ICE in rustc 1.59.0-beta.2

[coriolinus]

Code

src/lib.rs

#[derive(Debug, PartialEq, Eq)]
pub enum Classification {
    Abundant,
    Perfect,
    Deficient,
}


pub fn classify(num: u64) -> Option<Classification> {
    unimplemented!("classify {}", num);
}

tests/perfect-numbers.rs

use perfect_numbers::{classify, Classification};


macro_rules! tests {
    ($property_test_func:ident {
        $( $(#[$attr:meta])* $test_name:ident( $( $param:expr ),* ); )+
    }) => {
        $(
            $(#[$attr])*
            #[test]
            fn $test_name() {
                $property_test_func($( $param ),* )
            }
        )+
    }
}


fn test_classification(num: u64, result: Classification) {
    assert_eq!(classify(num), Some(result));
}


#[test]
fn basic() {
    assert_eq!(classify(0), None);
}


tests! {
    test_classification {
        #[ignore] test_1(1, Classification::Deficient);
        #[ignore] test_2(2, Classification::Deficient);
        #[ignore] test_4(4, Classification::Deficient);
        #[ignore] test_6(6, Classification::Perfect);
        #[ignore] test_12(12, Classification::Abundant);
        #[ignore] test_28(28, Classification::Perfect);
        #[ignore] test_30(30, Classification::Abundant);
        #[ignore] test_32(32, Classification::Deficient);
        #[ignore] test_33550335(33_550_335, Classification::Abundant);
        #[ignore] test_33550336(33_550_336, Classification::Perfect);
        #[ignore] test_33550337(33_550_337, Classification::Deficient);
    }
}

Meta

• Occurs with beta-x86_64-unknown-linux-gnu - rustc 1.59.0-beta.2 (98a701b42 2022-01-19)
• Command: cargo test --quiet --no-run here
• Executed in CI environment; log here.
• Not encountered with rustc 1.58.0 (02072b482 2022-01-11)

Error output

thread 'rustc' panicked at 'index out of bounds: the len is 11 but the index is 14', compiler/rustc_query_impl/src/on_disk_cache.rs:726:18
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

error: internal compiler error: unexpected panic

note: the compiler unexpectedly panicked. this is a bug.

note: we would appreciate a bug report: https://github.com/rust-lang/rust/issues/new?labels=C-bug%2C+I-ICE%2C+T-compiler&template=ice.md

note: rustc 1.59.0-beta.2 (98a701b42 2022-01-19) running on x86_64-unknown-linux-gnu

note: compiler flags: -C embed-bitcode=no -C debuginfo=2 -C incremental

note: some of the compiler flags provided by cargo are hidden

query stack during panic:
#0 [eval_to_allocation_raw] const-evaluating + checking `basic::promoted[0]`
#1 [optimized_mir] optimizing MIR for `basic`
end of query stack
error: could not compile `perfect_numbers`
error: build failed

No backtrace as yet because I don't have the beta version installed locally.

[5225225]

Looks like a form of incremental compilation bug, seeing as I can't reproduce it after installing beta on my machine, so it's not deterministic in any case.

See: #93029 , looks to have been already fixed, just the fix isn't on beta. Error message is the same, in any case.

[coriolinus]

Incremental compilation makes some sense; we encounter it in the middle of a fairly large build of a number of semi-related projects. Glad to hear it's already fixed! Any idea how to fix CI, or is the best bet to wait until the version bumps in a month?

[5225225]

Running cargo clean (maybe with -p for just the package you're compiling, maybe the whole thing is needed) seemed to work for the last incremental compilation issues.

Would slow down compilation a fair bit though. But I guess it would be bearable until the fix goes out. I'd start off with just cleaning your own package, and then if it still ICEs, just start running cargo clean and watching your CPU waste time rebuilding the packages over and over :D

(but no, I don't have any better advice than cargo clean -p <package> before every build)

[LoganDark]

Also experiencing this error. Happens whenever I do a git checkout to a slightly different version of the same overall codebase. Annoying.

[workingjubilee]

Closing as a duplicate of #94124
Or that's a duplicate of this? Either way. Two issues enter, one issue leaves.