Rollup of 8 pull requests

compiler/rustc_const_eval/src/interpret/place.rs

@@ -700,8 +700,13 @@ where
             assert_eq!(src.layout.size, dest.layout.size);
         }
 
+        // Setting `nonoverlapping` here only has an effect when we don't hit the fast-path above,
+        // but that should at least match what LLVM does where `memcpy` is also only used when the
+        // type does not have Scalar/ScalarPair layout.
+        // (Or as the `Assign` docs put it, assignments "not producing primitives" must be
+        // non-overlapping.)
         self.mem_copy(
-            src.ptr, src.align, dest.ptr, dest.align, dest_size, /*nonoverlapping*/ false,
+            src.ptr, src.align, dest.ptr, dest.align, dest_size, /*nonoverlapping*/ true,
         )
     }
 

compiler/rustc_hir_analysis/src/check/check.rs

@@ -722,7 +722,14 @@ pub(super) fn check_specialization_validity<'tcx>(
     let result = opt_result.unwrap_or(Ok(()));
 
     if let Err(parent_impl) = result {
-        report_forbidden_specialization(tcx, impl_item, parent_impl);
+        if !tcx.is_impl_trait_in_trait(impl_item) {
+            report_forbidden_specialization(tcx, impl_item, parent_impl);
+        } else {
+            tcx.sess.delay_span_bug(
+                DUMMY_SP,
+                format!("parent item: {:?} not marked as default", parent_impl),
+            );
+        }
     }
 }
 
@@ -1485,7 +1492,9 @@ fn opaque_type_cycle_error(
                 }
 
                 for closure_def_id in visitor.closures {
-                    let Some(closure_local_did) = closure_def_id.as_local() else { continue; };
+                    let Some(closure_local_did) = closure_def_id.as_local() else {
+                        continue;
+                    };
                     let typeck_results = tcx.typeck(closure_local_did);
 
                     let mut label_match = |ty: Ty<'_>, span| {

compiler/rustc_middle/src/ty/layout.rs

@@ -1110,12 +1110,11 @@ where
 ///
 /// This takes two primary parameters:
 ///
-/// * `codegen_fn_attr_flags` - these are flags calculated as part of the
-///   codegen attrs for a defined function. For function pointers this set of
-///   flags is the empty set. This is only applicable for Rust-defined
-///   functions, and generally isn't needed except for small optimizations where
-///   we try to say a function which otherwise might look like it could unwind
-///   doesn't actually unwind (such as for intrinsics and such).
+/// * `fn_def_id` - the `DefId` of the function. If this is provided then we can
+///   determine more precisely if the function can unwind. If this is not provided
+///   then we will only infer whether the function can unwind or not based on the
+///   ABI of the function. For example, a function marked with `#[rustc_nounwind]`
+///   is known to not unwind even if it's using Rust ABI.
 ///
 /// * `abi` - this is the ABI that the function is defined with. This is the
 ///   primary factor for determining whether a function can unwind or not.
@@ -1147,11 +1146,6 @@ where
 ///   aborts the process.
 /// * This affects whether functions have the LLVM `nounwind` attribute, which
 ///   affects various optimizations and codegen.
-///
-/// FIXME: this is actually buggy with respect to Rust functions. Rust functions
-/// compiled with `-Cpanic=unwind` and referenced from another crate compiled
-/// with `-Cpanic=abort` will look like they can't unwind when in fact they
-/// might (from a foreign exception or similar).
 #[inline]
 #[tracing::instrument(level = "debug", skip(tcx))]
 pub fn fn_can_unwind(tcx: TyCtxt<'_>, fn_def_id: Option<DefId>, abi: SpecAbi) -> bool {

compiler/rustc_monomorphize/src/collector.rs

@@ -231,8 +231,8 @@ impl<'tcx> UsageMap<'tcx> {
         assert!(self.used_map.insert(user_item, used_items).is_none());
     }
 
-    pub fn get_user_items(&self, item: MonoItem<'tcx>) -> Option<&[MonoItem<'tcx>]> {
-        self.user_map.get(&item).map(|items| items.as_slice())
+    pub fn get_user_items(&self, item: MonoItem<'tcx>) -> &[MonoItem<'tcx>] {
+        self.user_map.get(&item).map(|items| items.as_slice()).unwrap_or(&[])
     }
 
     /// Internally iterate over all inlined items used by `item`.

compiler/rustc_monomorphize/src/partitioning.rs

@@ -427,9 +427,9 @@ fn merge_codegen_units<'tcx>(
         // zero-padded suffixes, which means they are automatically sorted by
         // names. The numeric suffix width depends on the number of CGUs, which
         // is always greater than zero:
-        // - [1,9]     CGUS: `0`, `1`, `2`, ...
-        // - [10,99]   CGUS: `00`, `01`, `02`, ...
-        // - [100,999] CGUS: `000`, `001`, `002`, ...
+        // - [1,9]     CGUs: `0`, `1`, `2`, ...
+        // - [10,99]   CGUs: `00`, `01`, `02`, ...
+        // - [100,999] CGUs: `000`, `001`, `002`, ...
         // - etc.
         //
         // If we didn't zero-pad the sorted-by-name order would be `XYZ-cgu.0`,
@@ -458,29 +458,29 @@ fn internalize_symbols<'tcx>(
     /// used to keep track of that.
     #[derive(Clone, PartialEq, Eq, Debug)]
     enum MonoItemPlacement {
-        SingleCgu { cgu_name: Symbol },
+        SingleCgu(Symbol),
         MultipleCgus,
     }
 
     let mut mono_item_placements = FxHashMap::default();
     let single_codegen_unit = codegen_units.len() == 1;
 
     if !single_codegen_unit {
-        for cgu in codegen_units.iter_mut() {
+        for cgu in codegen_units.iter() {
             for item in cgu.items().keys() {
                 // If there is more than one codegen unit, we need to keep track
                 // in which codegen units each monomorphization is placed.
                 match mono_item_placements.entry(*item) {
                     Entry::Occupied(e) => {
                         let placement = e.into_mut();
                         debug_assert!(match *placement {
-                            MonoItemPlacement::SingleCgu { cgu_name } => cgu_name != cgu.name(),
+                            MonoItemPlacement::SingleCgu(cgu_name) => cgu_name != cgu.name(),
                             MonoItemPlacement::MultipleCgus => true,
                         });
                         *placement = MonoItemPlacement::MultipleCgus;
                     }
                     Entry::Vacant(e) => {
-                        e.insert(MonoItemPlacement::SingleCgu { cgu_name: cgu.name() });
+                        e.insert(MonoItemPlacement::SingleCgu(cgu.name()));
                     }
                 }
             }
@@ -490,7 +490,7 @@ fn internalize_symbols<'tcx>(
     // For each internalization candidates in each codegen unit, check if it is
     // used from outside its defining codegen unit.
     for cgu in codegen_units {
-        let home_cgu = MonoItemPlacement::SingleCgu { cgu_name: cgu.name() };
+        let home_cgu = MonoItemPlacement::SingleCgu(cgu.name());
 
         for (item, linkage_and_visibility) in cgu.items_mut() {
             if !internalization_candidates.contains(item) {
@@ -501,20 +501,20 @@ fn internalize_symbols<'tcx>(
             if !single_codegen_unit {
                 debug_assert_eq!(mono_item_placements[item], home_cgu);
 
-                if let Some(user_items) = cx.usage_map.get_user_items(*item) {
-                    if user_items
-                        .iter()
-                        .filter_map(|user_item| {
-                            // Some user mono items might not have been
-                            // instantiated. We can safely ignore those.
-                            mono_item_placements.get(user_item)
-                        })
-                        .any(|placement| *placement != home_cgu)
-                    {
-                        // Found a user from another CGU, so skip to the next item
-                        // without marking this one as internal.
-                        continue;
-                    }
+                if cx
+                    .usage_map
+                    .get_user_items(*item)
+                    .iter()
+                    .filter_map(|user_item| {
+                        // Some user mono items might not have been
+                        // instantiated. We can safely ignore those.
+                        mono_item_placements.get(user_item)
+                    })
+                    .any(|placement| *placement != home_cgu)
+                {
+                    // Found a user from another CGU, so skip to the next item
+                    // without marking this one as internal.
+                    continue;
                 }
             }
 

compiler/rustc_span/src/symbol.rs

@@ -759,7 +759,6 @@ symbols! {
         from_desugaring,
         from_fn,
         from_iter,
-        from_method,
         from_output,
         from_residual,
         from_size_align_unchecked,

compiler/rustc_trait_selection/src/traits/error_reporting/on_unimplemented.rs

@@ -41,7 +41,6 @@ pub trait TypeErrCtxtExt<'tcx> {
 static ALLOWED_FORMAT_SYMBOLS: &[Symbol] = &[
     kw::SelfUpper,
     sym::ItemContext,
-    sym::from_method,
     sym::from_desugaring,
     sym::direct,
     sym::cause,
@@ -172,23 +171,6 @@ impl<'tcx> TypeErrCtxtExt<'tcx> for TypeErrCtxt<'_, 'tcx> {
             }
         }
 
-        if let ObligationCauseCode::ItemObligation(item)
-        | ObligationCauseCode::BindingObligation(item, _)
-        | ObligationCauseCode::ExprItemObligation(item, ..)
-        | ObligationCauseCode::ExprBindingObligation(item, ..) = *obligation.cause.code()
-        {
-            // FIXME: maybe also have some way of handling methods
-            // from other traits? That would require name resolution,
-            // which we might want to be some sort of hygienic.
-            //
-            // Currently I'm leaving it for what I need for `try`.
-            if self.tcx.trait_of_item(item) == Some(trait_ref.def_id) {
-                let method = self.tcx.item_name(item);
-                flags.push((sym::from_method, None));
-                flags.push((sym::from_method, Some(method.to_string())));
-            }
-        }
-
         if let Some(k) = obligation.cause.span.desugaring_kind() {
             flags.push((sym::from_desugaring, None));
             flags.push((sym::from_desugaring, Some(format!("{:?}", k))));
@@ -672,7 +654,7 @@ impl<'tcx> OnUnimplementedFormatString {
                             None => {
                                 if let Some(val) = options.get(&s) {
                                     val
-                                } else if s == sym::from_desugaring || s == sym::from_method {
+                                } else if s == sym::from_desugaring {
                                     // don't break messages using these two arguments incorrectly
                                     &empty_string
                                 } else if s == sym::ItemContext {

src/librustdoc/clean/mod.rs

@@ -1479,8 +1479,97 @@ pub(crate) fn clean_middle_assoc_item<'tcx>(
     Item::from_def_id_and_parts(assoc_item.def_id, Some(assoc_item.name), kind, cx)
 }
 
+/// The goal of this function is to return the first `Path` which is not private (ie not private
+/// or `doc(hidden)`). If it's not possible, it'll return the "end type".
+///
+/// If the path is not a re-export or is public, it'll return `None`.
+fn first_non_private(
+    cx: &mut DocContext<'_>,
+    hir_id: hir::HirId,
+    path: &hir::Path<'_>,
+) -> Option<Path> {
+    let (parent_def_id, mut ident) = match &path.segments[..] {
+        [] => return None,
+        // Relative paths are available in the same scope as the owner.
+        [leaf] => (cx.tcx.local_parent(hir_id.owner.def_id), leaf.ident),
+        // So are self paths.
+        [parent, leaf] if parent.ident.name == kw::SelfLower => {
+            (cx.tcx.local_parent(hir_id.owner.def_id), leaf.ident)
+        }
+        // Crate paths are not. We start from the crate root.
+        [parent, leaf] if matches!(parent.ident.name, kw::Crate | kw::PathRoot) => {
+            (LOCAL_CRATE.as_def_id().as_local()?, leaf.ident)
+        }
+        [parent, leaf] if parent.ident.name == kw::Super => {
+            let parent_mod = cx.tcx.parent_module(hir_id);
+            if let Some(super_parent) = cx.tcx.opt_local_parent(parent_mod) {
+                (super_parent, leaf.ident)
+            } else {
+                // If we can't find the parent of the parent, then the parent is already the crate.
+                (LOCAL_CRATE.as_def_id().as_local()?, leaf.ident)
+            }
+        }
+        // Absolute paths are not. We start from the parent of the item.
+        [.., parent, leaf] => (parent.res.opt_def_id()?.as_local()?, leaf.ident),
+    };
+    let target_def_id = path.res.opt_def_id()?;
+    // First we try to get the `DefId` of the item.
+    for child in
+        cx.tcx.module_children_local(parent_def_id).iter().filter(move |c| c.ident == ident)
+    {
+        if let Res::Def(DefKind::Ctor(..), _) | Res::SelfCtor(..) = child.res {
+            continue;
+        }
+
+        if let Some(def_id) = child.res.opt_def_id() && target_def_id == def_id {
+            let mut last_path_res = None;
+            'reexps: for reexp in child.reexport_chain.iter() {
+                if let Some(use_def_id) = reexp.id() &&
+                    let Some(local_use_def_id) = use_def_id.as_local()
+                {
+                    let hir = cx.tcx.hir();
+                    for item_id in hir.module_items(cx.tcx.local_parent(local_use_def_id)) {
+                        let item = hir.item(item_id);
+                        if item.ident == ident && let hir::ItemKind::Use(path, _) = item.kind {
+                            for res in &path.res {
+                                if let Res::Def(DefKind::Ctor(..), _) | Res::SelfCtor(..) = res {
+                                    continue;
+                                }
+                                if !cx.tcx.is_doc_hidden(use_def_id) &&
+                                    cx.tcx.local_visibility(local_use_def_id).is_public() {
+                                    break 'reexps;
+                                }
+                                ident = path.segments.last().unwrap().ident;
+                                last_path_res = Some((path, res));
+                                continue 'reexps;
+                            }
+                        }
+                    }
+                }
+            }
+            if !child.reexport_chain.is_empty() {
+                // So in here, we use the data we gathered from iterating the reexports. If
+                // `last_path_res` is set, it can mean two things:
+                //
+                // 1. We found a public reexport.
+                // 2. We didn't find a public reexport so it's the "end type" path.
+                if let Some((path, res)) = last_path_res {
+                    let path = hir::Path { segments: path.segments, res: *res, span: path.span };
+                    return Some(clean_path(&path, cx));
+                }
+                // If `last_path_res` is `None`, it can mean two things:
+                //
+                // 1. The re-export is public, no need to change anything, just use the path as is.
+                // 2. Nothing was found, so let's just return the original path.
+                return None;
+            }
+        }
+    }
+    None
+}
+
 fn clean_qpath<'tcx>(hir_ty: &hir::Ty<'tcx>, cx: &mut DocContext<'tcx>) -> Type {
-    let hir::Ty { hir_id: _, span, ref kind } = *hir_ty;
+    let hir::Ty { hir_id, span, ref kind } = *hir_ty;
     let hir::TyKind::Path(qpath) = kind else { unreachable!() };
 
     match qpath {
@@ -1497,7 +1586,12 @@ fn clean_qpath<'tcx>(hir_ty: &hir::Ty<'tcx>, cx: &mut DocContext<'tcx>) -> Type
             if let Some(expanded) = maybe_expand_private_type_alias(cx, path) {
                 expanded
             } else {
-                let path = clean_path(path, cx);
+                // First we check if it's a private re-export.
+                let path = if let Some(path) = first_non_private(cx, hir_id, &path) {
+                    path
+                } else {
+                    clean_path(path, cx)
+                };
                 resolve_type(cx, path)
             }
         }
@@ -1649,7 +1743,7 @@ fn maybe_expand_private_type_alias<'tcx>(
         }
     }
 
-    Some(cx.enter_alias(substs, def_id.to_def_id(), |cx| clean_ty(ty, cx)))
+    Some(cx.enter_alias(substs, def_id.to_def_id(), |cx| clean_ty(&ty, cx)))
 }
 
 pub(crate) fn clean_ty<'tcx>(ty: &hir::Ty<'tcx>, cx: &mut DocContext<'tcx>) -> Type {

src/tools/miri/tests/fail/overlapping_assignment.rs

@@ -0,0 +1,23 @@
+#![feature(core_intrinsics)]
+#![feature(custom_mir)]
+
+use std::intrinsics::mir::*;
+
+// It's not that easy to fool the MIR validity check
+// which wants to prevent overlapping assignments...
+// So we use two separate pointer arguments, and then arrange for them to alias.
+#[custom_mir(dialect = "runtime", phase = "optimized")]
+pub fn self_copy(ptr1: *mut [i32; 4], ptr2: *mut [i32; 4]) {
+    mir! {
+        {
+            *ptr1 = *ptr2; //~ERROR: overlapping ranges
+            Return()
+        }
+    }
+}
+
+pub fn main() {
+    let mut x = [0; 4];
+    let ptr = std::ptr::addr_of_mut!(x);
+    self_copy(ptr, ptr);
+}

src/tools/miri/tests/fail/overlapping_assignment.stderr

@@ -0,0 +1,20 @@
+error: Undefined Behavior: `copy_nonoverlapping` called on overlapping ranges
+  --> $DIR/overlapping_assignment.rs:LL:CC
+   |
+LL |             *ptr1 = *ptr2;
+   |             ^^^^^^^^^^^^^ `copy_nonoverlapping` called on overlapping ranges
+   |
+   = help: this indicates a bug in the program: it performed an invalid operation, and caused Undefined Behavior
+   = help: see https://doc.rust-lang.org/nightly/reference/behavior-considered-undefined.html for further information
+   = note: BACKTRACE:
+   = note: inside `self_copy` at $DIR/overlapping_assignment.rs:LL:CC
+note: inside `main`
+  --> $DIR/overlapping_assignment.rs:LL:CC
+   |
+LL |     self_copy(ptr, ptr);
+   |     ^^^^^^^^^^^^^^^^^^^
+
+note: some details are omitted, run with `MIRIFLAGS=-Zmiri-backtrace=full` for a verbose backtrace
+
+error: aborting due to previous error
+