unistd: make getpeereid take AsFd rather than a raw FD

[woodruffw]

I noticed that the sockopt APIs now take AsFd, so I figured this API should be made to do the same 🙂

AsFd returns a BorrowedFd that is correct by construction, meaning that this API rejects more error states at the type level (such as passing in fd = -1). As such, I've removed the test_getpeereid_invalid_fd testcase, since its error state is no longer reachable from Rust users of this API.

[woodruffw]

NB: Unrelated formatting changes in this file (made automatically). Happy to revert if you'd prefer to leave them out.

[SteveLauC]

Ref #1750

[SteveLauC]

BTW, the module unistd is partially covered in PR #1863, and it does have some work that is not done yet, see this comment, would you like to finish that part in this PR, which would be great!

[woodruffw]

would you like to finish that part in this PR, which would be great!

It looks like there are some unresolved design questions in that PR's discussion, so I don't want to make this changeset harder to merge by adding those in.

Does that seem reasonable?

[SteveLauC]

Sorry for being unclear, I was suggesting finishing the part that is missing from that PR (i.e., the following functions):

1. pub fn tcgetpgrp(fd: c_int) -> Result<Pid>
2. pub fn tcsetpgrp(fd: c_int, pgrp: Pid) -> Result<()>
3. pub fn fpathconf(fd: RawFd, var: PathconfVar) -> Result<Option<c_long>>
4. pub fn ttyname(fd: RawFd) -> Result<PathBuf>
5. pub fn getpeereid(fd: RawFd) -> Result<(Uid, Gid)>

It should be okay to simply replace the RawFd (or c_int) with a trait generic argument, just like what's done in this PR:)

[woodruffw]

Sorry for being unclear, I was suggesting finishing the part that is missing from that PR (i.e., the following functions):

Oh, gotcha! Yeah, I can definitely do that; sorry for the confusion on my part!

[woodruffw]

Done -- all 5 of those APIs now take an AsFd bounds instead, and I've update the tests.

[woodruffw]

Gentle ping on this, now that #2132 is semi-resolved 🙂

[asomers]

LGTM.