Use SmallRye FFM to reliably get host name

[dmlloyd]

This is a quick example of how SmallRye FFM can be used. In this case, we have always relied on complicated heuristics to determine the system host name because Java has no API to access this information. But with SmallRye FFM, we can just make the native call directly if the JVM allows it.

This should not necessarily be considered for merging, at least until we have solutions for both Quarkus and WildFly to register native access. But eventually, doing something like this could replace most of the heuristics that we currently employ (like probing environment variables).

Note that while this is a large change, much of the change consists of things that should be moved to a parent POM: working around impsorter bugs, establishing common versions. Also, there are some legitimately missing tests in the net module relating to host name (it's hard to really test in general, but we can at least make sure we don't explode).

Edit: the Windows build doesn't presently work; this may be a problem on the SmallRye FFM side, but it can be ignored for now.

[Ladicek]

I understand that on classic JVM, only one of the native methods will ever get called, so only one of them will get linked, but with GraalVM native image, it seems to me you rely on OS.current() to be constant-propagated and therefore the unused native method being dead-code-eliminated and so not linked as well?

[dmlloyd]

It's okay if it did try to link it though.

[Ladicek]

But it wouldn't succeed then? The bootstrap method would, IDK, probably throw an exception? I don't know how native image would behave in that case.

Well, more of a theoretical concern. I was just wondering.

[dmlloyd]

The internal function resolution (via SymbolLookup) would return an empty Optional (or maybe not; it doesn't really matter), and if so, the bootstrap method would return a stub method which in turn throws an exception. The stub method would then be optimized away. (Assuming that the optimization happens after the bootstrap methods).

[Ladicek]

the bootstrap method would return a stub method which in turn throws an exception

Ah, clever! Nice!

[dmlloyd]

Hey, it works now. Cool.

[dmlloyd]

Updated to use the new @Lib annotation.

[dmlloyd]

Updated to reflect security manager removal. Also I realized that we wouldn't need the substitutions here anymore either with this change, so I removed that as well; a net reduction of around 45 lines of code in total compared to before.

[dmlloyd]

This PR could actually go in now, if we wanted to do it.

[dmlloyd]

I think I'll just go with it. We can always peel it out if it causes a problem.