(cherry picked from commit b791169)

LangItem might eventually be the type that forces us to support
`#[derive(StableHash)]` for enums with more than 256 variants.

This commit also removes `#[derive(Hash)]` from the affected types, since it is
no longer used.

If these types need a plain Hash impl is needed in the future, it could
potentially share the `stable_hash_inner` implementation.

By moving away from the old `impl_stable_traits_for_trivial_type!` macro, this
commit has the side-effect of removing unused StableOrd impls for some types.
Those impls could potentially be re-added if they are ever needed.

Add a test for a case that results in a recursive loop that runs until
it runs out of memory or gets killed. This used to take a long time to
compile because of nested recursion caused by dereferencing the type of
the struct.

Signed-off-by: Hardik Kumar <hardikxk@gmail.com>

This pulls some less-trivial changes out of the main migration to `Value<'ll>`.

derived impls kept the lifetime spans of the derive input so the external macro check missed them and `#[expect(redundant_lifetimes)]` on a derived struct went unfulfilled. the lint now skips items whose own span is in an external macro.

Just because the parser of a specific syntax fragment kind returns an
`Err(_)` doesn't automatically mean that the problematic token sequence
isn't "of" that fragment's kind. The token sequence might "almost" be of
that kind (i.e., share a non-empty prefix) but simply contain a syntax
error in it.

On a side note, it's rather odd that the now removed
`annotate_err_with_kind` only added the note "this macro call doesn't
expand to a $fragkind" if the target fragment kind was *type* or *pat*,
there are 20 kinds at the of writing.

…..)`

This also means that it's now available under `core::pat` instead of
`core` directly. This makes it consistent with the `std` re-export
which is only available under `std::pat` and not under `std` directly.

Moreover, drop library feature `pattern_type_macro` and demote language
feature `pattern_types` to a library feature. This is nice because it
means `std::pat::pattern_type` is no longer unnecessarily gated behind
two (!) feature gates.

This also means that it's now available under `core::view` instead of
`core` directly. This makes it consistent with the `std` re-export
which is only available under `std::view` and not under `std` directly.

Moreover, drop library feature `view_type_macro` and demote language
feature `view_types` to a library feature. This is nice because it
means `std::view::view_type` is no longer unnecessarily gated behind
two (!) feature gates.

Co-authored-by: xkevio <xkeviota@gmail.com>

style: use ecx stmt item method

cg_llvm: Use an invariant lifetime brand for LLVM values

This PR changes the type of LLVM value references from `&'ll Value` to `Value<'ll>` with an invariant lifetime brand, to help uphold soundness of the FFI bindings.

There should be no change to compiler output.

---

Value is one of the more widely-used types, much more so than Type, so unfortunately this step ends up needing to touch quite a lot of code. Thankfully almost all of the changes are trivial.

…rk-Simulacrum

Bump stage0 to 1.101.0-beta.1

https://forge.rust-lang.org/release/process.html#default-branch-bootstrap-update-and-crater-tuesday

- run replace-version-placeholder (cherry-picked from beta)
- bump stage0 to 1.101.0-beta.1
- update STAGE0_MISSING_TARGETS
- reformat with `cfg_select!` awareness
- update `cfg(bootstrap)`

Implement `core::ops::FnAddr`

This implements `FnAddr` as proposed in rust-lang/libs-team#837 (comment).

This is my first contribution to the rust compiler, so I'm not 100% familiar with the process.

All code was hand written by me.

cc rust-lang#148768

…syntax, r=mati865

Migrate built-in macros `pattern_type` & `view_type` to macros 2.0 that expand to internal syntax

Part of MCP-approved rust-lang#110680.

Best reviewed commit by commit. Please see individual commit messages for further details.

Pinging owner of pattern types @oli-obk & owner of view types @scrabsha for visibility.

<sub>(No LLM was or will be used by me during the entire creation process of this PR)</sub>

…ternal-macro, r=petrochenkov

do not lint `redundant_lifetimes` on items from external macros

derived impls copy the lifetime spans from the derive input so the external macro check on the lint span never fires and the lint is reported on the generated impl. that also left `#[expect(redundant_lifetimes)]` on a derived struct unfulfillable. now the lint skips items whose own span is in an external macro.

fixes rust-lang#150553
as suggested in rust-lang#150553 (comment)

r? @petrochenkov

feat: support nested functions with `#[offload_kernel]`

Related to rust-lang#131513.

Allows annotating nested functions with `#[offload_kernel]`, similar to autodiff.

Also fixes the error message when `#[offload_kernel]` is used on anything other than functions.

r? @ZuseZ4

(@Sa4dUs, ping)