Auto merge of #81172 - SimonSapin:ptr-metadata, r=oli-obk

Implement RFC 2580: Pointer metadata & VTable

RFC: rust-lang/rfcs#2580

~~Before merging this PR:~~

* [x] Wait for the end of the RFC’s [FCP to merge](rust-lang/rfcs#2580 (comment)).
* [x] Open a tracking issue: #81513
* [x] Update `#[unstable]` attributes in the PR with the tracking issue number

----

This PR extends the language with a new lang item for the `Pointee` trait which is special-cased in trait resolution to implement it for all types. Even in generic contexts, parameters can be assumed to implement it without a corresponding bound.

For this I mostly imitated what the compiler was already doing for the `DiscriminantKind` trait. I’m very unfamiliar with compiler internals, so careful review is appreciated.

This PR also extends the standard library with new unstable APIs in `core::ptr` and `std::ptr`:

```rust
pub trait Pointee {
    /// One of `()`, `usize`, or `DynMetadata<dyn SomeTrait>`
    type Metadata: Copy + Send + Sync + Ord + Hash + Unpin;
}

pub trait Thin = Pointee<Metadata = ()>;

pub const fn metadata<T: ?Sized>(ptr: *const T) -> <T as Pointee>::Metadata {}

pub const fn from_raw_parts<T: ?Sized>(*const (), <T as Pointee>::Metadata) -> *const T {}
pub const fn from_raw_parts_mut<T: ?Sized>(*mut (),<T as Pointee>::Metadata) -> *mut T {}

impl<T: ?Sized> NonNull<T> {
    pub const fn from_raw_parts(NonNull<()>, <T as Pointee>::Metadata) -> NonNull<T> {}

    /// Convenience for `(ptr.cast(), metadata(ptr))`
    pub const fn to_raw_parts(self) -> (NonNull<()>, <T as Pointee>::Metadata) {}
}

impl<T: ?Sized> *const T {
    pub const fn to_raw_parts(self) -> (*const (), <T as Pointee>::Metadata) {}
}

impl<T: ?Sized> *mut T {
    pub const fn to_raw_parts(self) -> (*mut (), <T as Pointee>::Metadata) {}
}

/// `<dyn SomeTrait as Pointee>::Metadata == DynMetadata<dyn SomeTrait>`
pub struct DynMetadata<Dyn: ?Sized> {
    // Private pointer to vtable
}

impl<Dyn: ?Sized> DynMetadata<Dyn> {
    pub fn size_of(self) -> usize {}
    pub fn align_of(self) -> usize {}
    pub fn layout(self) -> crate::alloc::Layout {}
}

unsafe impl<Dyn: ?Sized> Send for DynMetadata<Dyn> {}
unsafe impl<Dyn: ?Sized> Sync for DynMetadata<Dyn> {}
impl<Dyn: ?Sized> Debug for DynMetadata<Dyn> {}
impl<Dyn: ?Sized> Unpin for DynMetadata<Dyn> {}
impl<Dyn: ?Sized> Copy for DynMetadata<Dyn> {}
impl<Dyn: ?Sized> Clone for DynMetadata<Dyn> {}
impl<Dyn: ?Sized> Eq for DynMetadata<Dyn> {}
impl<Dyn: ?Sized> PartialEq for DynMetadata<Dyn> {}
impl<Dyn: ?Sized> Ord for DynMetadata<Dyn> {}
impl<Dyn: ?Sized> PartialOrd for DynMetadata<Dyn> {}
impl<Dyn: ?Sized> Hash for DynMetadata<Dyn> {}
```

API differences from the RFC, in areas noted as unresolved questions in the RFC:

* Module-level functions instead of associated `from_raw_parts` functions on `*const T` and `*mut T`, following the precedent of `null`, `slice_from_raw_parts`, etc.
* Added `to_raw_parts`

Author: bors

compiler/rustc_hir/src/lang_items.rs

@@ -201,6 +201,10 @@ language_item_table! {
     // The associated item of `trait DiscriminantKind`.
     Discriminant,            sym::discriminant_type,   discriminant_type,          Target::AssocTy;
 
+    PointeeTrait,            sym::pointee_trait,       pointee_trait,              Target::Trait;
+    Metadata,                sym::metadata_type,       metadata_type,              Target::AssocTy;
+    DynMetadata,             sym::dyn_metadata,        dyn_metadata,               Target::Struct;
+
     Freeze,                  sym::freeze,              freeze_trait,               Target::Trait;
 
     Drop,                    sym::drop,                drop_trait,                 Target::Trait;

compiler/rustc_middle/src/traits/mod.rs

@@ -479,6 +479,9 @@ pub enum ImplSource<'tcx, N> {
     /// ImplSource for a builtin `DeterminantKind` trait implementation.
     DiscriminantKind(ImplSourceDiscriminantKindData),
 
+    /// ImplSource for a builtin `Pointee` trait implementation.
+    Pointee(ImplSourcePointeeData),
+
     /// ImplSource automatically generated for a generator.
     Generator(ImplSourceGeneratorData<'tcx, N>),
 
@@ -497,7 +500,8 @@ impl<'tcx, N> ImplSource<'tcx, N> {
             ImplSource::Generator(c) => c.nested,
             ImplSource::Object(d) => d.nested,
             ImplSource::FnPointer(d) => d.nested,
-            ImplSource::DiscriminantKind(ImplSourceDiscriminantKindData) => Vec::new(),
+            ImplSource::DiscriminantKind(ImplSourceDiscriminantKindData)
+            | ImplSource::Pointee(ImplSourcePointeeData) => Vec::new(),
             ImplSource::TraitAlias(d) => d.nested,
         }
     }
@@ -512,7 +516,8 @@ impl<'tcx, N> ImplSource<'tcx, N> {
             ImplSource::Generator(c) => &c.nested[..],
             ImplSource::Object(d) => &d.nested[..],
             ImplSource::FnPointer(d) => &d.nested[..],
-            ImplSource::DiscriminantKind(ImplSourceDiscriminantKindData) => &[],
+            ImplSource::DiscriminantKind(ImplSourceDiscriminantKindData)
+            | ImplSource::Pointee(ImplSourcePointeeData) => &[],
             ImplSource::TraitAlias(d) => &d.nested[..],
         }
     }
@@ -557,6 +562,9 @@ impl<'tcx, N> ImplSource<'tcx, N> {
             ImplSource::DiscriminantKind(ImplSourceDiscriminantKindData) => {
                 ImplSource::DiscriminantKind(ImplSourceDiscriminantKindData)
             }
+            ImplSource::Pointee(ImplSourcePointeeData) => {
+                ImplSource::Pointee(ImplSourcePointeeData)
+            }
             ImplSource::TraitAlias(d) => ImplSource::TraitAlias(ImplSourceTraitAliasData {
                 alias_def_id: d.alias_def_id,
                 substs: d.substs,
@@ -635,6 +643,9 @@ pub struct ImplSourceFnPointerData<'tcx, N> {
 #[derive(Clone, Debug, PartialEq, Eq, TyEncodable, TyDecodable, HashStable)]
 pub struct ImplSourceDiscriminantKindData;
 
+#[derive(Clone, Debug, PartialEq, Eq, TyEncodable, TyDecodable, HashStable)]
+pub struct ImplSourcePointeeData;
+
 #[derive(Clone, PartialEq, Eq, TyEncodable, TyDecodable, HashStable, TypeFoldable, Lift)]
 pub struct ImplSourceTraitAliasData<'tcx, N> {
     pub alias_def_id: DefId,

compiler/rustc_middle/src/traits/select.rs

@@ -125,6 +125,9 @@ pub enum SelectionCandidate<'tcx> {
     /// Builtin implementation of `DiscriminantKind`.
     DiscriminantKindCandidate,
 
+    /// Builtin implementation of `Pointee`.
+    PointeeCandidate,
+
     TraitAliasCandidate(DefId),
 
     /// Matching `dyn Trait` with a supertrait of `Trait`. The index is the

compiler/rustc_middle/src/traits/structural_impls.rs

@@ -19,6 +19,8 @@ impl<'tcx, N: fmt::Debug> fmt::Debug for traits::ImplSource<'tcx, N> {
 
             super::ImplSource::DiscriminantKind(ref d) => write!(f, "{:?}", d),
 
+            super::ImplSource::Pointee(ref d) => write!(f, "{:?}", d),
+
             super::ImplSource::Object(ref d) => write!(f, "{:?}", d),
 
             super::ImplSource::Param(ref n, ct) => {
@@ -110,4 +112,5 @@ impl<'tcx, N: fmt::Debug> fmt::Debug for traits::ImplSourceTraitAliasData<'tcx,
 TrivialTypeFoldableAndLiftImpls! {
     super::IfExpressionCause,
     super::ImplSourceDiscriminantKindData,
+    super::ImplSourcePointeeData,
 }

compiler/rustc_middle/src/ty/sty.rs

@@ -2133,6 +2133,54 @@ impl<'tcx> TyS<'tcx> {
         }
     }
 
+    /// Returns the type of metadata for (potentially fat) pointers to this type.
+    pub fn ptr_metadata_ty(&'tcx self, tcx: TyCtxt<'tcx>) -> Ty<'tcx> {
+        // FIXME: should this normalize?
+        let tail = tcx.struct_tail_without_normalization(self);
+        match tail.kind() {
+            // Sized types
+            ty::Infer(ty::IntVar(_) | ty::FloatVar(_))
+            | ty::Uint(_)
+            | ty::Int(_)
+            | ty::Bool
+            | ty::Float(_)
+            | ty::FnDef(..)
+            | ty::FnPtr(_)
+            | ty::RawPtr(..)
+            | ty::Char
+            | ty::Ref(..)
+            | ty::Generator(..)
+            | ty::GeneratorWitness(..)
+            | ty::Array(..)
+            | ty::Closure(..)
+            | ty::Never
+            | ty::Error(_)
+            | ty::Foreign(..)
+            // If returned by `struct_tail_without_normalization` this is a unit struct
+            // without any fields, or not a struct, and therefore is Sized.
+            | ty::Adt(..)
+            // If returned by `struct_tail_without_normalization` this is the empty tuple,
+            // a.k.a. unit type, which is Sized
+            | ty::Tuple(..) => tcx.types.unit,
+
+            ty::Str | ty::Slice(_) => tcx.types.usize,
+            ty::Dynamic(..) => {
+                let dyn_metadata = tcx.lang_items().dyn_metadata().unwrap();
+                tcx.type_of(dyn_metadata).subst(tcx, &[tail.into()])
+            },
+
+            ty::Projection(_)
+            | ty::Param(_)
+            | ty::Opaque(..)
+            | ty::Infer(ty::TyVar(_))
+            | ty::Bound(..)
+            | ty::Placeholder(..)
+            | ty::Infer(ty::FreshTy(_) | ty::FreshIntTy(_) | ty::FreshFloatTy(_)) => {
+                bug!("`ptr_metadata_ty` applied to unexpected type: {:?}", tail)
+            }
+        }
+    }
+
     /// When we create a closure, we record its kind (i.e., what trait
     /// it implements) into its `ClosureSubsts` using a type
     /// parameter. This is kind of a phantom type, except that the

compiler/rustc_span/src/symbol.rs

@@ -475,6 +475,7 @@ symbols! {
         dropck_eyepatch,
         dropck_parametricity,
         dylib,
+        dyn_metadata,
         dyn_trait,
         edition_macro_pats,
         eh_catch_typeinfo,
@@ -709,6 +710,7 @@ symbols! {
         memory,
         message,
         meta,
+        metadata_type,
         min_align_of,
         min_align_of_val,
         min_const_fn,
@@ -831,6 +833,7 @@ symbols! {
         plugin,
         plugin_registrar,
         plugins,
+        pointee_trait,
         pointer,
         pointer_trait,
         pointer_trait_fmt,

compiler/rustc_trait_selection/src/traits/project.rs

@@ -12,7 +12,7 @@ use super::SelectionContext;
 use super::SelectionError;
 use super::{
     ImplSourceClosureData, ImplSourceDiscriminantKindData, ImplSourceFnPointerData,
-    ImplSourceGeneratorData, ImplSourceUserDefinedData,
+    ImplSourceGeneratorData, ImplSourcePointeeData, ImplSourceUserDefinedData,
 };
 use super::{Normalized, NormalizedTy, ProjectionCacheEntry, ProjectionCacheKey};
 
@@ -1069,6 +1069,51 @@ fn assemble_candidates_from_impls<'cx, 'tcx>(
                     | ty::Error(_) => false,
                 }
             }
+            super::ImplSource::Pointee(..) => {
+                // While `Pointee` is automatically implemented for every type,
+                // the concrete metadata type may not be known yet.
+                //
+                // Any type with multiple potential metadata types is therefore not eligible.
+                let self_ty = selcx.infcx().shallow_resolve(obligation.predicate.self_ty());
+
+                // FIXME: should this normalize?
+                let tail = selcx.tcx().struct_tail_without_normalization(self_ty);
+                match tail.kind() {
+                    ty::Bool
+                    | ty::Char
+                    | ty::Int(_)
+                    | ty::Uint(_)
+                    | ty::Float(_)
+                    | ty::Foreign(_)
+                    | ty::Str
+                    | ty::Array(..)
+                    | ty::Slice(_)
+                    | ty::RawPtr(..)
+                    | ty::Ref(..)
+                    | ty::FnDef(..)
+                    | ty::FnPtr(..)
+                    | ty::Dynamic(..)
+                    | ty::Closure(..)
+                    | ty::Generator(..)
+                    | ty::GeneratorWitness(..)
+                    | ty::Never
+                    // If returned by `struct_tail_without_normalization` this is a unit struct
+                    // without any fields, or not a struct, and therefore is Sized.
+                    | ty::Adt(..)
+                    // If returned by `struct_tail_without_normalization` this is the empty tuple.
+                    | ty::Tuple(..)
+                    // Integers and floats are always Sized, and so have unit type metadata.
+                    | ty::Infer(ty::InferTy::IntVar(_) | ty::InferTy::FloatVar(..)) => true,
+
+                    ty::Projection(..)
+                    | ty::Opaque(..)
+                    | ty::Param(..)
+                    | ty::Bound(..)
+                    | ty::Placeholder(..)
+                    | ty::Infer(..)
+                    | ty::Error(_) => false,
+                }
+            }
             super::ImplSource::Param(..) => {
                 // This case tell us nothing about the value of an
                 // associated type. Consider:
@@ -1169,6 +1214,7 @@ fn confirm_select_candidate<'cx, 'tcx>(
         super::ImplSource::DiscriminantKind(data) => {
             confirm_discriminant_kind_candidate(selcx, obligation, data)
         }
+        super::ImplSource::Pointee(data) => confirm_pointee_candidate(selcx, obligation, data),
         super::ImplSource::Object(_)
         | super::ImplSource::AutoImpl(..)
         | super::ImplSource::Param(..)
@@ -1256,6 +1302,26 @@ fn confirm_discriminant_kind_candidate<'cx, 'tcx>(
     confirm_param_env_candidate(selcx, obligation, ty::Binder::dummy(predicate), false)
 }
 
+fn confirm_pointee_candidate<'cx, 'tcx>(
+    selcx: &mut SelectionContext<'cx, 'tcx>,
+    obligation: &ProjectionTyObligation<'tcx>,
+    _: ImplSourcePointeeData,
+) -> Progress<'tcx> {
+    let tcx = selcx.tcx();
+
+    let self_ty = selcx.infcx().shallow_resolve(obligation.predicate.self_ty());
+    let substs = tcx.mk_substs([self_ty.into()].iter());
+
+    let metadata_def_id = tcx.require_lang_item(LangItem::Metadata, None);
+
+    let predicate = ty::ProjectionPredicate {
+        projection_ty: ty::ProjectionTy { substs, item_def_id: metadata_def_id },
+        ty: self_ty.ptr_metadata_ty(tcx),
+    };
+
+    confirm_param_env_candidate(selcx, obligation, ty::Binder::bind(predicate), false)
+}
+
 fn confirm_fn_pointer_candidate<'cx, 'tcx>(
     selcx: &mut SelectionContext<'cx, 'tcx>,
     obligation: &ProjectionTyObligation<'tcx>,

compiler/rustc_trait_selection/src/traits/select/candidate_assembly.rs

@@ -267,6 +267,9 @@ impl<'cx, 'tcx> SelectionContext<'cx, 'tcx> {
         } else if lang_items.discriminant_kind_trait() == Some(def_id) {
             // `DiscriminantKind` is automatically implemented for every type.
             candidates.vec.push(DiscriminantKindCandidate);
+        } else if lang_items.pointee_trait() == Some(def_id) {
+            // `Pointee` is automatically implemented for every type.
+            candidates.vec.push(PointeeCandidate);
         } else if lang_items.sized_trait() == Some(def_id) {
             // Sized is never implementable by end-users, it is
             // always automatically computed.

compiler/rustc_trait_selection/src/traits/select/confirmation.rs

@@ -30,7 +30,8 @@ use crate::traits::{BuiltinDerivedObligation, ImplDerivedObligation};
 use crate::traits::{
     ImplSourceAutoImplData, ImplSourceBuiltinData, ImplSourceClosureData,
     ImplSourceDiscriminantKindData, ImplSourceFnPointerData, ImplSourceGeneratorData,
-    ImplSourceObjectData, ImplSourceTraitAliasData, ImplSourceUserDefinedData,
+    ImplSourceObjectData, ImplSourcePointeeData, ImplSourceTraitAliasData,
+    ImplSourceUserDefinedData,
 };
 use crate::traits::{ObjectCastObligation, PredicateObligation, TraitObligation};
 use crate::traits::{Obligation, ObligationCause};
@@ -99,6 +100,8 @@ impl<'cx, 'tcx> SelectionContext<'cx, 'tcx> {
                 Ok(ImplSource::DiscriminantKind(ImplSourceDiscriminantKindData))
             }
 
+            PointeeCandidate => Ok(ImplSource::Pointee(ImplSourcePointeeData)),
+
             TraitAliasCandidate(alias_def_id) => {
                 let data = self.confirm_trait_alias_candidate(obligation, alias_def_id);
                 Ok(ImplSource::TraitAlias(data))

compiler/rustc_trait_selection/src/traits/select/mod.rs

@@ -1318,8 +1318,8 @@ impl<'cx, 'tcx> SelectionContext<'cx, 'tcx> {
         let is_global =
             |cand: &ty::PolyTraitRef<'_>| cand.is_global() && !cand.has_late_bound_regions();
 
-        // (*) Prefer `BuiltinCandidate { has_nested: false }` and `DiscriminantKindCandidate`
-        // to anything else.
+        // (*) Prefer `BuiltinCandidate { has_nested: false }`, `PointeeCandidate`,
+        // and `DiscriminantKindCandidate` to anything else.
         //
         // This is a fix for #53123 and prevents winnowing from accidentally extending the
         // lifetime of a variable.
@@ -1332,8 +1332,18 @@ impl<'cx, 'tcx> SelectionContext<'cx, 'tcx> {
             }
 
             // (*)
-            (BuiltinCandidate { has_nested: false } | DiscriminantKindCandidate, _) => true,
-            (_, BuiltinCandidate { has_nested: false } | DiscriminantKindCandidate) => false,
+            (
+                BuiltinCandidate { has_nested: false }
+                | DiscriminantKindCandidate
+                | PointeeCandidate,
+                _,
+            ) => true,
+            (
+                _,
+                BuiltinCandidate { has_nested: false }
+                | DiscriminantKindCandidate
+                | PointeeCandidate,
+            ) => false,
 
             (ParamCandidate(other), ParamCandidate(victim)) => {
                 if other.value == victim.value && victim.constness == Constness::NotConst {

compiler/rustc_ty_utils/src/instance.rs

@@ -275,7 +275,8 @@ fn resolve_associated_item<'tcx>(
         traits::ImplSource::AutoImpl(..)
         | traits::ImplSource::Param(..)
         | traits::ImplSource::TraitAlias(..)
-        | traits::ImplSource::DiscriminantKind(..) => None,
+        | traits::ImplSource::DiscriminantKind(..)
+        | traits::ImplSource::Pointee(..) => None,
     })
 }
 

compiler/rustc_typeck/src/coherence/mod.rs

@@ -48,7 +48,20 @@ fn enforce_trait_manually_implementable(
     let did = Some(trait_def_id);
     let li = tcx.lang_items();
 
-    // Disallow *all* explicit impls of `DiscriminantKind`, `Sized` and `Unsize` for now.
+    // Disallow *all* explicit impls of `Pointee`, `DiscriminantKind`, `Sized` and `Unsize` for now.
+    if did == li.pointee_trait() {
+        let span = impl_header_span(tcx, impl_def_id);
+        struct_span_err!(
+            tcx.sess,
+            span,
+            E0322,
+            "explicit impls for the `Pointee` trait are not permitted"
+        )
+        .span_label(span, "impl of 'Pointee' not allowed")
+        .emit();
+        return;
+    }
+
     if did == li.discriminant_kind_trait() {
         let span = impl_header_span(tcx, impl_def_id);
         struct_span_err!(