Rollup of 5 pull requests
Successful merges:
- #128778 (atomics: allow atomic and non-atomic reads to race)
- #130918 (simplify LLVM submodule handling)
- #130960 (Only add an automatic SONAME for Rust dylibs)
- #130973 (compiletest: rename "runtest/crash.rs" to "runtest/crashes.rs" to be in line with the test directory)
- #130976 (remove couple redundant clones)
r? `@ghost`
`@rustbot` modify labels: rollup
Only add an automatic SONAME for Rust dylibs
#126094 added an automatic relative `SONAME` to all dynamic libraries, but it was really only needed for Rust `--crate-type="dylib"`. In Fedora, it was a surprise to see `SONAME` on `"cdylib"` libraries like Python modules, especially because that generates an undesirable RPM `Provides`. We can instead add a `SONAME` just for Rust dylibs by passing the crate-type argument farther.
Ref: https://bugzilla.redhat.com/show_bug.cgi?id=2314879
atomics: allow atomic and non-atomic reads to race
We currently define our atomics in terms of C++ `atomic_ref`. That has the unfortunate side-effect of making it UB for an atomic and a non-atomic read to race (concretely, [this code](https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=d1a743774e60923db33def7fe314d754) has UB). There's really no good reason for this, all the academic models of the C++ memory model I am aware of allow this -- C++ just disallows this because of their insistence on an "object model" with typed memory, where `atomic_ref` temporarily creates an "atomic object" that may not be accesses via regular non-atomic operations.
So instead of tying our operations to `atomic_ref`, let us tie them directly to the underlying C++ memory model. I am not sure what is the best way to phrase this, so here's a first attempt.
We also carve out an exception from the "no mixed-size atomic accesses" rule to permit mixed-size atomic reads -- given that we permit mixed-size non-atomic reads, it seems odd that this would be disallowed for atomic reads. However, when an atomic write races with any other atomic operation, they must use the same size.
With this change, it is finally the case that every non-atomic access can be replaced by an atomic access without introducing UB.
Cc `@rust-lang/opsem` `@chorman0773` `@m-ou-se` `@WaffleLapkin` `@Amanieu`
Fixes https://github.com/rust-lang/unsafe-code-guidelines/issues/483
add tests for validity of Box with custom allocator
Ensure that the validity visitor visits both parts of a box with custom allocator using the right types.
Rename a few tests to make tidy happier
A somewhat random smattering of tests that I have recently looked at, and thus had cause to research and write down the reason for their existence.
Allow instantiating trait object binder in ptr-to-ptr casts
For unsizing coercions between trait objects with the same principal, we already allow instantiating the for binder. For example, coercing `Box<dyn for<'a> Trait<'a>` to `Box<dyn Trait<'static>>` is allowed.
Since ptr-to-ptr casts will insert an unsizing coercion before the cast if possible, this has the consequence that the following compiles already:
```rust
// This compiles today.
fn cast<'b>(x: *mut dyn for<'a> Trait<'a>) -> *mut dyn Trait<'b> {
// lowered as (roughly)
// tmp: *mut dyn Trait<'?0> = Unsize(x) // requires dyn for<'a> Trait<'a> <: dyn Trait<'?0>
// ret: *mut dyn Trait<'b> = PtrToPtr(tmp) // requires dyn Trait<'?0> == dyn Trait<'b>
x as _
}
```
However, if no unsizing coercion is inserted then this currently fails to compile as one type is more general than the other. This PR will allow this code to compile, too, by changing ptr-to-ptr casts of pointers with vtable metadata to use sutyping instead of type equality.
```rust
// This will compile after this PR.
fn cast<'b>(x: *mut dyn for<'a> Trait<'a>) -> *mut Wrapper<dyn Trait<'b>> {
// lowered as (roughly)
// no Unsize here!
// ret: *mut Wrapper<dyn Trait<'b>> = PtrToPtr(x) // requires dyn for<'a> Trait<'a> == dyn Trait<'b>
x as _
}
```
Note that it is already possible to work around the current restrictions and make the code compile before this PR by splitting the cast in two, so this shouldn't allow a new class of programs to compile:
```rust
// Workaround that compiles today.
fn cast<'b>(x: *mut dyn for<'a> Trait<'a>) -> *mut Wrapper<dyn Trait<'b>> {
x as *mut dyn Trait<'_> as _
}
```
r? `@compiler-errors`
cc `@WaffleLapkin`
Revert space-saving operations
The "all of our artifacts" `mv` seems like it may save enough space to matter sometimes, since it can range up to a gigabyte of difference, if memory serves. For the rest, I think we're good.
try-job: dist-aarch64-apple
rustdoc: update `ProcMacro` docs section on helper attributes
I believe the mention of attribute macros in the section on proc macro helper attributes is erroneous. As far as I can tell, attribute macros cannot define helper attributes.
The following attribute macro is not valid (fails to build), no matter how I try to define (or skip defining) the helpers:
```rust
#[proc_macro_attribute(attributes(helper))]
pub fn attribute_helpers(_attr: TokenStream, item: TokenStream) -> TokenStream {
item
}
```
The [language reference](https://doc.rust-lang.org/reference/procedural-macros.html#attribute-macros) also doesn't seem to mention attribute macro helpers. The helpers subsection is inside the section on derive macros.
Make clashing_extern_declarations considering generic args for ADT field
In following example, G<u16> should be recognized as different from G<u32> :
```rust
#[repr(C)] pub struct G<T> { g: [T; 4] }
pub mod x { extern "C" { pub fn g(_: super::G<u16>); } }
pub mod y { extern "C" { pub fn g(_: super::G<u32>); } }
```
fixes#130851
Allow instantiating object trait binder when upcasting
This PR fixes two bugs (that probably need an FCP).
### We use equality rather than subtyping for upcasting dyn conversions
This code should be valid:
```rust
#![feature(trait_upcasting)]
trait Foo: for<'h> Bar<'h> {}
trait Bar<'a> {}
fn foo(x: &dyn Foo) {
let y: &dyn Bar<'static> = x;
}
```
But instead:
```
error[E0308]: mismatched types
--> src/lib.rs:7:32
|
7 | let y: &dyn Bar<'static> = x;
| ^ one type is more general than the other
|
= note: expected existential trait ref `for<'h> Bar<'h>`
found existential trait ref `Bar<'_>`
```
And so should this:
```rust
#![feature(trait_upcasting)]
fn foo(x: &dyn for<'h> Fn(&'h ())) {
let y: &dyn FnOnce(&'static ()) = x;
}
```
But instead:
```
error[E0308]: mismatched types
--> src/lib.rs:4:39
|
4 | let y: &dyn FnOnce(&'static ()) = x;
| ^ one type is more general than the other
|
= note: expected existential trait ref `for<'h> FnOnce<(&'h (),)>`
found existential trait ref `FnOnce<(&(),)>`
```
Specifically, both of these fail because we use *equality* when comparing the supertrait to the *target* of the unsize goal. For the first example, since our supertrait is `for<'h> Bar<'h>` but our target is `Bar<'static>`, there's a higher-ranked type mismatch even though we *should* be able to instantiate that supertrait binder when upcasting. Similarly for the second example.
### New solver uses equality rather than subtyping for no-op (i.e. non-upcasting) dyn conversions
This code should be valid in the new solver, like it is with the old solver:
```rust
// -Znext-solver
fn foo<'a>(x: &mut for<'h> dyn Fn(&'h ())) {
let _: &mut dyn Fn(&'a ()) = x;
}
```
But instead:
```
error: lifetime may not live long enough
--> <source>:2:11
|
1 | fn foo<'a>(x: &mut dyn for<'h> Fn(&'h ())) {
| -- lifetime `'a` defined here
2 | let _: &mut dyn Fn(&'a ()) = x;
| ^^^^^^^^^^^^^^^^^^^ type annotation requires that `'a` must outlive `'static`
|
= note: requirement occurs because of a mutable reference to `dyn Fn(&())`
```
Specifically, this fails because we try to coerce `&mut dyn for<'h> Fn(&'h ())` to `&mut dyn Fn(&'a ())`, which registers an `dyn for<'h> Fn(&'h ()): dyn Fn(&'a ())` goal. This fails because the new solver uses *equating* rather than *subtyping* in `Unsize` goals.
This is *mostly* not a problem... You may wonder why the same code passes on the new solver for immutable references:
```
// -Znext-solver
fn foo<'a>(x: &dyn Fn(&())) {
let _: &dyn Fn(&'a ()) = x; // works
}
```
That's because in this case, we first try to coerce via `Unsize`, but due to the leak check the goal fails. Then, later in coercion, we fall back to a simple subtyping operation, which *does* work.
Since `&T` is covariant over `T`, but `&mut T` is invariant, that's where the discrepancy between these two examples crops up.
---
r? lcnr or reassign :D
The only requirement for `f16` support, aside from LLVM not crashing and
no ABI issues, is that symbols to convert to and from `f32` are
available. Since the update to compiler-builtins in [1], we now provide
these on all platforms.
This also enables `f16` math since there are no further requirements.
Still excluded are platforms for which LLVM emits infinitely-recursing
code.
[1]: https://github.com/rust-lang/rust/pull/125016
etc: Add sample rust-analyzer configs for eglot & helix
LSP configuration in editors like Emacs (eglot) and helix does not
use the same JSON format as vscode and vim do. It is not obvious
how to set up LSP for rustc in those editors and the dev guide currently
does not cover them. Adding sample configuration files for those editors
alongside the currently existing JSON one would be helpful.
I figured having those included in the repo like the JSON one might save
someone some time and frustration otherwise spent on trying to get the
more niche editors' LSP to work properly. I'll add a section in the dev
guide too.
