Don't add empty target features for target-cpu=native on macOS
LLVM does not support host feature detection (only host cpu detection) on apple platforms. As such, the returned feature string will be empty. Adding this empty string to the target-features attribute results in a verifier error on LLVM 22.
Fix this by not adding the empty string to the target features. The reason why this was not caught by the target-cpu-native test is that it requires a function that adds *some* target features, otherwise the attribute is omitted entirely. We achieve this with a somewhat peculiar construction that enables `neon` if it's already enabled. (This is to avoid enabling it on softfloat targets.)
Fixes https://github.com/rust-lang/rust/issues/153397.
Work around a false `err.emit();` type error in rust-analyzer
For whatever reason, rust-analyzer doesn't see that these calls to `err.emit();` are diverging, so the trailing semicolon makes r-a complain about a type mismatch between `()` and some other type.
Removing the trailing semicolon makes no functional difference (because the emit still unwinds the stack), but seems to be enough to allow rust-analyzer to see that emitting an error returns `!` in these cases, which silences the false error.
rustc_public: remove the `CrateDefItems` trait
This trait feels sus since its documentation says it's for 'retrieving all items from a definition' however it only provides an `associated_items` method (??), which I believe should only be valid for `impl`s and `trait`s.
Clarify a confusing green-path function
The current name of this function, `try_load_from_disk_and_cache_in_memory`, is confusing on a number of levels:
- Trying to mark the node green is load-bearing even for queries that never cache to disk.
- It will only return None if it fails to mark the query green; the subsequent parts always return Some.
- If it cannot load a value from disk, it will obtain a value by invoking the query provider.
- It is not actually responsible for storing values in the in-memory cache; that is handled by an outer layer.
This PR therefore:
- Hoists the try-mark-green out of that function, making the function always return a value.
- Renames the function to `load_from_disk_or_invoke_provider_green `.
- Renames a few other local variables in passing.
There should be no change to compiler behaviour.
Stop using `LinkedGraph` in `lexical_region_resolve`
There are only two users of the older `LinkedGraph` data structure, and this is one.
It turns out that this diagnostic-related code doesn't need any non-trivial graph operations (since it does its own graph traversal); it just needs the ability to get a list of in-edges or out-edges (constraints) for any particular node. That's easy enough to do with a simple custom data structure.
Inspired by https://github.com/rust-lang/rust/pull/152621, which wants to make changes to `LinkedGraph` that wouldn't make sense for this use-site.
explicit tail calls: support indirect arguments
tracking issue: https://github.com/rust-lang/rust/issues/112788
After discussion in https://github.com/rust-lang/rust/issues/144855, I was wrong and it is actually possible to support tail calls with `PassMode::Indirect { on_stack: false, .. }` arguments.
Normally an indirect argument with `on_stack: false` would be passed as a pointer into the caller's stack frame. For tail calls, that would be unsound, because the caller's stack frame is overwritten by the callee's stack frame.
Therefore we store the argument for the callee in the corresponding caller's slot. Because guaranteed tail calls demand that the caller's signature matches the callee's, the corresponding slot has the correct type.
To handle cases like the one below, the tail call arguments must first be copied to a temporary, and can only then be copied to the caller's argument slots.
```rust
// A struct big enough that it is not passed via registers.
pub struct Big([u64; 4]);
fn swapper(a: Big, b: Big) -> (Big, Big) {
become swapper_helper(b, a);
}
```
---
I'm not really familiar with MIR and what tricks/helpers we have, so I'm just cobbling this together. Hopefully we can arrive at something more elegant.
Clean up some code related to `QueryVTable::execute_query_fn`
This PR is an assortment of small cleanups to code that interacts with `execute_query_fn` in the query vtable.
I also experimented with trying to replace the macro-generated `__rust_end_short_backtrace` functions with a single shared generic function, but I couldn't manage to avoid breaking short backtraces, so I left a note behind to document my attempt.
r? nnethercote (or compiler)
Port `#[rustc_autodiff]` to the attribute parser infrastructure
For https://github.com/rust-lang/rust/issues/131229
r? @jdonszelmann
cc @ZuseZ4
`autodiff_forward` and `autodiff_reverse` can be ported in a seperate PR, but these are expanded in the AST and don't exist anymore in the HIR so this is a bit more of a challenge.
Avoid duplicate `requirement` diag args in `RegionOriginNote`
Fixesrust-lang/rust#143872
`RegionOriginNote::WithRequirement` can be emitted multiple times for one
diagnostic. I fixed the ICE by scoping per note fluent args in `RegionOriginNote::WithRequirement` with `store_args` and `restore_args`.
It's currently in `dep_node.rs`, along with a blanket impl. Specific
impls are in `dep_node_key.rs`. This commit moves the trait and the
blanket impl into `dep_node_key.rs`, so everything is in one place.
This function is tricky to document, and there's more that could be said here,
but I don't want to take up too much vertical space, or add too much risk of
details becoming inaccurate over time.
All three of these functions were previously taking `cache` and
`execute_query_fn` as separate arguments, but after some other recent
simplifications we can just pass `&'tcx QueryVTable<'tcx, C>` instead.
This makes it easier to see where `execute_query_fn` is actually called.
Rémy Rakic