Failed to compile ring under nightly Rust (1.61.0, 9f4dc0b4d 2022-03-23)

[messense]

Cross post from briansmith/ring#1469

    Checking ring v0.16.20
error: no rules expected the token `aarch64_apple`
   --> /root/.cargo/registry/src/rsproxy.cn-8f6827c7555bfaf8/ring-0.16.20/src/cpu.rs:257:13
    |
165 |     macro_rules! features {
    |     --------------------- when calling this macro
...
257 |             aarch64_apple: true,
    |             ^^^^^^^^^^^^^ no rules expected this token in macro call

error[E0425]: cannot find value `AES` in module `cpu::arm`
   --> /root/.cargo/registry/src/rsproxy.cn-8f6827c7555bfaf8/ring-0.16.20/src/aead/aes.rs:381:65
    |
381 |         if cpu::intel::AES.available(cpu_features) || cpu::arm::AES.available(cpu_features) {
    |                                                                 ^^^ not found in `cpu::arm`
    |
help: consider importing this constant
    |
15  | use crate::cpu::intel::AES;
    |

error[E0425]: cannot find value `PMULL` in module `cpu::arm`
   --> /root/.cargo/registry/src/rsproxy.cn-8f6827c7555bfaf8/ring-0.16.20/src/aead/gcm.rs:315:26
    |
315 |             || cpu::arm::PMULL.available(cpu_features)
    |                          ^^^^^ not found in `cpu::arm`

error[E0425]: cannot find value `ARMCAP_STATIC` in this scope
   --> /root/.cargo/registry/src/rsproxy.cn-8f6827c7555bfaf8/ring-0.16.20/src/cpu.rs:235:41
    |
235 |             if self.mask == self.mask & ARMCAP_STATIC {
    |                                         ^^^^^^^^^^^^^ not found in this scope

Works fine on stable Rust.

https://github.com/briansmith/ring/blob/9cc0d45f4d8521f467bb3a621e74b1535e118188/src/cpu.rs#L165-L224

[zjp-CN]

That post is written by me.
I've opened a PR briansmith/ring#1470, and made some comments there.
I'm not an expert on Rust, actually I'm an amateur programmer, and this is what I can do for now.

[ctz]

1.61.0-nightly (5f37001 2022-03-22) works so this is a very recent nightly regression.

[kaimast]

I see that too when compiling maturin

[hunjixin]

same issue on ubuntu20

[zjp-CN]

Maybe related to this issue.
Here is a snippet of macro with /// on the match side:

#[rustfmt::skip]
macro_rules! f {
    (
        /// ab
    ) => {};
}

fn main() {
    f!();
}

Pass with rustc 1.61.0-nightly (5f37001 2022-03-22), you can try this snippet via

rustup default nightly-2022-03-23
# nightly-2022-03-23-x86_64-unknown-linux-gnu unchanged - rustc 1.61.0-nightly (5f3700105 2022-03-22)

Failed with rustc 1.61.0-nightly (9f4dc0b 2022-03-23), you can try this snippet via

rustup default nightly-2022-03-24
# nightly-2022-03-24-x86_64-unknown-linux-gnu unchanged - rustc 1.61.0-nightly (9f4dc0b4d 2022-03-23)

error: unexpected end of macro invocation
 --> src/main.rs:9:5
  |
2 | macro_rules! f {
  | -------------- when calling this macro
...
9 |     f!();
  |     ^^^^ missing tokens in macro arguments

---

Surprisingly, the code can be expanded on playground with:
Expand macros in code using the nightly compiler.
1.61.0-nightly (2022-03-19 8d60bf4)

[lqd]

The piece of code looks surprising, but bisection lead to #95159

[zjp-CN]

The piece of code looks surprising, but bisection lead to #95159

@lqd

Ok, I found it 10644e0

😄

[nnethercote]

Thanks for the report. Should be an easy fix, I will get to it on Monday.

[zjp-CN]

I think 10644e0 is right to some extent.
Even in previous Rust version, /// stuff in the matcher is not so useful and will be ignored like normal comments //:

#[rustfmt::skip]
macro_rules! f {
    (
        /// doc
    ) => { println!("/// doc"); };
    (#[$m:meta]) => { println!("#[$m:meta]"); };
    () => { println!("()"); };
}

fn main() {
    f!(
        /// doc
    );
    f!();
}

With rustc 1.61.0-nightly (5f37001 2022-03-22) or other versions before, the result is:

#[$m:meta]
/// doc

With rustc 1.61.0-nightly (9f4dc0b 2022-03-23), the result is:

#[$m:meta]
()

[shepmaster]

Surprisingly, the code can be expanded on playground with:

Note that the playground's nightly is a few days old and cannot be upgraded because of how this issue affects ring ;-)

[apiraino]

Assigning priority as discussed in the Zulip thread of the Prioritization Working Group.

@rustbot label -I-prioritize +P-critical

[nnethercote]

Even in previous Rust version, /// stuff in the matcher is not so useful and will be ignored like normal comments //:

That's an interesting observation! But I'm not a language expert and I never intended to change behaviour in #95159, so I will fix this problem by going back to the old behaviour.