An example of an FFI-safe enum conversion is when converting
Option<NonZeroUsize> to usize. Because the Some value must be non-zero,
rustc can use 0 to represent the None variant, making this conversion is
safe. Furthermore, it can be relied on (and removing this optimisation
already would be a breaking change).
Make closures and generators a must use types
Warn about unused expressions with closure or generator type. This follows
existing precedence of must use annotations present on `FnOnce`, `FnMut`, `Fn`
traits, which already indirectly apply to closures in some cases, e.g.,:
```rust
fn f() -> impl FnOnce() {
|| {}
}
fn main() {
// an existing warning: unused implementer of `std::ops::FnOnce` that must be used:
f();
// a new warning: unused closure that must be used:
|| {};
}
```
Closes#74691.
Stabilize const_type_id feature
The tracking issue for `const_type_id` points to the ill-fated #41875. So I'm re-energizing `TypeId` shenanigans by opening this one up to see if there's anything blocking us from stabilizing the constification of type ids.
Will wait for CI before pinging teams/groups.
-----
This PR stabilizes the `const_type_id` feature, which allows `TypeId::of` (and the underlying unstable intrinsic) to be called in constant contexts.
There are some [sanity tests](https://github.com/rust-lang/rust/blob/master/src/test/ui/consts/const-typeid-of-rpass.rs) that demonstrate its usage, but I’ve included some more below.
As a simple example, you could create a constant item that contains some type ids:
```rust
use std::any::TypeId;
const TYPE_IDS: [TypeId; 2] = [
TypeId::of::<u32>(),
TypeId::of::<i32>(),
];
assert_eq!(TypeId::of::<u32>(), TYPE_IDS[0]);
```
Type ids can also now appear in associated constants. You could create a trait that associates each type with its constant type id:
```rust
trait Any where Self: 'static {
const TYPE_ID: TypeId = TypeId::of::<Self>();
}
impl<T: 'static> Any for T { }
assert_eq!(TypeId::of::<usize>(), usize::TYPE_ID);
```
`TypeId::of` is generic, which we saw above in the way the generic `Self` argument was used. This has some implications for const evaluation. It means we can make trait impls evaluate differently depending on information that wasn't directly passed through the trait system. This violates the _parametricity_ property, which requires all instances of a generic function to behave the same way with respect to its generic parameters. That's not unique to `TypeId::of`, other generic const functions based on compiler intrinsics like `mem::align_of` can also violate parametricity. In practice Rust doesn't really have type parametricity anyway since it monomorphizes generics into concrete functions, so violating it using type ids isn’t new.
As an example of how impls can behave differently, you could combine constant type ids with the `const_if_match` feature to dispatch calls based on the type id of the generic `Self`, rather than based on information about `Self` that was threaded through trait bounds. It's like a rough-and-ready form of specialization:
```rust
#![feature(const_if_match)]
trait Specialized where Self: 'static {
// An associated constant that determines the function to call
// at compile-time based on `TypeId::of::<Self>`.
const CALL: fn(&Self) = {
const USIZE: TypeId = TypeId::of::<usize>();
match TypeId::of::<Self>() {
// Use a closure for `usize` that transmutes the generic `Self` to
// a concrete `usize` and dispatches to `Self::usize`.
USIZE => |x| Self::usize(unsafe { &*(x as *const Self as *const usize) }),
// For other types, dispatch to the generic `Self::default`.
_ => Self::default,
}
};
fn call(&self) {
// Call the function we determined at compile-time
(Self::CALL)(self)
}
fn default(x: &Self);
fn usize(x: &usize);
}
// Implement our `Specialized` trait for any `Debug` type.
impl<T: fmt::Debug + 'static> Specialized for T {
fn default(x: &Self) {
println!("default: {:?}", x);
}
fn usize(x: &usize) {
println!("usize: {:?}", x);
}
}
// Will print "usize: 42"
Specialized::call(&42usize);
// Will print "default: ()"
Specialized::call(&());
```
Type ids have some edges that this stabilization exposes to more contexts. It's possible for type ids to collide (but this is a bug). Since they can change between compiler versions, it's never valid to cast a type id to its underlying value.
Warn about unused expressions with closure or generator type. This follows
existing precedence of must use annotations present on `FnOnce`, `FnMut`, `Fn`
traits, which already indirectly apply to closures in some cases, e.g.,:
```rust
fn f() -> impl FnOnce() {
|| {}
}
fn main() {
// an existing warning: unused implementer of `std::ops::FnOnce` that must be used:
f();
// a new warning: unused closure that must be used:
|| {};
}
```
Forbid generic parameters in anon consts inside of type defaults
Emit a resolution error for `struct Foo<T, U = [u8; std::mem::size_of::<T>()]>`.
We are unable to support this with the way `ty::Generics` is currently used,
so let's just forbid it entirely for now.
Fixes some ICE on stable, e.g.
```rust
struct Foo<T, U = [u8; std::mem::size_of::<*mut T>()]>(T, U);
```
r? @varkor @eddyb
Miri: replace canonical_alloc_id mechanism by extern_static_alloc_id
We only have to call `extern_static_alloc_id` when a `Pointer` is "imported" from the `tcx` to the machine, not on each access. Also drop the old hook for TLS handling, it is not needed any more.
The Miri side of this is at https://github.com/rust-lang/miri/pull/1489.
Fixes https://github.com/rust-lang/rust/issues/71194
r? @oli-obk
Rollup of 6 pull requests
Successful merges:
- #74088 (Avoid writes without any data in `Write::write_all_vectored`)
- #74598 (Fix sync_once_cell_does_not_leak_partially_constructed_boxes)
- #74750 (Clean up some uses of logging in ui tests)
- #74783 (python codes cleanup)
- #74790 (Don't italicize comments in ayu theme)
- #74799 (Fixed typo in `closure`)
Failed merges:
r? @ghost
Clean up some uses of logging in ui tests
The removed test can't possibly trigger anything today as we don't have logging in libstd.
The `exec-env` flag was mistakenly used for adding env vars to rustc invocations both in test and in the test suite and there were some accidental renames from RUST_LOG to RUSTC_LOG that I reverted.
Pull out some duplicated code into a new function
I debated pulling the actual struct_span_err calls into the new method, but I felt like having to pass in multiple arguments for it and wiring up string formatting outweighed the benefits.
Viewing the diff with whitespace ignored is recommended.
Serialize span hygiene data
Fixes#68686Fixes#70963
This PR serializies global hygiene data into both the incremental compilation cache and the crate metadata. This allows hygiene information to be preserved across compilation sessions (both incremental and cross-crate).
When serializing a `SyntaxContext`, we simply write out the raw id from the current compilation session. Whenever we deserialize a `SyntaxContext`, we 'remap' the id to a fresh id in our current compilation session, and load the associated `SyntaxContextData`.
As a result, some 'upstream' `SyntaxContextData` will end up getting duplicated in 'downstream' crates. This only happens when we actually need to use an 'upstream' `SyntaxContext`, which occurs when we deserialize a `Span` that requires it.
We serialize an `ExpnData` into the metadata of the crate which generated it. An `ExpnId` is serialized as a reference into the crate which 'owns' the corresponding `ExpnData`, which avoids duplication in downstream crates.
I've included a macros 2.0 test which requires hygiene serialization to compile successfully.
TODO:
- [x] <strike>Determine how many additional `DefId`s we end up creating for `ExpnId`s - this may be significant for `libcore`, which uses macros heavily. Alternatively, we could try to compute a `DefPathHash` without making a corresponding `DefId` - however, this might significantly complicate the implementation.</strike> (We no longer create `DefId`s)
- [x] Investigate the overhead of duplicating `SyntaxContextData` in crate metadata.
- [x] Investigate how `resolve_crate_root` behaves with deserialized hygiene data - the current logic may be wrong.
- [x] Add additional tests. The effects of this PR are usually only noticeable when working with headache-inducing macro expansions (e.g. macros expanding to macros), so there are lots of corner cases to test.
- [x] Determine what to do about this:
4774f9b523/src/librustc_resolve/build_reduced_graph.rs (L892)
- [x] Determine if we need to do anything here - I think the fact that `src/test/ui/hygiene/cross_crate_hygiene.rs` passes means that this is working.
3d5d0f898c/src/librustc_resolve/imports.rs (L1389-L1392)
A raw SyntaxContext id is implicitly dependent on the target platform,
since libstd and libcore have platform-dependent #[cfg]s which affect
which macros are invoked. As a result, we must strip out any
SyntaxContext ids from test output to ensure that the captured stdout is
not platform-dependent.
Remove trait LengthAtMost32
This is a continuation of https://github.com/rust-lang/rust/pull/74026 preserving the original burrbull's commit.
I talked to @burrbull, he suggested me to finish his PR.
Normalize bounds fully when checking defaulted types
When checking that the default type for `<T as X>::Y` is valid in this example:
```
trait X { type Y: PartialEq<<Self as X>::Y> }
impl X for T { default type Y = S; }
```
We will have to prove the bound `S: PartialEq<<T as X>::Y>`. In this case
we want `<T as X>::Y` to normalize to `S`. This is valid because we are
checking the default value specifically here. Add `<T as X>::Y = S` to the
ParamEnv for normalization _of the bound we are checking_ only.
Fixes#73818.
---
I noticed that adding this to the env for bounds checking didn't break any tests. Not sure if this is because we can't rely on it to prove anything, or because of missing test coverage.
r? @matthewjasper, @nikomatsakis
Optimize away BitAnd and BitOr when possible
This PR lets `const_prop` optimize away `a | true == true` , `a & false == false` and `a * 0 = 0`. While I was writing this I've realized that constant propagation misses a lot of opportunities. For example: https://play.rust-lang.org/?version=stable&mode=debug&edition=2018&gist=2a4b45e772f214210a36749b27223bb0
Constant propagation doesn't seem to... propagate constants, additionally the way constant propagation is currently setup makes it tricky to add cases like `a | false == a`.
I tried to organize `eval_rvalue_with_identities` to make the pattern of the optimizations easier to see but it still obscurs what should be a simple peephole optmization.
cc @oli-obk
correctly deal with unsorted generic parameters
We now stop sorting generic params and instead correctly handle unsorted params in the rest of the compiler.
We still restrict const params to come after type params though, so this PR does not change anything which
is visible to users.
This might slightly influence perf, so let's prevent any unintentional rollups. @bors rollup=never
r? @varkor
Fix ICE while building MIR with type errors
See https://github.com/rust-lang/rust/issues/74047#issuecomment-663290913 for background. Replacing a binding with `PatKind::Wild` (introduced in #51789 and later refactored in #67439) caused an ICE downstream while building MIR.
I noticed that taking this code out no longer triggers the ICEs it was added to prevent. I'm not sure what else changed, or if this change is _correct_, but it does seem to be passing ui tests at least.
r? @oli-obk
cc @estebank
Fixes#74047.
