error with #[rustc_safe_intrinsic] in Rust 1.84.0

[cuviper]

FYI, this showed up in the 1.84.0-beta crater run: rust-lang/rust#134138

https://crater-reports.s3.amazonaws.com/beta-1.84.0-4-retry2/beta-2024-12-08/gh/whitequark.superlinker/log.txt

[INFO] [stderr]   error: attributes starting with `rustc` are reserved for use by the `rustc` compiler
[INFO] [stderr]      --> shim/main.rs:100:11
[INFO] [stderr]       |
[INFO] [stderr]   100 |         #[rustc_safe_intrinsic]
[INFO] [stderr]       |           ^^^^^^^^^^^^^^^^^^^^
[INFO] [stderr] 
[INFO] [stderr]   error: cannot find attribute `rustc_safe_intrinsic` in this scope
[INFO] [stderr]      --> shim/main.rs:100:11
[INFO] [stderr]       |
[INFO] [stderr]   100 |         #[rustc_safe_intrinsic]
[INFO] [stderr]       |           ^^^^^^^^^^^^^^^^^^^^ help: a built-in attribute with a similar name exists: `rustc_intrinsic`
[INFO] [stderr] 
[INFO] [stderr]   error: requires `legacy_receiver` lang_item
[INFO] [stderr]      --> shim/main.rs:144:11
[INFO] [stderr]       |
[INFO] [stderr]   144 |     fn eq(&self, other: &usize) -> bool {
[INFO] [stderr]       |           ^^^^^
[INFO] [stderr] 
[INFO] [stderr]   error: requires `legacy_receiver` lang_item
[INFO] [stderr]      --> shim/main.rs:147:11
[INFO] [stderr]       |
[INFO] [stderr]   147 |     fn ne(&self, other: &usize) -> bool {
[INFO] [stderr]       |           ^^^^^
[INFO] [stderr] 
[INFO] [stderr]   error: intrinsic safety mismatch between list of intrinsics within the compiler and core library intrinsics for intrinsic `wrapping_add`
[INFO] [stderr]      --> shim/main.rs:102:9
[INFO] [stderr]       |
[INFO] [stderr]   102 |         pub(super) fn wrapping_add<T: Copy>(a: T, b: T) -> T;
[INFO] [stderr]       |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[INFO] [stderr] 
[INFO] [stderr]   error: aborting due to 5 previous errors

Since you're using RUSTC_BOOTSTRAP=1 for unstable features in this shim, it doesn't count as a Rust regression.

[whitequark]

Thank you. Just to be clear, this project is a fun experimental thing that works shockingly well for something that wasn't really written seriously, and you can entirely ignore it for the purpose of crater statistics.

[whitequark]

(The reason we wrote it that way is because we needed to avoid having any other function be included in the binary, or something like that, using Rust as a sort of macroassembler.)

I'll keep the issue open since it's a regression on our side.

[dramforever]

welp, so much for having fun with #[no_core]

[whitequark]

I wonder what legacy_receiver even is