SOCK_NONBLOCK should re-export libc::SOCK_NONBLOCK

[oh-its-jimjam]

I believe there is a bug for SOCK_NONBLOCK as the value is not the same for all architectures so will break if it's used on a non-x86 architecture - like MIPS. For example the flag value is 128 on MIPS.

[oh-its-jimjam]

I just noticed that you have a dependency to libc. I submitted a PR to add SOCK_NONBLOCK and SOCK_CLOEXEC flags to libc so you can potentially reference them and don't have to split your code into different architectures.

I'm happy to submit a PR for it when it's available in libc.

[kamalmarhubi]

Thanks for taking a look. Bringing in the value from libc is probably the right thing to do.

[oh-its-jimjam]

Alex has already merged the change into libc so I can submit a PR to you - hopefully tonight if I get time.

[kamalmarhubi]

There's no release out with it yet, from what I could see, but thanks for tracking it!

[kamalmarhubi]

@jamperry I just checked, both your commits made it into libc 0.2.6 which is out.

[Susurrus]

@jamperry Any interest in filing a PR for this? Seems pretty straightforward since the constant is properly defined in libc.

[Susurrus]

I believe this is resolved as of f067e73.

[kristate]

@kamalmarhubi @Susurrus
Does f067e73 account for target_os = "macos" target_os = "ios" ?

On MacOS:

error[E0599]: no associated item named `SOCK_NONBLOCK` found for type `nix::sys::socket::SockFlag` in the current scope

[kristate]

@kamalmarhubi @Susurrus
My apologies, it seems that SOCK_NONBLOCK does not exist in darwin.
I will stick to using fcntl() and friends, thanks.

[kristate]

@kamalmarhubi @Susurrus
Now I'm curious: nix::sys::socket::socket provides a fourth parameter that calls fcntl() but seemingly does not support macos. Opened an Issue (#861) for clarification. thanks.