feat: add cygwin support

[Berrysoft]

What does this PR do

This PR adds cygwin target support.

Tracking:

• Add cygwin support rust-lang/libc#4279
• Cygwin: Add PTY and group API rust-lang/libc#4309
• cygwin: fix member types of statfs rust-lang/libc#4324
• Add std support to cygwin target rust-lang/rust#137621

Checklist:

• I have read CONTRIBUTING.md
• I have written necessary tests and rustdoc comments
• A change log has been added if this PR modifies nix's API

[Berrysoft]

The CI is broken:(

[SteveLauC]

The CI is broken:(

Weird, let me take a look

[Berrysoft]

Weird, let me take a look

Hello, @SteveLauC , if you are still looking at this:) I think it's because the latest version of once_cell updates its MSRV to 1.70, larger than the MSRV of nix, 1.69.

[SteveLauC]

I think it's because the latest version of once_cell updates its MSRV

So was this newer version of once_cell updated by the libc crate?

[Berrysoft]

So was this newer version of once_cell updated by the libc crate?

I don't think so. libc doesn't depend on once_cell. It is tempfile.

This repo doesn't commit the Cargo.lock. In the CI, cargo tries to find the latest version that satisifies the Cargo.toml, without considering the rust-version field in Cargo.toml (this problem has been solved with resolver 3, which is the default resolver in edition 2024).

However, it seems that once_cell published a new release recently to decrease the MSRV to 1.65. So a re-run of CI should solve the problem:)

[SteveLauC]

Just re-ran the CI, and they failed because SOCK_PACKET was deprecated by rust-lang/libc#4267, but since it is used explicitly for testing so I think we can #[allow(deprecated)] here

[SteveLauC]

Weird, the FreeBSD CI was still using once_cell 1.21.0 🤔

https://github.com/nix-rust/nix/pull/2610/checks?check_run_id=38692242745

[Berrysoft]

Weird, the FreeBSD CI was still using once_cell 1.21.0

Failed 5 days ago. Not today.

[SteveLauC]

Failed 5 days ago. Not today.

Hmm, sorry for missing that:D

[Berrysoft]

Thank you @SteveLauC , now the CI passes again.

[SteveLauC]

Sorry for the late review, the code part looks good. For the tests, some of them are "ignored" even though the corresponding symbols are available, I would like to know why, and if possible, it would be great to comment on the reason there, which would help maintenance a lot! :)

[SteveLauC]

Let me fix the unrelated nightly clippy warning

[SteveLauC]

Thanks

[ognevny]

@Berrysoft why unistd::sync was not added in this PR? it seems that it's present in libc

[Berrysoft]

I think I have checked most features, but sorry if I missed something...