind of answered the question; it's ensuring they don't include the changes and still work. I think it's a good test that our changes aren't breaking things or making assumptions that the FIPS mode always operates a certain way. We've had issues raised before where we broke the scenario where FIPS was disabled on Linux. This is why the PRs now test both on and off, and I think unavailable is a good additional test as well. Prior to this change, we couldn't even build on the other platforms and so also couldn't test that the code still operated correctly.

[gnu-andrew]

Thanks for the review. Merging.