1.67 release broken on illumos when stdout is not a tty

[davepacheco]

Problem

After installing toolchain 1.67 on illumos, all cargo invocations with stdout not being a tty fail with:

$ cargo version > /dev/null
thread 'main' panicked at 'unexpected error from isatty: 0', /cargo/registry/src/proxy-git.cwkhome.fun-1ecc6299db9ec823/rustix-0.36.3/src/backend/libc/termios/syscalls.rs:141:20
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

It's fine if stdout is a tty:

$ cargo version
cargo 1.67.0 (8ecd4f20a 2023-01-10)

This is unfortunately a showstopper for us because stdout is not a tty in CI, so we can't use this toolchain in CI.

Steps

On an illumos system:

1. rustup toolchain update stable
2. cargo version (to make sure you got 1.67)
3. cargo version > /dev/null (should reproduce the problem)

Possible Solution(s)

We believe this is bytecodealliance/rustix#467, which was introduced to Cargo by 48895b1. This has been fixed upstream.

Notes

No response

Version

$ cargo version --verbose
cargo 1.67.0 (8ecd4f20a 2023-01-10)
release: 1.67.0
commit-hash: 8ecd4f20a9efb626975ac18a016d480dc7183d9b
commit-date: 2023-01-10
host: x86_64-unknown-illumos
libgit2: 1.5.0 (sys:0.16.0 vendored)
libcurl: 7.86.0-DEV (sys:0.4.59+curl-7.86.0 vendored ssl:OpenSSL/1.1.1q)
os: illumos helios-1.0.21408 [64-bit]

(thanks @jgallagher for having found this before, fixed it upstream, and identified it here)

[weihanglo]

Sorry for this bug affecting you. We could had included this fix rust-lang/rust#105511 but unfortunately it got merged after 1.67 branched from master to beta. And the Cargo team wasn't aware of this situation 😞.

I don't really know about illumos, so don't know how to help. As always, if permitted, I would recommend every project runs tests on beta channel in CI pipelines.

[Brooooooklyn]

Same on FreeBSD: https://github.com/Brooooooklyn/Image/actions/runs/4032326915/jobs/6932135051

[weihanglo]

@Brooooooklyn Sorry for that. Could you provide more context? I don't really know how to interpret the job failure log.

[jgallagher]

The build step from the job failure shows the same panic message as above (although with a different errno):

Type Error: Could not parse the Cargo.toml: Error: Command failed: cargo metadata --format-version 1 --manifest-path "/Users/runner/work/Image/Image/packages/binding/Cargo.toml"
thread 'main' panicked at 'unexpected error from isatty: 45', /cargo/registry/src/proxy-git.cwkhome.fun-1ecc6299db9ec823/rustix-0.36.3/src/backend/libc/termios/syscalls.rs:141:20
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

thread 'main' panicked at 'unexpected error from isatty: 45', /cargo/registry/src/proxy-git.cwkhome.fun-1ecc6299db9ec823/rustix-0.36.3/src/backend/libc/termios/syscalls.rs:141:20
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

[weihanglo]

BTW, could anyone check if the bug is fixed under the current beta?
(probably cargo 1.68.0-beta.2 (0762e3bbd 2023-01-30))

[jgallagher]

It is fixed in the current beta (and was also fixed in the initial beta). On an illumos system:

john@helios1:~$ cargo -V
cargo 1.67.0 (8ecd4f20a 2023-01-10)
john@helios1:~$ cargo -V >/dev/null
thread 'main' panicked at 'unexpected error from isatty: 0', /cargo/registry/src/proxy-git.cwkhome.fun-1ecc6299db9ec823/rustix-0.36.3/src/backend/libc/termios/syscalls.rs:141:20
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
john@helios1:~$ cargo +beta -V
cargo 1.68.0-beta.2 (0762e3bbd 2023-01-30)
john@helios1:~$ cargo +beta -V >/dev/null
john@helios1:~$

[weihanglo]

Thank you for testing the fix on beta! I am going to close this, and sorry for the inconvenience.