Add GNU/Hurd support

[sthibaul]

No description provided.

[sthibaul]

error: this loop never actually loops
    --> test/sys/test_socket.rs:1305:9
     |
1305 | /         for _ in msg.cmsgs() {
1306 | |             panic!("unexpected cmsg");
1307 | |         }
     | |_________^
     |
     = help: for further information visit https://rust-lang.github.io/rust-clippy/master/index.html#never_loop
     = note: `#[deny(clippy::never_loop)]` on by default

This is unrelated. This actually looks to me like a false positive from clippy: the code does mean not to loop and just panic if we make even just one iteration.

[SteveLauC]

Hi, thanks for your contribution to nix! Unfortunately, our Cirrus CI is running out of its credits, so I manually stopped the CI for this PR.

After our migration to GitHub action, I will take a look at this PR and re-run the CI job

[sthibaul]

I have rebased on top of the github action commit

[sthibaul]

And I have now added i686-unknown-hurd-gnu as tier3 in CI. Obviously, in the meanwhile some errors have popped up, I'll have a look.

[sthibaul]

Some fixes need to be done in libc (and wait for a libc release...)

[SteveLauC]

Some fixes need to be done in libc (and wait for a libc release...)

Ok, then I will label this W-blocking-upstream

[asomers]

You'll have to rebase to fix CI.

[sthibaul]

You'll have to rebase to fix CI.

ok, done so!

[asomers]

@SteveLauC do you need to change the minimum libc version? CI's minver check won't catch an error there, if the error only affects Hurd.

[SteveLauC]

hi @sthibaul, is this PR blocked by these changes?

• libc#3430: hurd: Fix C API interface completion
• libc#3428: hurd: Complete C API interface

These changes haven't been released yet, so I guess we still need to wait for a new libc release

[SteveLauC]

Hi @sthibaul, now we are ok to use the libc dependency from git so that we don't have to wait for a libc release.

And, rebasing your branch should make the CI pass since the commit used already has these 2 PRs covered

I am sorry about those merge conflicts, we have switched to cfg alias to make our code clean, and this PR is mostly about cfg, so...

[sthibaul]

Ok, at last it does pass...

[SteveLauC]

Thanks!

[sthibaul]

AIUI it is waiting for @asomers ' ack?

[SteveLauC]

AIUI it is waiting for @asomers ' ack?

Yeah, that would be best, let's wait for asomers's review:)

[asomers]

LGTM