Rollup merge of #99511 - RalfJung:raw_eq, r=wesleywiser

make raw_eq precondition more restrictive

Specifically, don't allow comparing pointers that way. Comparing pointers is subtle because you have to talk about what happens to the provenance.

This matches what [Miri already implements](https://play.rust-lang.org/?version=nightly&mode=debug&edition=2021&gist=9eb1dfb8a61b5a2d4a7cee43df2717af), and all existing users are fine with this.

If raw_eq on pointers is ever desired, we can adjust the intrinsic spec and Miri implementation as needed, but for now that seems just unnecessary. Also, this is a const intrinsic, and in const, comparing pointers this way is *not possible* -- so if we allow the intrinsic to compare pointers in general, we need to impose an extra restrictions saying that in const-context, pointers are *not* okay.

Author: matthiaskrgr

library/core/src/array/equality.rs

@@ -173,13 +173,14 @@ macro_rules! is_raw_eq_comparable {
     )+};
 }
 
-// SAFETY: All the ordinary integer types allow all bit patterns as distinct values
+// SAFETY: All the ordinary integer types have no padding, and are not pointers.
 is_raw_eq_comparable!(u8, u16, u32, u64, u128, usize, i8, i16, i32, i64, i128, isize);
 
-// SAFETY: bool and char have *niches*, but no *padding*, so this is sound
+// SAFETY: bool and char have *niches*, but no *padding* (and these are not pointer types), so this
+// is sound
 is_raw_eq_comparable!(bool, char);
 
-// SAFETY: Similarly, the non-zero types have a niche, but no undef,
+// SAFETY: Similarly, the non-zero types have a niche, but no undef and no pointers,
 // and they compare like their underlying numeric type.
 is_raw_eq_comparable!(
     NonZeroU8,

library/core/src/intrinsics.rs

@@ -2282,7 +2282,8 @@ extern "rust-intrinsic" {
     ///
     /// # Safety
     ///
-    /// It's UB to call this if any of the *bytes* in `*a` or `*b` are uninitialized.
+    /// It's UB to call this if any of the *bytes* in `*a` or `*b` are uninitialized or carry a
+    /// pointer value.
     /// Note that this is a stricter criterion than just the *values* being
     /// fully-initialized: if `T` has padding, it's UB to call this intrinsic.
     ///