FreeBSD support

[ydirson]

rtnetlink has 2 problems building on FreeBSD:

• no support for Linux namespaces and traffic control, addressed in first commit
• lack of AddressFamily::Bridge in netlink-packet-route, addressed in second commit

I originally thought about hiding namespaces and traffic control behind a ns feature, what do you think?

That second point brings me back to my original suggestion of having enum values defined once, regardless of target platform support.

[cathay4t]

@ydirson Please fix the CI failure.

Meanwhile, please include https://github.com/rust-netlink/netlink-packet-route/blob/main/.github/workflows/build.yml to this PR, so we know this project can actually compile on FreeBSD. (You may only keep x86_64-unknown-freebsd and x86_64-unknown-linux-gnu).

Thank you!

[ydirson]

@ydirson Please fix the CI failure.

Meanwhile, please include https://github.com/rust-netlink/netlink-packet-route/blob/main/.github/workflows/build.yml to this PR, so we know this project can actually compile on FreeBSD. (You may only keep x86_64-unknown-freebsd and x86_64-unknown-linux-gnu).

The libc support still has not been merged, so this will not build yet out of the fox on freebsd. Things more very slowly there, but it's not clear if I should resubmit my patch for 0.2 instead but protected by a feature flag.

[ydirson]

@ydirson Please fix the CI failure.

I'm unsure how to handle examples that can only build under a given cfg(), since examples (and tests etc) are implicitly discovered. I just cannot find any hint of cargo supporting this use-case.

I tried guarding the example code with #[cfg(any(linux, android))], but that ends with a build failure when there is no valid main(). Is the only way out to have an empty main() for unsupported configuration?

[cathay4t]

@ydirson Please fix the CI failure.

I'm unsure how to handle examples that can only build under a given cfg(), since examples (and tests etc) are implicitly discovered. I just cannot find any hint of cargo supporting this use-case.

I tried guarding the example code with #[cfg(any(linux, android))], but that ends with a build failure when there is no valid main(). Is the only way out to have an empty main() for unsupported configuration?

Instead of #[cfg(any(linux, android))], can we do #[cfg(not(freebsd))], the former one ruled out some os(e.g. fuchsia) from compiling this crate which might works well before.

[ydirson]

Instead of #[cfg(any(linux, android))], can we do #[cfg(not(freebsd))], the former one ruled out some os(e.g. fuchsia) from compiling this crate which might works well before.

Done - that had to be #[cfg(target_os = "freebsd")], though

[cathay4t]

@ydirson The 0.14.1 release contains this PR now. Thanks for the contribution!