Auto merge of rust-lang#123676 - GuillaumeGomez:rollup-1hurixy, r=GuillaumeGomez

Rollup of 8 pull requests

Successful merges:

 - rust-lang#123254 (Do not allocate for ZST ThinBox (attempt 2 using const_allocate))
 - rust-lang#123626 (Add MC/DC support to coverage test tools)
 - rust-lang#123638 (rustdoc: synthetic auto: filter out clauses from the implementor's ParamEnv)
 - rust-lang#123653 (Split `non_local_definitions` lint tests in separate test files)
 - rust-lang#123658 (Stop making any assumption about the projections applied to the upvars in the `ByMoveBody` pass)
 - rust-lang#123662 (Don't rely on upvars being assigned just because coroutine-closure kind is assigned)
 - rust-lang#123665 (Fix typo in `Future::poll()` docs)
 - rust-lang#123672 (compiletest: unset `RUSTC_LOG_COLOR`)

r? `@ghost`
`@rustbot` modify labels: rollup

Author: bors

compiler/rustc_middle/src/ty/sty.rs

@@ -2231,7 +2231,7 @@ impl<'tcx> Ty<'tcx> {
     pub fn tuple_fields(self) -> &'tcx List<Ty<'tcx>> {
         match self.kind() {
             Tuple(args) => args,
-            _ => bug!("tuple_fields called on non-tuple"),
+            _ => bug!("tuple_fields called on non-tuple: {self:?}"),
         }
     }
 

compiler/rustc_mir_transform/src/coroutine/by_move_body.rs

@@ -281,14 +281,14 @@ impl<'tcx> MutVisitor<'tcx> for MakeByMoveBody<'tcx> {
         if place.local == ty::CAPTURE_STRUCT_LOCAL
             && let Some((&mir::ProjectionElem::Field(idx, _), projection)) =
                 place.projection.split_first()
-            && let Some(&(remapped_idx, remapped_ty, needs_deref, additional_projections)) =
+            && let Some(&(remapped_idx, remapped_ty, needs_deref, bridging_projections)) =
                 self.field_remapping.get(&idx)
         {
             // As noted before, if the parent closure captures a field by value, and
             // the child captures a field by ref, then for the by-move body we're
             // generating, we also are taking that field by value. Peel off a deref,
-            // since a layer of reffing has now become redundant.
-            let final_deref = if needs_deref {
+            // since a layer of ref'ing has now become redundant.
+            let final_projections = if needs_deref {
                 let Some((mir::ProjectionElem::Deref, projection)) = projection.split_first()
                 else {
                     bug!(
@@ -302,20 +302,18 @@ impl<'tcx> MutVisitor<'tcx> for MakeByMoveBody<'tcx> {
                 projection
             };
 
-            // The only thing that should be left is a deref, if the parent captured
-            // an upvar by-ref.
-            std::assert_matches::assert_matches!(final_deref, [] | [mir::ProjectionElem::Deref]);
-
-            // For all of the additional projections that come out of precise capturing,
-            // re-apply these projections.
-            let additional_projections =
-                additional_projections.iter().map(|elem| match elem.kind {
-                    ProjectionKind::Deref => mir::ProjectionElem::Deref,
-                    ProjectionKind::Field(idx, VariantIdx::ZERO) => {
-                        mir::ProjectionElem::Field(idx, elem.ty)
-                    }
-                    _ => unreachable!("precise captures only through fields and derefs"),
-                });
+            // These projections are applied in order to "bridge" the local that we are
+            // currently transforming *from* the old upvar that the by-ref coroutine used
+            // to capture *to* the upvar of the parent coroutine-closure. For example, if
+            // the parent captures `&s` but the child captures `&(s.field)`, then we will
+            // apply a field projection.
+            let bridging_projections = bridging_projections.iter().map(|elem| match elem.kind {
+                ProjectionKind::Deref => mir::ProjectionElem::Deref,
+                ProjectionKind::Field(idx, VariantIdx::ZERO) => {
+                    mir::ProjectionElem::Field(idx, elem.ty)
+                }
+                _ => unreachable!("precise captures only through fields and derefs"),
+            });
 
             // We start out with an adjusted field index (and ty), representing the
             // upvar that we get from our parent closure. We apply any of the additional
@@ -326,8 +324,8 @@ impl<'tcx> MutVisitor<'tcx> for MakeByMoveBody<'tcx> {
                 projection: self.tcx.mk_place_elems_from_iter(
                     [mir::ProjectionElem::Field(remapped_idx, remapped_ty)]
                         .into_iter()
-                        .chain(additional_projections)
-                        .chain(final_deref.iter().copied()),
+                        .chain(bridging_projections)
+                        .chain(final_projections.iter().copied()),
                 ),
             };
         }

compiler/rustc_trait_selection/src/solve/assembly/structural_traits.rs

@@ -292,7 +292,9 @@ pub(in crate::solve) fn extract_tupled_inputs_and_output_from_callable<'tcx>(
             let kind_ty = args.kind_ty();
             let sig = args.coroutine_closure_sig().skip_binder();
 
-            let coroutine_ty = if let Some(closure_kind) = kind_ty.to_opt_closure_kind() {
+            let coroutine_ty = if let Some(closure_kind) = kind_ty.to_opt_closure_kind()
+                && !args.tupled_upvars_ty().is_ty_var()
+            {
                 if !closure_kind.extends(goal_kind) {
                     return Err(NoSolution);
                 }
@@ -401,7 +403,9 @@ pub(in crate::solve) fn extract_tupled_inputs_and_output_from_async_callable<'tc
             let kind_ty = args.kind_ty();
             let sig = args.coroutine_closure_sig().skip_binder();
             let mut nested = vec![];
-            let coroutine_ty = if let Some(closure_kind) = kind_ty.to_opt_closure_kind() {
+            let coroutine_ty = if let Some(closure_kind) = kind_ty.to_opt_closure_kind()
+                && !args.tupled_upvars_ty().is_ty_var()
+            {
                 if !closure_kind.extends(goal_kind) {
                     return Err(NoSolution);
                 }

compiler/rustc_trait_selection/src/solve/normalizes_to/mod.rs

@@ -487,6 +487,11 @@ impl<'tcx> assembly::GoalKind<'tcx> for NormalizesTo<'tcx> {
             bug!();
         };
 
+        // Bail if the upvars haven't been constrained.
+        if tupled_upvars_ty.expect_ty().is_ty_var() {
+            return ecx.evaluate_added_goals_and_make_canonical_response(Certainty::AMBIGUOUS);
+        }
+
         let Some(closure_kind) = closure_fn_kind_ty.expect_ty().to_opt_closure_kind() else {
             // We don't need to worry about the self type being an infer var.
             return Err(NoSolution);

compiler/rustc_trait_selection/src/traits/project.rs

@@ -1601,7 +1601,10 @@ fn confirm_closure_candidate<'cx, 'tcx>(
                 // If we know the kind and upvars, use that directly.
                 // Otherwise, defer to `AsyncFnKindHelper::Upvars` to delay
                 // the projection, like the `AsyncFn*` traits do.
-                let output_ty = if let Some(_) = kind_ty.to_opt_closure_kind() {
+                let output_ty = if let Some(_) = kind_ty.to_opt_closure_kind()
+                    // Fall back to projection if upvars aren't constrained
+                    && !args.tupled_upvars_ty().is_ty_var()
+                {
                     sig.to_coroutine_given_kind_and_upvars(
                         tcx,
                         args.parent_args(),
@@ -1731,7 +1734,10 @@ fn confirm_async_closure_candidate<'cx, 'tcx>(
 
             let term = match item_name {
                 sym::CallOnceFuture | sym::CallRefFuture => {
-                    if let Some(closure_kind) = kind_ty.to_opt_closure_kind() {
+                    if let Some(closure_kind) = kind_ty.to_opt_closure_kind()
+                        // Fall back to projection if upvars aren't constrained
+                        && !args.tupled_upvars_ty().is_ty_var()
+                    {
                         if !closure_kind.extends(goal_kind) {
                             bug!("we should not be confirming if the closure kind is not met");
                         }

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

@@ -400,39 +400,36 @@ impl<'cx, 'tcx> SelectionContext<'cx, 'tcx> {
                 }
             }
             ty::CoroutineClosure(def_id, args) => {
+                let args = args.as_coroutine_closure();
                 let is_const = self.tcx().is_const_fn_raw(def_id);
-                match self.infcx.closure_kind(self_ty) {
-                    Some(closure_kind) => {
-                        let no_borrows = match self
-                            .infcx
-                            .shallow_resolve(args.as_coroutine_closure().tupled_upvars_ty())
-                            .kind()
-                        {
-                            ty::Tuple(tys) => tys.is_empty(),
-                            ty::Error(_) => false,
-                            _ => bug!("tuple_fields called on non-tuple"),
-                        };
-                        // A coroutine-closure implements `FnOnce` *always*, since it may
-                        // always be called once. It additionally implements `Fn`/`FnMut`
-                        // only if it has no upvars (therefore no borrows from the closure
-                        // that would need to be represented with a lifetime) and if the
-                        // closure kind permits it.
-                        // FIXME(async_closures): Actually, it could also implement `Fn`/`FnMut`
-                        // if it takes all of its upvars by copy, and none by ref. This would
-                        // require us to record a bit more information during upvar analysis.
-                        if no_borrows && closure_kind.extends(kind) {
-                            candidates.vec.push(ClosureCandidate { is_const });
-                        } else if kind == ty::ClosureKind::FnOnce {
-                            candidates.vec.push(ClosureCandidate { is_const });
-                        }
+                if let Some(closure_kind) = self.infcx.closure_kind(self_ty)
+                    // Ambiguity if upvars haven't been constrained yet
+                    && !args.tupled_upvars_ty().is_ty_var()
+                {
+                    let no_borrows = match args.tupled_upvars_ty().kind() {
+                        ty::Tuple(tys) => tys.is_empty(),
+                        ty::Error(_) => false,
+                        _ => bug!("tuple_fields called on non-tuple"),
+                    };
+                    // A coroutine-closure implements `FnOnce` *always*, since it may
+                    // always be called once. It additionally implements `Fn`/`FnMut`
+                    // only if it has no upvars (therefore no borrows from the closure
+                    // that would need to be represented with a lifetime) and if the
+                    // closure kind permits it.
+                    // FIXME(async_closures): Actually, it could also implement `Fn`/`FnMut`
+                    // if it takes all of its upvars by copy, and none by ref. This would
+                    // require us to record a bit more information during upvar analysis.
+                    if no_borrows && closure_kind.extends(kind) {
+                        candidates.vec.push(ClosureCandidate { is_const });
+                    } else if kind == ty::ClosureKind::FnOnce {
+                        candidates.vec.push(ClosureCandidate { is_const });
                     }
-                    None => {
-                        if kind == ty::ClosureKind::FnOnce {
-                            candidates.vec.push(ClosureCandidate { is_const });
-                        } else {
-                            // This stays ambiguous until kind+upvars are determined.
-                            candidates.ambiguous = true;
-                        }
+                } else {
+                    if kind == ty::ClosureKind::FnOnce {
+                        candidates.vec.push(ClosureCandidate { is_const });
+                    } else {
+                        // This stays ambiguous until kind+upvars are determined.
+                        candidates.ambiguous = true;
                     }
                 }
             }

library/alloc/src/boxed/thin.rs

@@ -4,10 +4,14 @@
 use crate::alloc::{self, Layout, LayoutError};
 use core::error::Error;
 use core::fmt::{self, Debug, Display, Formatter};
+#[cfg(not(no_global_oom_handling))]
+use core::intrinsics::const_allocate;
 use core::marker::PhantomData;
 #[cfg(not(no_global_oom_handling))]
 use core::marker::Unsize;
-use core::mem::{self, SizedTypeProperties};
+use core::mem;
+#[cfg(not(no_global_oom_handling))]
+use core::mem::SizedTypeProperties;
 use core::ops::{Deref, DerefMut};
 use core::ptr::Pointee;
 use core::ptr::{self, NonNull};
@@ -109,9 +113,14 @@ impl<Dyn: ?Sized> ThinBox<Dyn> {
     where
         T: Unsize<Dyn>,
     {
-        let meta = ptr::metadata(&value as &Dyn);
-        let ptr = WithOpaqueHeader::new(meta, value);
-        ThinBox { ptr, _marker: PhantomData }
+        if mem::size_of::<T>() == 0 {
+            let ptr = WithOpaqueHeader::new_unsize_zst::<Dyn, T>(value);
+            ThinBox { ptr, _marker: PhantomData }
+        } else {
+            let meta = ptr::metadata(&value as &Dyn);
+            let ptr = WithOpaqueHeader::new(meta, value);
+            ThinBox { ptr, _marker: PhantomData }
+        }
     }
 }
 
@@ -200,6 +209,16 @@ impl WithOpaqueHeader {
         Self(ptr.0)
     }
 
+    #[cfg(not(no_global_oom_handling))]
+    fn new_unsize_zst<Dyn, T>(value: T) -> Self
+    where
+        Dyn: ?Sized,
+        T: Unsize<Dyn>,
+    {
+        let ptr = WithHeader::<<Dyn as Pointee>::Metadata>::new_unsize_zst::<Dyn, T>(value);
+        Self(ptr.0)
+    }
+
     fn try_new<H, T>(header: H, value: T) -> Result<Self, core::alloc::AllocError> {
         WithHeader::try_new(header, value).map(|ptr| Self(ptr.0))
     }
@@ -288,6 +307,58 @@ impl<H> WithHeader<H> {
         }
     }
 
+    // `Dyn` is `?Sized` type like `[u32]`, and `T` is ZST type like `[u32; 0]`.
+    #[cfg(not(no_global_oom_handling))]
+    fn new_unsize_zst<Dyn, T>(value: T) -> WithHeader<H>
+    where
+        Dyn: Pointee<Metadata = H> + ?Sized,
+        T: Unsize<Dyn>,
+    {
+        assert!(mem::size_of::<T>() == 0);
+
+        const fn max(a: usize, b: usize) -> usize {
+            if a > b { a } else { b }
+        }
+
+        // Compute a pointer to the right metadata. This will point to the beginning
+        // of the header, past the padding, so the assigned type makes sense.
+        // It also ensures that the address at the end of the header is sufficiently
+        // aligned for T.
+        let alloc: &<Dyn as Pointee>::Metadata = const {
+            // FIXME: just call `WithHeader::alloc_layout` with size reset to 0.
+            // Currently that's blocked on `Layout::extend` not being `const fn`.
+
+            let alloc_align =
+                max(mem::align_of::<T>(), mem::align_of::<<Dyn as Pointee>::Metadata>());
+
+            let alloc_size =
+                max(mem::align_of::<T>(), mem::size_of::<<Dyn as Pointee>::Metadata>());
+
+            unsafe {
+                // SAFETY: align is power of two because it is the maximum of two alignments.
+                let alloc: *mut u8 = const_allocate(alloc_size, alloc_align);
+
+                let metadata_offset =
+                    alloc_size.checked_sub(mem::size_of::<<Dyn as Pointee>::Metadata>()).unwrap();
+                // SAFETY: adding offset within the allocation.
+                let metadata_ptr: *mut <Dyn as Pointee>::Metadata =
+                    alloc.add(metadata_offset).cast();
+                // SAFETY: `*metadata_ptr` is within the allocation.
+                metadata_ptr.write(ptr::metadata::<Dyn>(ptr::dangling::<T>() as *const Dyn));
+
+                // SAFETY: we have just written the metadata.
+                &*(metadata_ptr)
+            }
+        };
+
+        // SAFETY: `alloc` points to `<Dyn as Pointee>::Metadata`, so addition stays in-bounds.
+        let value_ptr =
+            unsafe { (alloc as *const <Dyn as Pointee>::Metadata).add(1) }.cast::<T>().cast_mut();
+        debug_assert!(value_ptr.is_aligned());
+        mem::forget(value);
+        WithHeader(NonNull::new(value_ptr.cast()).unwrap(), PhantomData)
+    }
+
     // Safety:
     // - Assumes that either `value` can be dereferenced, or is the
     //   `NonNull::dangling()` we use when both `T` and `H` are ZSTs.
@@ -300,20 +371,19 @@ impl<H> WithHeader<H> {
 
         impl<H> Drop for DropGuard<H> {
             fn drop(&mut self) {
+                // All ZST are allocated statically.
+                if self.value_layout.size() == 0 {
+                    return;
+                }
+
                 unsafe {
                     // SAFETY: Layout must have been computable if we're in drop
                     let (layout, value_offset) =
                         WithHeader::<H>::alloc_layout(self.value_layout).unwrap_unchecked();
 
-                    // Note: Don't deallocate if the layout size is zero, because the pointer
-                    // didn't come from the allocator.
-                    if layout.size() != 0 {
-                        alloc::dealloc(self.ptr.as_ptr().sub(value_offset), layout);
-                    } else {
-                        debug_assert!(
-                            value_offset == 0 && H::IS_ZST && self.value_layout.size() == 0
-                        );
-                    }
+                    // Since we only allocate for non-ZSTs, the layout size cannot be zero.
+                    debug_assert!(layout.size() != 0);
+                    alloc::dealloc(self.ptr.as_ptr().sub(value_offset), layout);
                 }
             }
         }

library/alloc/src/lib.rs

@@ -108,8 +108,10 @@
 #![feature(const_box)]
 #![feature(const_cow_is_borrowed)]
 #![feature(const_eval_select)]
+#![feature(const_heap)]
 #![feature(const_maybe_uninit_as_mut_ptr)]
 #![feature(const_maybe_uninit_write)]
+#![feature(const_option)]
 #![feature(const_pin)]
 #![feature(const_refs_to_cell)]
 #![feature(const_size_of_val)]

library/core/src/future/future.rs

@@ -81,7 +81,7 @@ pub trait Future {
     /// An implementation of `poll` should strive to return quickly, and should
     /// not block. Returning quickly prevents unnecessarily clogging up
     /// threads or event loops. If it is known ahead of time that a call to
-    /// `poll` may end up taking awhile, the work should be offloaded to a
+    /// `poll` may end up taking a while, the work should be offloaded to a
     /// thread pool (or something similar) to ensure that `poll` can return
     /// quickly.
     ///

src/librustdoc/clean/auto_trait.rs

@@ -168,7 +168,7 @@ fn clean_param_env<'tcx>(
 
     // FIXME(#111101): Incorporate the explicit predicates of the item here...
     let item_predicates: FxIndexSet<_> =
-        tcx.predicates_of(item_def_id).predicates.iter().map(|(pred, _)| pred).collect();
+        tcx.param_env(item_def_id).caller_bounds().iter().collect();
     let where_predicates = param_env
         .caller_bounds()
         .iter()