Rollup merge of rust-lang#137180 - compiler-errors:sym-regions, r=oli-obk

Give `global_asm` a fake body to store typeck results, represent `sym fn` as a hir expr to fix `sym fn` operands with lifetimes

There are a few intertwined problems with `sym fn` operands in both inline and global asm macros.

Specifically, unlike other anon consts, they may evaluate to a type with free regions in them without actually having an item-level type annotation to give them a "proper" type. This is in contrast to named constants, which always have an item-level type annotation, or unnamed constants which are constrained by their position (e.g. a const arg in a turbofish, or a const array length).

Today, we infer the type of the operand by looking at the HIR typeck results; however, those results are region-erased, so during borrowck we ICE since we don't expect to encounter erased regions. We can't just fill this type with something like `'static`, since we may want to use real (free) regions:

```rust
fn foo<'a>() {
  asm!("/* ... */", sym bar::<&'a ()>);
}
```

The first idea may be to represent `sym fn` operands using *inline* consts instead of anon consts. This makes sense, since inline consts can reference regions from the parent body (like the `'a` in the example above). However, this introduces a problem with `global_asm!`, which doesn't *have* a parent body; inline consts *must* be associated with a parent body since they are not a body owner of their own. In rust-lang#116087, I attempted to fix this by using two separate `sym` operands for global and inline asm. However, this led to a lot of confusion and also some unattractive code duplication.

In this PR, I adjust the lowering of `global_asm!` so that it's lowered in a "fake" HIR body. This body contains a single expression which is `ExprKind::InlineAsm`; we don't *use* this HIR body, but it's used in typeck and borrowck so that we can properly infer and validate the the lifetimes of `sym fn` operands.

I then adjust the lowering of `sym fn` to instead be represented with a HIR expression. This is both because it's no longer necessary to represent this operand as an anon const, since it's *just* a path expression, and also more importantly to sidestep yet another ICE (rust-lang#137179), which has to do with the existing code breaking an invariant of def-id creation and anon consts. Specifically, we are not allowed to synthesize a def-id for an anon const when that anon const contains expressions with def-ids whose parent is *not* that anon const. This is somewhat related to rust-lang#130443 (comment), which is also a place in the compiler where synthesizing anon consts leads to def-id parenting issue.

As a side-effect, this consolidates the type checking for inline and global asm, so it allows us to simplify `InlineAsmCtxt` a bit. It also allows us to delete a bit of hacky code from anon const `type_of` which was there to detect `sym fn` operands specifically. This also could be generalized to support `const` asm operands with types with lifetimes in them. Since we specifically reject these consts today, I'm not going to change the representation of those consts (but they'd just be turned into inline consts).

r? oli-obk -- mostly b/c you're patient and also understand the breadth of the code that this touches, please reassign if you don't want to review this.

Fixes rust-lang#111709
Fixes rust-lang#96304
Fixes rust-lang#137179

Author: matthiaskrgr

compiler/rustc_ast_lowering/src/asm.rs

@@ -1,13 +1,12 @@
 use std::collections::hash_map::Entry;
 use std::fmt::Write;
 
-use rustc_ast::ptr::P;
 use rustc_ast::*;
 use rustc_data_structures::fx::{FxHashMap, FxHashSet, FxIndexMap};
 use rustc_hir as hir;
 use rustc_hir::def::{DefKind, Res};
 use rustc_session::parse::feature_err;
-use rustc_span::{Span, kw, sym};
+use rustc_span::{Span, sym};
 use rustc_target::asm;
 
 use super::LoweringContext;
@@ -230,20 +229,7 @@ impl<'a, 'hir> LoweringContext<'a, 'hir> {
                                 tokens: None,
                             };
 
-                            // Wrap the expression in an AnonConst.
-                            let parent_def_id = self.current_hir_id_owner.def_id;
-                            let node_id = self.next_node_id();
-                            self.create_def(
-                                parent_def_id,
-                                node_id,
-                                kw::Empty,
-                                DefKind::AnonConst,
-                                *op_sp,
-                            );
-                            let anon_const = AnonConst { id: node_id, value: P(expr) };
-                            hir::InlineAsmOperand::SymFn {
-                                anon_const: self.lower_anon_const_to_anon_const(&anon_const),
-                            }
+                            hir::InlineAsmOperand::SymFn { expr: self.lower_expr(&expr) }
                         }
                     }
                     InlineAsmOperand::Label { block } => {

compiler/rustc_ast_lowering/src/item.rs

@@ -251,7 +251,12 @@ impl<'hir> LoweringContext<'_, 'hir> {
                     .arena
                     .alloc_from_iter(fm.items.iter().map(|x| self.lower_foreign_item_ref(x))),
             },
-            ItemKind::GlobalAsm(asm) => hir::ItemKind::GlobalAsm(self.lower_inline_asm(span, asm)),
+            ItemKind::GlobalAsm(asm) => {
+                let asm = self.lower_inline_asm(span, asm);
+                let fake_body =
+                    self.lower_body(|this| (&[], this.expr(span, hir::ExprKind::InlineAsm(asm))));
+                hir::ItemKind::GlobalAsm { asm, fake_body }
+            }
             ItemKind::TyAlias(box TyAlias { generics, where_clauses, ty, .. }) => {
                 // We lower
                 //

compiler/rustc_borrowck/src/universal_regions.rs

@@ -126,6 +126,11 @@ pub(crate) enum DefiningTy<'tcx> {
     /// The MIR represents an inline const. The signature has no inputs and a
     /// single return value found via `InlineConstArgs::ty`.
     InlineConst(DefId, GenericArgsRef<'tcx>),
+
+    // Fake body for a global asm. Not particularly useful or interesting,
+    // but we need it so we can properly store the typeck results of the asm
+    // operands, which aren't associated with a body otherwise.
+    GlobalAsm(DefId),
 }
 
 impl<'tcx> DefiningTy<'tcx> {
@@ -138,9 +143,10 @@ impl<'tcx> DefiningTy<'tcx> {
             DefiningTy::Closure(_, args) => args.as_closure().upvar_tys(),
             DefiningTy::CoroutineClosure(_, args) => args.as_coroutine_closure().upvar_tys(),
             DefiningTy::Coroutine(_, args) => args.as_coroutine().upvar_tys(),
-            DefiningTy::FnDef(..) | DefiningTy::Const(..) | DefiningTy::InlineConst(..) => {
-                ty::List::empty()
-            }
+            DefiningTy::FnDef(..)
+            | DefiningTy::Const(..)
+            | DefiningTy::InlineConst(..)
+            | DefiningTy::GlobalAsm(_) => ty::List::empty(),
         }
     }
 
@@ -152,7 +158,10 @@ impl<'tcx> DefiningTy<'tcx> {
             DefiningTy::Closure(..)
             | DefiningTy::CoroutineClosure(..)
             | DefiningTy::Coroutine(..) => 1,
-            DefiningTy::FnDef(..) | DefiningTy::Const(..) | DefiningTy::InlineConst(..) => 0,
+            DefiningTy::FnDef(..)
+            | DefiningTy::Const(..)
+            | DefiningTy::InlineConst(..)
+            | DefiningTy::GlobalAsm(_) => 0,
         }
     }
 
@@ -171,7 +180,8 @@ impl<'tcx> DefiningTy<'tcx> {
             | DefiningTy::Coroutine(def_id, ..)
             | DefiningTy::FnDef(def_id, ..)
             | DefiningTy::Const(def_id, ..)
-            | DefiningTy::InlineConst(def_id, ..) => def_id,
+            | DefiningTy::InlineConst(def_id, ..)
+            | DefiningTy::GlobalAsm(def_id) => def_id,
         }
     }
 }
@@ -411,6 +421,7 @@ impl<'tcx> UniversalRegions<'tcx> {
                     tcx.def_path_str_with_args(def_id, args),
                 ));
             }
+            DefiningTy::GlobalAsm(_) => unreachable!(),
         }
     }
 
@@ -633,6 +644,8 @@ impl<'cx, 'tcx> UniversalRegionsBuilder<'cx, 'tcx> {
                     DefiningTy::InlineConst(self.mir_def.to_def_id(), args)
                 }
             }
+
+            BodyOwnerKind::GlobalAsm => DefiningTy::GlobalAsm(self.mir_def.to_def_id()),
         }
     }
 
@@ -666,6 +679,8 @@ impl<'cx, 'tcx> UniversalRegionsBuilder<'cx, 'tcx> {
             }
 
             DefiningTy::FnDef(_, args) | DefiningTy::Const(_, args) => args,
+
+            DefiningTy::GlobalAsm(_) => ty::List::empty(),
         };
 
         let global_mapping = iter::once((tcx.lifetimes.re_static, fr_static));
@@ -802,6 +817,10 @@ impl<'cx, 'tcx> UniversalRegionsBuilder<'cx, 'tcx> {
                 let ty = args.as_inline_const().ty();
                 ty::Binder::dummy(tcx.mk_type_list(&[ty]))
             }
+
+            DefiningTy::GlobalAsm(def_id) => {
+                ty::Binder::dummy(tcx.mk_type_list(&[tcx.type_of(def_id).instantiate_identity()]))
+            }
         };
 
         // FIXME(#129952): We probably want a more principled approach here.

compiler/rustc_codegen_cranelift/src/global_asm.rs

@@ -16,7 +16,7 @@ use crate::prelude::*;
 
 pub(crate) fn codegen_global_asm_item(tcx: TyCtxt<'_>, global_asm: &mut String, item_id: ItemId) {
     let item = tcx.hir_item(item_id);
-    if let rustc_hir::ItemKind::GlobalAsm(asm) = item.kind {
+    if let rustc_hir::ItemKind::GlobalAsm { asm, .. } = item.kind {
         let is_x86 =
             matches!(tcx.sess.asm_arch.unwrap(), InlineAsmArch::X86 | InlineAsmArch::X86_64);
 
@@ -55,15 +55,15 @@ pub(crate) fn codegen_global_asm_item(tcx: TyCtxt<'_>, global_asm: &mut String,
                                 }
                             }
                         }
-                        InlineAsmOperand::SymFn { anon_const } => {
+                        InlineAsmOperand::SymFn { expr } => {
                             if cfg!(not(feature = "inline_asm_sym")) {
                                 tcx.dcx().span_err(
                                     item.span,
                                     "asm! and global_asm! sym operands are not yet supported",
                                 );
                             }
 
-                            let ty = tcx.typeck_body(anon_const.body).node_type(anon_const.hir_id);
+                            let ty = tcx.typeck(item_id.owner_id).expr_ty(expr);
                             let instance = match ty.kind() {
                                 &ty::FnDef(def_id, args) => Instance::new(def_id, args),
                                 _ => span_bug!(op_sp, "asm sym is not a function"),

compiler/rustc_codegen_ssa/src/mono_item.rs

@@ -36,7 +36,7 @@ impl<'a, 'tcx: 'a> MonoItemExt<'a, 'tcx> for MonoItem<'tcx> {
             }
             MonoItem::GlobalAsm(item_id) => {
                 let item = cx.tcx().hir_item(item_id);
-                if let hir::ItemKind::GlobalAsm(asm) = item.kind {
+                if let hir::ItemKind::GlobalAsm { asm, .. } = item.kind {
                     let operands: Vec<_> = asm
                         .operands
                         .iter()
@@ -71,11 +71,8 @@ impl<'a, 'tcx: 'a> MonoItemExt<'a, 'tcx> for MonoItem<'tcx> {
                                     }
                                 }
                             }
-                            hir::InlineAsmOperand::SymFn { ref anon_const } => {
-                                let ty = cx
-                                    .tcx()
-                                    .typeck_body(anon_const.body)
-                                    .node_type(anon_const.hir_id);
+                            hir::InlineAsmOperand::SymFn { expr } => {
+                                let ty = cx.tcx().typeck(item_id.owner_id).expr_ty(expr);
                                 let instance = match ty.kind() {
                                     &ty::FnDef(def_id, args) => Instance::new(def_id, args),
                                     _ => span_bug!(*op_sp, "asm sym is not a function"),

compiler/rustc_hir/src/hir.rs

@@ -1913,13 +1913,18 @@ pub enum BodyOwnerKind {
 
     /// Initializer of a `static` item.
     Static(Mutability),
+
+    /// Fake body for a global asm to store its const-like value types.
+    GlobalAsm,
 }
 
 impl BodyOwnerKind {
     pub fn is_fn_or_closure(self) -> bool {
         match self {
             BodyOwnerKind::Fn | BodyOwnerKind::Closure => true,
-            BodyOwnerKind::Const { .. } | BodyOwnerKind::Static(_) => false,
+            BodyOwnerKind::Const { .. } | BodyOwnerKind::Static(_) | BodyOwnerKind::GlobalAsm => {
+                false
+            }
         }
     }
 }
@@ -3420,7 +3425,7 @@ pub enum InlineAsmOperand<'hir> {
         anon_const: &'hir AnonConst,
     },
     SymFn {
-        anon_const: &'hir AnonConst,
+        expr: &'hir Expr<'hir>,
     },
     SymStatic {
         path: QPath<'hir>,
@@ -3848,7 +3853,7 @@ impl<'hir> Item<'hir> {
         expect_foreign_mod, (ExternAbi, &'hir [ForeignItemRef]),
             ItemKind::ForeignMod { abi, items }, (*abi, items);
 
-        expect_global_asm, &'hir InlineAsm<'hir>, ItemKind::GlobalAsm(asm), asm;
+        expect_global_asm, &'hir InlineAsm<'hir>, ItemKind::GlobalAsm { asm, .. }, asm;
 
         expect_ty_alias, (&'hir Ty<'hir>, &'hir Generics<'hir>),
             ItemKind::TyAlias(ty, generics), (ty, generics);
@@ -4015,7 +4020,15 @@ pub enum ItemKind<'hir> {
     /// An external module, e.g. `extern { .. }`.
     ForeignMod { abi: ExternAbi, items: &'hir [ForeignItemRef] },
     /// Module-level inline assembly (from `global_asm!`).
-    GlobalAsm(&'hir InlineAsm<'hir>),
+    GlobalAsm {
+        asm: &'hir InlineAsm<'hir>,
+        /// A fake body which stores typeck results for the global asm's sym_fn
+        /// operands, which are represented as path expressions. This body contains
+        /// a single [`ExprKind::InlineAsm`] which points to the asm in the field
+        /// above, and which is typechecked like a inline asm expr just for the
+        /// typeck results.
+        fake_body: BodyId,
+    },
     /// A type alias, e.g., `type Foo = Bar<u8>`.
     TyAlias(&'hir Ty<'hir>, &'hir Generics<'hir>),
     /// An enum definition, e.g., `enum Foo<A, B> {C<A>, D<B>}`.
@@ -4081,7 +4094,7 @@ impl ItemKind<'_> {
             ItemKind::Macro(..) => "macro",
             ItemKind::Mod(..) => "module",
             ItemKind::ForeignMod { .. } => "extern block",
-            ItemKind::GlobalAsm(..) => "global asm item",
+            ItemKind::GlobalAsm { .. } => "global asm item",
             ItemKind::TyAlias(..) => "type alias",
             ItemKind::Enum(..) => "enum",
             ItemKind::Struct(..) => "struct",
@@ -4540,6 +4553,10 @@ impl<'hir> Node<'hir> {
                 ..
             }) => Some((owner_id.def_id, *body)),
 
+            Node::Item(Item {
+                owner_id, kind: ItemKind::GlobalAsm { asm: _, fake_body }, ..
+            }) => Some((owner_id.def_id, *fake_body)),
+
             Node::Expr(Expr { kind: ExprKind::Closure(Closure { def_id, body, .. }), .. }) => {
                 Some((*def_id, *body))
             }

compiler/rustc_hir/src/intravisit.rs

@@ -573,9 +573,13 @@ pub fn walk_item<'v, V: Visitor<'v>>(visitor: &mut V, item: &'v Item<'v>) -> V::
             try_visit!(visitor.visit_id(item.hir_id()));
             walk_list!(visitor, visit_foreign_item_ref, items);
         }
-        ItemKind::GlobalAsm(asm) => {
+        ItemKind::GlobalAsm { asm: _, fake_body } => {
             try_visit!(visitor.visit_id(item.hir_id()));
-            try_visit!(visitor.visit_inline_asm(asm, item.hir_id()));
+            // Visit the fake body, which contains the asm statement.
+            // Therefore we should not visit the asm statement again
+            // outside of the body, or some visitors won't have their
+            // typeck results set correctly.
+            try_visit!(visitor.visit_nested_body(fake_body));
         }
         ItemKind::TyAlias(ref ty, ref generics) => {
             try_visit!(visitor.visit_id(item.hir_id()));
@@ -1442,10 +1446,12 @@ pub fn walk_inline_asm<'v, V: Visitor<'v>>(
                 try_visit!(visitor.visit_expr(in_expr));
                 visit_opt!(visitor, visit_expr, out_expr);
             }
-            InlineAsmOperand::Const { anon_const, .. }
-            | InlineAsmOperand::SymFn { anon_const, .. } => {
+            InlineAsmOperand::Const { anon_const, .. } => {
                 try_visit!(visitor.visit_anon_const(anon_const));
             }
+            InlineAsmOperand::SymFn { expr, .. } => {
+                try_visit!(visitor.visit_expr(expr));
+            }
             InlineAsmOperand::SymStatic { path, .. } => {
                 try_visit!(visitor.visit_qpath(path, id, *op_sp));
             }

compiler/rustc_hir/src/target.rs

@@ -110,7 +110,7 @@ impl Target {
             ItemKind::Macro(..) => Target::MacroDef,
             ItemKind::Mod(..) => Target::Mod,
             ItemKind::ForeignMod { .. } => Target::ForeignMod,
-            ItemKind::GlobalAsm(..) => Target::GlobalAsm,
+            ItemKind::GlobalAsm { .. } => Target::GlobalAsm,
             ItemKind::TyAlias(..) => Target::TyAlias,
             ItemKind::Enum(..) => Target::Enum,
             ItemKind::Struct(..) => Target::Struct,

compiler/rustc_hir_analysis/src/check/check.rs

@@ -15,7 +15,6 @@ use rustc_lint_defs::builtin::{
 use rustc_middle::hir::nested_filter;
 use rustc_middle::middle::resolve_bound_vars::ResolvedArg;
 use rustc_middle::middle::stability::EvalResult;
-use rustc_middle::span_bug;
 use rustc_middle::ty::error::TypeErrorToStringExt;
 use rustc_middle::ty::fold::{BottomUpFolder, fold_regions};
 use rustc_middle::ty::layout::{LayoutError, MAX_SIMD_LANES};
@@ -35,7 +34,6 @@ use {rustc_attr_parsing as attr, rustc_hir as hir};
 
 use super::compare_impl_item::check_type_bounds;
 use super::*;
-use crate::check::intrinsicck::InlineAsmCtxt;
 
 pub fn check_abi(tcx: TyCtxt<'_>, span: Span, abi: ExternAbi) {
     if !tcx.sess.target.is_abi_supported(abi) {
@@ -895,13 +893,6 @@ pub(crate) fn check_item_type(tcx: TyCtxt<'_>, def_id: LocalDefId) {
                 }
             }
         }
-        DefKind::GlobalAsm => {
-            let it = tcx.hir().expect_item(def_id);
-            let hir::ItemKind::GlobalAsm(asm) = it.kind else {
-                span_bug!(it.span, "DefKind::GlobalAsm but got {:#?}", it)
-            };
-            InlineAsmCtxt::new_global_asm(tcx).check_asm(asm, def_id);
-        }
         _ => {}
     }
 }