Possible regression with #[rustfmt::skip]

[steveklabnik]

Hi folks! I am getting users reporting a regression in the semver crates, see:

• build error on #![rustfmt::skip] steveklabnik/semver-parser#52
• semver-parser v0.10.1 build failure on stable dtolnay/semver#218

It appears that the rustfmt attributes are no longer considered "built-in"? Or am I completely mis-remembering this?

Sorry for not being 100% sure about the regression here, but I don't want this to get lost.

@rustbot modify labels: +regression-from-stable-to-stable -regression-untriaged

[ehuss]

@steveklabnik I believe this was changed in 1.44 probably via #69838. Do you happen to know which version the reporter was using? 1.44 or newer should build ok.

There are still some oddities with inner tool attributes like #74087, so it is possible it could be that.

[steveklabnik]

I don't know, but let's see if they can tell us: @dwarburt and @tarcieri (as well as @krisfris, who gave a 👍 on tony's issue), what Rust versions are you using?

[tarcieri]

The main failure I have handy is 1.41.0:

https://github.com/RustSec/rustsec-crate/runs/1318970855?check_suite_focus=true

[dwarburt]

I'm using the toolchain that the brave-browser build scripts downloads

✔ ~/git/fork/brave-browser [master|✚ 1]$ ./src/brave/build/rustup/0.1.4/bin/rustc --version
rustc 1.47.0 (18bf6b4f0 2020-10-07)
✔ ~/git/fork/brave-browser [master|✚ 1]$ 

[djc]

I ran into this with 1.34.0 (which is the MSRV for err-derive, which used to depend on semver-parser via rustc-version).

[steveklabnik]

It appears this is actually my fault: steveklabnik/semver-parser#52 (comment)

sorry about the noise.