Compile-time regression with mir-opt-level=2

[RalfJung]

The following rustc floating point test case used to work just fine with -Zmir-opt-level=3, but now even just calling rustc -Zmir-opt-level=2 on it diverges:

#![feature(track_caller, stmt_expr_attributes)]

// Poor-man's black-box
#[inline(never)]
fn black_box<T>(x: T) -> T { x }

macro_rules! test {
    ($val:expr, $src_ty:ident -> $dest_ty:ident, $expected:expr) => (
        // black_box disables constant evaluation to test run-time conversions:
        assert_eq!(black_box::<$src_ty>($val) as $dest_ty, $expected,
                    "run-time {} -> {}", stringify!($src_ty), stringify!($dest_ty));

        {
            const X: $src_ty = $val;
            const Y: $dest_ty = X as $dest_ty;
            assert_eq!(Y, $expected,
                        "const eval {} -> {}", stringify!($src_ty), stringify!($dest_ty));
        }
    );

    ($fval:expr, f* -> $ity:ident, $ival:expr) => (
        test!($fval, f32 -> $ity, $ival);
        test!($fval, f64 -> $ity, $ival);
    )
}

macro_rules! common_fptoi_tests {
    ($fty:ident -> $($ity:ident)+) => ({ $(
        test!($fty::NAN, $fty -> $ity, 0);
        test!($fty::INFINITY, $fty -> $ity, $ity::MAX);
        test!($fty::NEG_INFINITY, $fty -> $ity, $ity::MIN);
        // These two tests are not solely float->int tests, in particular the latter relies on
        // `u128::MAX as f32` not being UB. But that's okay, since this file tests int->float
        // as well, the test is just slightly misplaced.
        test!($ity::MIN as $fty, $fty -> $ity, $ity::MIN);
        test!($ity::MAX as $fty, $fty -> $ity, $ity::MAX);
        test!(0., $fty -> $ity, 0);
        test!($fty::MIN_POSITIVE, $fty -> $ity, 0);
        test!(-0.9, $fty -> $ity, 0);
        test!(1., $fty -> $ity, 1);
        test!(42., $fty -> $ity, 42);
    )+ });

    (f* -> $($ity:ident)+) => ({
        common_fptoi_tests!(f32 -> $($ity)+);
        common_fptoi_tests!(f64 -> $($ity)+);
    })
}

macro_rules! fptoui_tests {
    ($fty: ident -> $($ity: ident)+) => ({ $(
        test!(-0., $fty -> $ity, 0);
        test!(-$fty::MIN_POSITIVE, $fty -> $ity, 0);
        test!(-0.99999994, $fty -> $ity, 0);
        test!(-1., $fty -> $ity, 0);
        test!(-100., $fty -> $ity, 0);
        test!(#[allow(overflowing_literals)] -1e50, $fty -> $ity, 0);
        test!(#[allow(overflowing_literals)] -1e130, $fty -> $ity, 0);
    )+ });

    (f* -> $($ity:ident)+) => ({
        fptoui_tests!(f32 -> $($ity)+);
        fptoui_tests!(f64 -> $($ity)+);
    })
}

fn main() {
    common_fptoi_tests!(f* -> i8 i16 i32 i64 u8 u16 u32 u64);
    fptoui_tests!(f* -> u8 u16 u32 u64);
    common_fptoi_tests!(f* -> i128 u128);
    fptoui_tests!(f* -> u128);
}

Found through the Miri test suite. Unfortunately we have barely any test coverage on the rustc side for mir-opt-level 2 and higher.

Cc @rust-lang/wg-mir-opt

[spastorino]

I was too fast to request prioritization but then realized that this is requires-nightly. Anyway we've briefly discussed this on Zulip.

[RalfJung]

Ah I should also add the regression range: 033013c was still good. 67100f6 has the bug.

Is it appropriate to summon the ICE breakers for this?