Avoid a doc alias attribute error.

[ratmice]

"error: doc alias attribute expects a string: #[doc(alias = "a")]"
This error can be seen in the tests at
rust-lang/rust#72780

[koutheir]

Thank you for your efforts!

Do you see this error when building with a nightly toolchain?

[ratmice]

Nope, it was with

$ rustc --version
rustc 1.51.0 (2fd73fabe 2021-03-23)

[koutheir]

So is this for compatibility with that version of Rust?

[ratmice]

Yes, I did a little bisecting rustup and it compiles as low as 1.48 now, but cargo test still requires a newer version though. Tests get or-patterns syntax is experimental, but I saw this was from a recent commit for reducing the MSRV and figured it might reduce further. I don't in particular have any issues doing an upgrade to a more recent compiler (So no harm if you want to close this and stay with where it is at).

error[E0710]: an unknown tool name found in scoped lint: `rustdoc::invalid_html_tags`

[koutheir]

Again, thank you for your investigation!

I simply wanted to get a clearer picture on what is going on. Your changes make sense to me.

[ratmice]

I believe i've fixed all the testsuite errors,
which were all in dev-dependencies: rsix, and unsafe-io depended on by socketpair.
Those have been merged and whenever they hit crates.io and socketpair upgrates hopefully testing will just work.

[koutheir]

I don't know which errors you're referring to. I updated the socketpair and published version 0.2.4 to crates.io 4 hours ago. See this commit. Tests build fine on Rust 1.55.0.

[ratmice]

Sorry, could have explained better, as of this commit on Rust 1.51.0, cargo build works, but cargo test still fails to compile tests, so while it compiles with 1.51 and 1.48. I haven't actually managed to test it with these compilers.

sunfishcode/io-extras#10
bytecodealliance/rustix#71

[koutheir]

I see your point. I didn't observe these errors, as I only build and test the latest stable.

The issue comes from a dependency, so I think it's reasonable to wait for them to fix this issue. If that is not fixed within a reasonable duration, then I might consider removing the dependency and reimplementing what I use from it (currently the security_context_of_peer_socket test), or forking and maintaining the dependency. Also, if compatibility with Rust 1.51.0 becomes a non-issue, then this problem also becomes out of scope. Let's wait and see.

If this problem is very important to you, then feel free to create a new issue about this crate's tests not building on Rust 1.51.0, so you can track the issue.

[ratmice]

Not of any importance, I just wanted to document the current situation, since I may not, or probably won't notice once the crates are updated.