Rollup merge of #120248 - WaffleLapkin:bonk-ptr-object-casts, r=compiler-errors,oli-obk,lnicola

Make casts of pointers to trait objects stricter

This is an attempt to `fix` https://github.com/rust-lang/rust/issues/120222 and https://github.com/rust-lang/rust/issues/120217.

This is done by adding restrictions on casting pointers to trait objects.

Before this PR the rules were as follows:

> When casting `*const X<dyn A>` -> `*const Y<dyn B>`, principal traits in `A` and `B` must refer to the same trait definition (or no trait).

With this PR the rules are changed to

> When casting `*const X<dyn Src>` -> `*const Y<dyn Dst>`
> - if `Dst` has a principal trait `DstP`,
>   - `Src` must have a principal trait `SrcP`
>   - `dyn SrcP` and `dyn DstP` must be the same type (modulo the trait object lifetime, `dyn T+'a` -> `dyn T+'b` is allowed)
>   - Auto traits in `Dst` must be a subset of auto traits in `Src`
>     - Not adhering to this is currently a FCW (warn-by-default + `FutureReleaseErrorReportInDeps`), instead of an error
> - if `Src` has a principal trait `Dst` must as well
>   - this restriction will be removed in a follow up PR

This ensures that
1. Principal trait's generic arguments match (no `*const dyn Tr<A>` -> `*const dyn Tr<B>` casts, which are a problem for [#120222](https://github.com/rust-lang/rust/issues/120222))
2. Principal trait's lifetime arguments match (no `*const dyn Tr<'a>` -> `*const dyn Tr<'b>` casts, which are a problem for [#120217](https://github.com/rust-lang/rust/issues/120217))
3. No auto traits can be _added_ (this is a problem for arbitrary self types, see [this comment](https://github.com/rust-lang/rust/pull/120248#discussion_r1463835350))

Some notes:
 - We only care about the metadata/last field, so you can still cast `*const dyn T` to `*const WithHeader<dyn T>`, etc
- The lifetime of the trait object itself (`dyn A + 'lt`) is not checked, so you can still cast `*mut FnOnce() + '_` to `*mut FnOnce() + 'static`, etc
  - This feels fishy, but I couldn't come up with a reason it must be checked

The diagnostics are currently not great, to say the least, but as far as I can tell this correctly fixes the issues.

cc `@oli-obk` `@compiler-errors` `@lcnr`
This commit is contained in:
Matthias Krüger 2024-07-08 16:28:15 +02:00 committed by GitHub
commit 4bd78b0a8d
No known key found for this signature in database
GPG key ID: B5690EEEBB952194

View file

@ -68,8 +68,6 @@ pub type RawMap<A> = hash_map::HashMap<TypeId, Box<A>, BuildHasherDefault<TypeId
/// The type parameter `A` allows you to use a different value type; normally you will want /// The type parameter `A` allows you to use a different value type; normally you will want
/// it to be `core::any::Any` (also known as `std::any::Any`), but there are other choices: /// it to be `core::any::Any` (also known as `std::any::Any`), but there are other choices:
/// ///
/// - If you want the entire map to be cloneable, use `CloneAny` instead of `Any`; with
/// that, you can only add types that implement `Clone` to the map.
/// - You can add on `+ Send` or `+ Send + Sync` (e.g. `Map<dyn Any + Send>`) to add those /// - You can add on `+ Send` or `+ Send + Sync` (e.g. `Map<dyn Any + Send>`) to add those
/// auto traits. /// auto traits.
/// ///
@ -79,9 +77,6 @@ pub type RawMap<A> = hash_map::HashMap<TypeId, Box<A>, BuildHasherDefault<TypeId
/// also spelled [`AnyMap`] for convenience. /// also spelled [`AnyMap`] for convenience.
/// - <code>[Map]&lt;dyn [core::any::Any] + Send&gt;</code> /// - <code>[Map]&lt;dyn [core::any::Any] + Send&gt;</code>
/// - <code>[Map]&lt;dyn [core::any::Any] + Send + Sync&gt;</code> /// - <code>[Map]&lt;dyn [core::any::Any] + Send + Sync&gt;</code>
/// - <code>[Map]&lt;dyn [CloneAny]&gt;</code>
/// - <code>[Map]&lt;dyn [CloneAny] + Send&gt;</code>
/// - <code>[Map]&lt;dyn [CloneAny] + Send + Sync&gt;</code>
/// ///
/// ## Example /// ## Example
/// ///
@ -205,12 +200,6 @@ mod tests {
assert_debug::<Map<dyn Any>>(); assert_debug::<Map<dyn Any>>();
assert_debug::<Map<dyn Any + Send>>(); assert_debug::<Map<dyn Any + Send>>();
assert_debug::<Map<dyn Any + Send + Sync>>(); assert_debug::<Map<dyn Any + Send + Sync>>();
assert_send::<Map<dyn CloneAny + Send>>();
assert_send::<Map<dyn CloneAny + Send + Sync>>();
assert_sync::<Map<dyn CloneAny + Send + Sync>>();
assert_debug::<Map<dyn CloneAny>>();
assert_debug::<Map<dyn CloneAny + Send>>();
assert_debug::<Map<dyn CloneAny + Send + Sync>>();
} }
#[test] #[test]
@ -232,53 +221,6 @@ mod tests {
} }
} }
// impl some traits for dyn Any
use core::fmt;
#[doc(hidden)]
pub trait CloneToAny {
/// Clone `self` into a new `Box<dyn CloneAny>` object.
fn clone_to_any(&self) -> Box<dyn CloneAny>;
}
impl<T: Any + Clone> CloneToAny for T {
#[inline]
fn clone_to_any(&self) -> Box<dyn CloneAny> {
Box::new(self.clone())
}
}
macro_rules! impl_clone {
($t:ty) => {
impl Clone for Box<$t> {
#[inline]
fn clone(&self) -> Box<$t> {
// SAFETY: this dance is to reapply any Send/Sync marker. Im not happy about this
// approach, given that I used to do it in safe code, but then came a dodgy
// future-compatibility warning where_clauses_object_safety, which is spurious for
// auto traits but still super annoying (future-compatibility lints seem to mean
// your bin crate needs a corresponding allow!). Although I explained my plight¹
// and it was all explained and agreed upon, no action has been taken. So I finally
// caved and worked around it by doing it this way, which matches whats done for
// core::any², so its probably not *too* bad.
//
// ¹ https://github.com/rust-lang/rust/issues/51443#issuecomment-421988013
// ² https://github.com/rust-lang/rust/blob/e7825f2b690c9a0d21b6f6d84c404bb53b151b38/library/alloc/src/boxed.rs#L1613-L1616
let clone: Box<dyn CloneAny> = (**self).clone_to_any();
let raw: *mut dyn CloneAny = Box::into_raw(clone);
unsafe { Box::from_raw(raw as *mut $t) }
}
}
impl fmt::Debug for $t {
#[inline]
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
f.pad(stringify!($t))
}
}
};
}
/// Methods for downcasting from an `Any`-like trait object. /// Methods for downcasting from an `Any`-like trait object.
/// ///
/// This should only be implemented on trait objects for subtraits of `Any`, though you can /// This should only be implemented on trait objects for subtraits of `Any`, though you can
@ -350,16 +292,3 @@ macro_rules! implement {
implement!(Any); implement!(Any);
implement!(Any + Send); implement!(Any + Send);
implement!(Any + Send + Sync); implement!(Any + Send + Sync);
/// [`Any`], but with cloning.
///
/// Every type with no non-`'static` references that implements `Clone` implements `CloneAny`.
/// See [`core::any`] for more details on `Any` in general.
pub trait CloneAny: Any + CloneToAny {}
impl<T: Any + Clone> CloneAny for T {}
implement!(CloneAny);
implement!(CloneAny + Send);
implement!(CloneAny + Send + Sync);
impl_clone!(dyn CloneAny);
impl_clone!(dyn CloneAny + Send);
impl_clone!(dyn CloneAny + Send + Sync);