Auto merge of #65454 - tmandry:rollup-0k6jiik, r=tmandry

Rollup of 14 pull requests

Successful merges:

 - #64603 (Reducing spurious unused lifetime warnings.)
 - #64623 (Remove last uses of gensyms)
 - #65235 (don't assume we can *always* find a return type hint in async fn)
 - #65242 (Fix suggestion to constrain trait for method to be found)
 - #65265 (Cleanup librustc mir err codes)
 - #65293 (Optimize `try_expand_impl_trait_type`)
 - #65307 (Try fix incorrect "explicit lifetime name needed")
 - #65308 (Add long error explanation for E0574)
 - #65353 (save-analysis: Don't ICE when resolving qualified type paths in struct members)
 - #65389 (Return `false` from `needs_drop` for all zero-sized arrays.)
 - #65402 (Add troubleshooting section to PGO chapter in rustc book.)
 - #65425 (Optimize `BitIter`)
 - #65438 (Organize `never_type`  tests)
 - #65444 (Implement AsRef<[T]> for List<T>)

Failed merges:

 - #65390 (Add long error explanation for E0576)

r? @ghost

src/doc/rustc/src/profile-guided-optimization.md

@@ -125,6 +125,17 @@ RUSTFLAGS="-Cprofile-use=/tmp/pgo-data/merged.profdata" \
     cargo build --release --target=x86_64-unknown-linux-gnu
 ```
 
+### Troubleshooting
+
+- It is recommended to pass `-Cllvm-args=-pgo-warn-missing-function` during the
+  `-Cprofile-use` phase. LLVM by default does not warn if it cannot find
+  profiling data for a given function. Enabling this warning will make it
+  easier to spot errors in your setup.
+
+- There is a [known issue](https://github.com/rust-lang/cargo/issues/7416) in
+  Cargo prior to version 1.39 that will prevent PGO from working correctly. Be
+  sure to use Cargo 1.39 or newer when doing PGO.
+
 ## Further Reading
 
 `rustc`'s PGO support relies entirely on LLVM's implementation of the feature

src/librustc/hir/lowering.rs

@@ -3291,10 +3291,14 @@ impl<'a> LoweringContext<'a> {
                 let id = self.sess.next_node_id();
                 self.new_named_lifetime(id, span, hir::LifetimeName::Error)
             }
-            // This is the normal case.
-            AnonymousLifetimeMode::PassThrough => self.new_implicit_lifetime(span),
-
-            AnonymousLifetimeMode::ReportError => self.new_error_lifetime(None, span),
+            // `PassThrough` is the normal case.
+            // `new_error_lifetime`, which would usually be used in the case of `ReportError`,
+            // is unsuitable here, as these can occur from missing lifetime parameters in a
+            // `PathSegment`, for which there is no associated `'_` or `&T` with no explicit
+            // lifetime. Instead, we simply create an implicit lifetime, which will be checked
+            // later, at which point a suitable error will be emitted.
+          | AnonymousLifetimeMode::PassThrough
+          | AnonymousLifetimeMode::ReportError => self.new_implicit_lifetime(span),
         }
     }
 

src/librustc/middle/resolve_lifetime.rs

@@ -708,15 +708,22 @@ impl<'a, 'tcx> Visitor<'tcx> for LifetimeContext<'a, 'tcx> {
                     match param.kind {
                         GenericParamKind::Lifetime { .. } => {
                             let (name, reg) = Region::early(&self.tcx.hir(), &mut index, &param);
+                            let def_id = if let Region::EarlyBound(_ ,def_id , _) = reg {
+                                def_id
+                            } else {
+                                bug!();
+                            };
                             if let hir::ParamName::Plain(param_name) = name {
                                 if param_name.name == kw::UnderscoreLifetime {
                                     // Pick the elided lifetime "definition" if one exists
                                     // and use it to make an elision scope.
+                                    self.lifetime_uses.insert(def_id.clone(), LifetimeUseSet::Many);
                                     elision = Some(reg);
                                 } else {
                                     lifetimes.insert(name, reg);
                                 }
                             } else {
+                                self.lifetime_uses.insert(def_id.clone(), LifetimeUseSet::Many);
                                 lifetimes.insert(name, reg);
                             }
                         }
@@ -1615,7 +1622,6 @@ impl<'a, 'tcx> LifetimeContext<'a, 'tcx> {
                         _ => None,
                     } {
                         debug!("id = {:?} span = {:?} name = {:?}", id, span, name);
-
                         if name.name == kw::UnderscoreLifetime {
                             continue;
                         }

src/librustc/ty/mod.rs

@@ -701,6 +701,13 @@ impl<T> Deref for List<T> {
     type Target = [T];
     #[inline(always)]
     fn deref(&self) -> &[T] {
+        self.as_ref()
+    }
+}
+
+impl<T> AsRef<[T]> for List<T> {
+    #[inline(always)]
+    fn as_ref(&self) -> &[T] {
         unsafe {
             slice::from_raw_parts(self.data.as_ptr(), self.len)
         }

src/librustc/ty/print/pretty.rs

@@ -1483,7 +1483,7 @@ impl<F: fmt::Write> FmtPrinter<'_, 'tcx, F> {
         }
 
         // Replace any anonymous late-bound regions with named
-        // variants, using gensym'd identifiers, so that we can
+        // variants, using new unique identifiers, so that we can
         // clearly differentiate between named and unnamed regions in
         // the output. We'll probably want to tweak this over time to
         // decide just how much information to give.

src/librustc/ty/util.rs

@@ -697,6 +697,9 @@ impl<'tcx> TyCtxt<'tcx> {
             // that type, and when we finish expanding that type we remove the
             // its DefId.
             seen_opaque_tys: FxHashSet<DefId>,
+            // Cache of all expansions we've seen so far. This is a critical
+            // optimization for some large types produced by async fn trees.
+            expanded_cache: FxHashMap<(DefId, SubstsRef<'tcx>), Ty<'tcx>>,
             primary_def_id: DefId,
             found_recursion: bool,
             tcx: TyCtxt<'tcx>,
@@ -713,9 +716,16 @@ impl<'tcx> TyCtxt<'tcx> {
                 }
                 let substs = substs.fold_with(self);
                 if self.seen_opaque_tys.insert(def_id) {
-                    let generic_ty = self.tcx.type_of(def_id);
-                    let concrete_ty = generic_ty.subst(self.tcx, substs);
-                    let expanded_ty = self.fold_ty(concrete_ty);
+                    let expanded_ty = match self.expanded_cache.get(&(def_id, substs)) {
+                        Some(expanded_ty) => expanded_ty,
+                        None => {
+                            let generic_ty = self.tcx.type_of(def_id);
+                            let concrete_ty = generic_ty.subst(self.tcx, substs);
+                            let expanded_ty = self.fold_ty(concrete_ty);
+                            self.expanded_cache.insert((def_id, substs), expanded_ty);
+                            expanded_ty
+                        }
+                    };
                     self.seen_opaque_tys.remove(&def_id);
                     Some(expanded_ty)
                 } else {
@@ -735,14 +745,17 @@ impl<'tcx> TyCtxt<'tcx> {
             fn fold_ty(&mut self, t: Ty<'tcx>) -> Ty<'tcx> {
                 if let ty::Opaque(def_id, substs) = t.kind {
                     self.expand_opaque_ty(def_id, substs).unwrap_or(t)
-                } else {
+                } else if t.has_projections() {
                     t.super_fold_with(self)
+                } else {
+                    t
                 }
             }
         }
 
         let mut visitor = OpaqueTypeExpander {
             seen_opaque_tys: FxHashSet::default(),
+            expanded_cache: FxHashMap::default(),
             primary_def_id: def_id,
             found_recursion: false,
             tcx: self,
@@ -1096,6 +1109,9 @@ fn needs_drop_raw<'tcx>(tcx: TyCtxt<'tcx>, query: ty::ParamEnvAnd<'tcx, Ty<'tcx>
 
         ty::UnnormalizedProjection(..) => bug!("only used with chalk-engine"),
 
+        // Zero-length arrays never contain anything to drop.
+        ty::Array(_, len) if len.try_eval_usize(tcx, param_env) == Some(0) => false,
+
         // Structural recursion.
         ty::Array(ty, _) | ty::Slice(ty) => needs_drop(ty),
 

src/librustc_index/bit_set.rs

@@ -168,11 +168,7 @@ impl<T: Idx> BitSet<T> {
     /// Iterates over the indices of set bits in a sorted order.
     #[inline]
     pub fn iter(&self) -> BitIter<'_, T> {
-        BitIter {
-            cur: None,
-            iter: self.words.iter().enumerate(),
-            marker: PhantomData,
-        }
+        BitIter::new(&self.words)
     }
 
     /// Duplicates the set as a hybrid set.
@@ -291,26 +287,55 @@ impl<T: Idx> ToString for BitSet<T> {
 }
 
 pub struct BitIter<'a, T: Idx> {
-    cur: Option<(Word, usize)>,
-    iter: iter::Enumerate<slice::Iter<'a, Word>>,
+    /// A copy of the current word, but with any already-visited bits cleared.
+    /// (This lets us use `trailing_zeros()` to find the next set bit.) When it
+    /// is reduced to 0, we move onto the next word.
+    word: Word,
+
+    /// The offset (measured in bits) of the current word.
+    offset: usize,
+
+    /// Underlying iterator over the words.
+    iter: slice::Iter<'a, Word>,
+
     marker: PhantomData<T>
 }
 
+impl<'a, T: Idx> BitIter<'a, T> {
+    #[inline]
+    fn new(words: &'a [Word]) -> BitIter<'a, T> {
+        // We initialize `word` and `offset` to degenerate values. On the first
+        // call to `next()` we will fall through to getting the first word from
+        // `iter`, which sets `word` to the first word (if there is one) and
+        // `offset` to 0. Doing it this way saves us from having to maintain
+        // additional state about whether we have started.
+        BitIter {
+            word: 0,
+            offset: std::usize::MAX - (WORD_BITS - 1),
+            iter: words.iter(),
+            marker: PhantomData,
+        }
+    }
+}
+
 impl<'a, T: Idx> Iterator for BitIter<'a, T> {
     type Item = T;
     fn next(&mut self) -> Option<T> {
         loop {
-            if let Some((ref mut word, offset)) = self.cur {
-                let bit_pos = word.trailing_zeros() as usize;
-                if bit_pos != WORD_BITS {
-                    let bit = 1 << bit_pos;
-                    *word ^= bit;
-                    return Some(T::new(bit_pos + offset))
-                }
+            if self.word != 0 {
+                // Get the position of the next set bit in the current word,
+                // then clear the bit.
+                let bit_pos = self.word.trailing_zeros() as usize;
+                let bit = 1 << bit_pos;
+                self.word ^= bit;
+                return Some(T::new(bit_pos + self.offset))
             }
 
-            let (i, word) = self.iter.next()?;
-            self.cur = Some((*word, WORD_BITS * i));
+            // Move onto the next word. `wrapping_add()` is needed to handle
+            // the degenerate initial value given to `offset` in `new()`.
+            let word = self.iter.next()?;
+            self.word = *word;
+            self.offset = self.offset.wrapping_add(WORD_BITS);
         }
     }
 }
@@ -851,11 +876,7 @@ impl<R: Idx, C: Idx> BitMatrix<R, C> {
     pub fn iter(&self, row: R) -> BitIter<'_, C> {
         assert!(row.index() < self.num_rows);
         let (start, end) = self.range(row);
-        BitIter {
-            cur: None,
-            iter: self.words[start..end].iter().enumerate(),
-            marker: PhantomData,
-        }
+        BitIter::new(&self.words[start..end])
     }
 
     /// Returns the number of elements in `row`.

src/librustc_mir/dataflow/impls/indirect_mutation.rs

@@ -104,25 +104,16 @@ impl<'tcx> TransferFunction<'_, '_, 'tcx> {
         kind: mir::BorrowKind,
         borrowed_place: &mir::Place<'tcx>,
     ) -> bool {
-        let borrowed_ty = borrowed_place.ty(self.body, self.tcx).ty;
-
-        // Zero-sized types cannot be mutated, since there is nothing inside to mutate.
-        //
-        // FIXME: For now, we only exempt arrays of length zero. We need to carefully
-        // consider the effects before extending this to all ZSTs.
-        if let ty::Array(_, len) = borrowed_ty.kind {
-            if len.try_eval_usize(self.tcx, self.param_env) == Some(0) {
-                return false;
-            }
-        }
-
         match kind {
             mir::BorrowKind::Mut { .. } => true,
 
             | mir::BorrowKind::Shared
             | mir::BorrowKind::Shallow
             | mir::BorrowKind::Unique
-            => !borrowed_ty.is_freeze(self.tcx, self.param_env, DUMMY_SP),
+            => !borrowed_place
+                .ty(self.body, self.tcx)
+                .ty
+                .is_freeze(self.tcx, self.param_env, DUMMY_SP),
         }
     }
 }