OpenBSD socket address is padded with zeroes

[notgull]

I tried this code on my Linux machine:

use std::os::unix::net::UnixListener;

fn main() {
    let listener = UnixListener::bind("/tmp/rust-uds.sock").unwrap();
    println!("{:?}", listener.local_addr().unwrap());
}

I got this output:

"/tmp/rust-uds.sock" (pathname)

However, on OpenBSD 7, I got this output:

"/tmp/rust-uds.sock\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0" (pathname)

I expected the output to be the same as Linux: not padded with zeroes like this.

Meta

rustc --version --verbose:

$ rustc --version --verbose
rustc 1.63.0
binary: rustc
commit-hash: unknown
commit-date: unknown
host: x86_64-unknown-openbsd
release: 1.63.0
LLVM version: 13.0.0

[notgull]

This is an OpenBSD bug. I've reported it here: https://marc.info/?l=openbsd-bugs&m=170105481926736&w=2

Still, we should work around it in the standard library.

[cjeker]

Just to be clear, this is not an OpenBSD bug. OpenBSD decided to return always the full struct sockaddr_un.
This is all within the specification of getsockname imposed by POSIX.
The embedded path sun_path inside struct sockaddr_un is always a valid C string (NUL terminated).
There is therefor no need to look at the address_len argument as long as the passed sockaddr buffer is at least sizeof(struct sockaddr_un) bytes. Developers are highly encouraged to use struct sockaddr_storage to ensure there is no clipping. The reason for this behavior is to reduce the risk of bugs because of the scary getsockname spec.