Fix nightly CI

[JurajSadel]

Gets rid of

error: unexpected `cfg` condition name: `test`

closes #4384

[bugadani]

Do we know why exactly this warning is triggered? I remember debugging the same not too long ago, but I don't remember what came of it :/

[MabezDev]

Hmm, I'm not sure this is the real fix. Isn't this likely related to the fact that when building no_std the compiler/cargo won't emit the check cfg(test)? cfg(test) usage also needs to be cfg_attr'd on no_std?

[JurajSadel]

I spent some time analyzing this and here is the summary:

1.) This is on by default

rustc -W help

....
unexpected-cfgs  warn     detects unexpected names and values in `#[cfg]` conditions
....

2.) cargo build/check which is ran in cargo xtask ci --toolchain nightly doesn't know about test - I guess it's because of test is not part of known names?

Related from https://doc.rust-lang.org/rustc/check-cfg.html:

Starting with 1.85.0, the test cfg is considered to be a "userspace" config despite being also set by rustc and should be managed by the build system itself.

So, if I'm not missing something crucial here, I don't know how to deal with the errors besides what this PR does (adding println!("cargo::rustc-check-cfg=cfg(test)"); might be overkill since not all packages have build.rs)

EDIT: when it started to happen: rustc 1.92.0-nightly (f04e3dfc8 2025-10-19)
cc @bugadani @MabezDev

[bjoernQ]

this is a bit wordy but probably fine

[bjoernQ]

I guess this won't get any better the longer we wait - we should investigate this further and eventually come back to the unexpeced-cfg "workaround" but it won't harm to have it

[bugadani]

I don't really understand why this started happening now (the relevant changes were introduced in ~February), and why only on Linux - I can't reproduce any of the warnings (except for the actualy clippy lint) on Windows. Having this workaround makes me sad, and we should revisit in the future in case it can be removed.