[AST] Add RecoveryExpr to retain expressions on semantic errors

Normally clang avoids creating expressions when it encounters semantic
errors, even if the parser knows which expression to produce.

This works well for the compiler. However, this is not ideal for
source-level tools that have to deal with broken code, e.g. clangd is
not able to provide navigation features even for names that compiler
knows how to resolve.

The new RecoveryExpr aims to capture the minimal set of information
useful for the tools that need to deal with incorrect code:

source range of the expression being dropped,
subexpressions of the expression.
We aim to make constructing RecoveryExprs as simple as possible to
ensure writing code to avoid dropping expressions is easy.

Producing RecoveryExprs can result in new code paths being taken in the
frontend. In particular, clang can produce some new diagnostics now and
we aim to suppress bogus ones based on Expr::containsErrors.

We deliberately produce RecoveryExprs only in the parser for now to
minimize the code affected by this patch. Producing RecoveryExprs in
Sema potentially allows to preserve more information (e.g. type of an
expression), but also results in more code being affected. E.g.
SFINAE checks will have to take presence of RecoveryExprs into account.

Initial implementation only works in C++ mode, as it relies on compiler
postponing diagnostics on dependent expressions. C and ObjC often do not
do this, so they require more work to make sure we do not produce too
many bogus diagnostics on the new expressions.

See documentation of RecoveryExpr for more details.

original patch from Ilya
This change is based on https://reviews.llvm.org/D61722

Reviewers: sammccall, rsmith

Reviewed By: sammccall, rsmith

Tags: #clang

Differential Revision: https://reviews.llvm.org/D69330

Author: hokein

clang/include/clang/AST/ComputeDependence.h

@@ -45,6 +45,7 @@ class ExtVectorElementExpr;
 class BlockExpr;
 class AsTypeExpr;
 class DeclRefExpr;
+class RecoveryExpr;
 class CXXRewrittenBinaryOperator;
 class CXXStdInitializerListExpr;
 class CXXTypeidExpr;
@@ -122,6 +123,7 @@ ExprDependence computeDependence(ExtVectorElementExpr *E);
 ExprDependence computeDependence(BlockExpr *E);
 ExprDependence computeDependence(AsTypeExpr *E);
 ExprDependence computeDependence(DeclRefExpr *E, const ASTContext &Ctx);
+ExprDependence computeDependence(RecoveryExpr *E);
 ExprDependence computeDependence(CXXRewrittenBinaryOperator *E);
 ExprDependence computeDependence(CXXStdInitializerListExpr *E);
 ExprDependence computeDependence(CXXTypeidExpr *E);

clang/include/clang/AST/Expr.h

@@ -5911,6 +5911,69 @@ class TypoExpr : public Expr {
   }
 
 };
+
+/// Frontend produces RecoveryExprs on semantic errors that prevent creating
+/// other well-formed expressions. E.g. when type-checking of a binary operator
+/// fails, we cannot produce a BinaryOperator expression. Instead, we can choose
+/// to produce a recovery expression storing left and right operands.
+///
+/// RecoveryExpr does not have any semantic meaning in C++, it is only useful to
+/// preserve expressions in AST that would otherwise be dropped. It captures
+/// subexpressions of some expression that we could not construct and source
+/// range covered by the expression.
+///
+/// For now, RecoveryExpr is type-, value- and instantiation-dependent to take
+/// advantage of existing machinery to deal with dependent code in C++, e.g.
+/// RecoveryExpr is preserved in `decltype(<broken-expr>)` as part of the
+/// `DependentDecltypeType`. In addition to that, clang does not report most
+/// errors on dependent expressions, so we get rid of bogus errors for free.
+/// However, note that unlike other dependent expressions, RecoveryExpr can be
+/// produced in non-template contexts.
+///
+/// One can also reliably suppress all bogus errors on expressions containing
+/// recovery expressions by examining results of Expr::containsErrors().
+class RecoveryExpr final : public Expr,
+                           private llvm::TrailingObjects<RecoveryExpr, Expr *> {
+public:
+  static RecoveryExpr *Create(ASTContext &Ctx, SourceLocation BeginLoc,
+                              SourceLocation EndLoc, ArrayRef<Expr *> SubExprs);
+  static RecoveryExpr *CreateEmpty(ASTContext &Ctx, unsigned NumSubExprs);
+
+  ArrayRef<Expr *> subExpressions() {
+    auto *B = getTrailingObjects<Expr *>();
+    return llvm::makeArrayRef(B, B + NumExprs);
+  }
+
+  ArrayRef<const Expr *> subExpressions() const {
+    return const_cast<RecoveryExpr *>(this)->subExpressions();
+  }
+
+  child_range children() {
+    Stmt **B = reinterpret_cast<Stmt **>(getTrailingObjects<Expr *>());
+    return child_range(B, B + NumExprs);
+  }
+
+  SourceLocation getBeginLoc() const { return BeginLoc; }
+  SourceLocation getEndLoc() const { return EndLoc; }
+
+  static bool classof(const Stmt *T) {
+    return T->getStmtClass() == RecoveryExprClass;
+  }
+
+private:
+  RecoveryExpr(ASTContext &Ctx, SourceLocation BeginLoc, SourceLocation EndLoc,
+               ArrayRef<Expr *> SubExprs);
+  RecoveryExpr(EmptyShell Empty) : Expr(RecoveryExprClass, Empty) {}
+
+  size_t numTrailingObjects(OverloadToken<Stmt *>) const { return NumExprs; }
+
+  SourceLocation BeginLoc, EndLoc;
+  unsigned NumExprs;
+  friend TrailingObjects;
+  friend class ASTStmtReader;
+  friend class ASTStmtWriter;
+};
+
 } // end namespace clang
 
 #endif // LLVM_CLANG_AST_EXPR_H

clang/include/clang/AST/RecursiveASTVisitor.h

@@ -2668,6 +2668,7 @@ DEF_TRAVERSE_STMT(CXXRewrittenBinaryOperator, {
 })
 DEF_TRAVERSE_STMT(OpaqueValueExpr, {})
 DEF_TRAVERSE_STMT(TypoExpr, {})
+DEF_TRAVERSE_STMT(RecoveryExpr, {})
 DEF_TRAVERSE_STMT(CUDAKernelCallExpr, {})
 
 // These operators (all of them) do not need any action except

clang/include/clang/Basic/LangOptions.def

@@ -148,6 +148,8 @@ LANGOPT(RelaxedTemplateTemplateArgs, 1, 0, "C++17 relaxed matching of template t
 
 LANGOPT(DoubleSquareBracketAttributes, 1, 0, "'[[]]' attributes extension for all language standard modes")
 
+COMPATIBLE_LANGOPT(RecoveryAST, 1, 0, "Preserve expressions in AST when encountering errors")
+
 BENIGN_LANGOPT(ThreadsafeStatics , 1, 1, "thread-safe static initializers")
 LANGOPT(POSIXThreads      , 1, 0, "POSIX thread support")
 LANGOPT(Blocks            , 1, 0, "blocks extension to C")

clang/include/clang/Basic/StmtNodes.td

@@ -195,6 +195,7 @@ def ConvertVectorExpr : StmtNode<Expr>;
 def BlockExpr : StmtNode<Expr>;
 def OpaqueValueExpr : StmtNode<Expr>;
 def TypoExpr : StmtNode<Expr>;
+def RecoveryExpr : StmtNode<Expr>;
 def BuiltinBitCastExpr : StmtNode<ExplicitCastExpr>;
 
 // Microsoft Extensions.

clang/include/clang/Driver/CC1Options.td

@@ -566,6 +566,11 @@ def fno_concept_satisfaction_caching : Flag<["-"],
                                             "fno-concept-satisfaction-caching">,
   HelpText<"Disable satisfaction caching for C++2a Concepts.">;
 
+def frecovery_ast : Flag<["-"], "frecovery-ast">,
+  HelpText<"Preserve expressions in AST rather than dropping them when "
+           "encountering semantic errors">;
+def fno_recovery_ast : Flag<["-"], "fno-recovery-ast">;
+
 let Group = Action_Group in {
 
 def Eonly : Flag<["-"], "Eonly">,

clang/include/clang/Sema/Sema.h

@@ -3885,6 +3885,10 @@ class Sema final {
   void DiagnoseAmbiguousLookup(LookupResult &Result);
   //@}
 
+  /// Attempts to produce a RecoveryExpr after some AST node cannot be created.
+  ExprResult CreateRecoveryExpr(SourceLocation Begin, SourceLocation End,
+                                ArrayRef<Expr *> SubExprs);
+
   ObjCInterfaceDecl *getObjCInterfaceDecl(IdentifierInfo *&Id,
                                           SourceLocation IdLoc,
                                           bool TypoCorrection = false);

clang/include/clang/Serialization/ASTBitCodes.h

@@ -1634,6 +1634,9 @@ namespace serialization {
       /// An AtomicExpr record.
       EXPR_ATOMIC,
 
+      /// A RecoveryExpr record.
+      EXPR_RECOVERY,
+
       // Objective-C
 
       /// An ObjCStringLiteral record.

clang/lib/AST/ComputeDependence.cpp

@@ -456,6 +456,15 @@ ExprDependence clang::computeDependence(DeclRefExpr *E, const ASTContext &Ctx) {
   return Deps;
 }
 
+ExprDependence clang::computeDependence(RecoveryExpr *E) {
+  // FIXME: drop type+value+instantiation once Error is sufficient to suppress
+  // bogus dianostics.
+  auto D = ExprDependence::TypeValueInstantiation | ExprDependence::Error;
+  for (auto *S : E->subExpressions())
+    D |= S->getDependence();
+  return D;
+}
+
 ExprDependence clang::computeDependence(PredefinedExpr *E) {
   return toExprDependence(E->getType()->getDependence()) &
          ~ExprDependence::UnexpandedPack;

clang/lib/AST/Expr.cpp

@@ -2375,6 +2375,7 @@ bool Expr::isUnusedResultAWarning(const Expr *&WarnE, SourceLocation &Loc,
   // If we don't know precisely what we're looking at, let's not warn.
   case UnresolvedLookupExprClass:
   case CXXUnresolvedConstructExprClass:
+  case RecoveryExprClass:
     return false;
 
   case CXXTemporaryObjectExprClass:
@@ -3227,6 +3228,7 @@ bool Expr::HasSideEffects(const ASTContext &Ctx,
   case SubstNonTypeTemplateParmPackExprClass:
   case FunctionParmPackExprClass:
   case TypoExprClass:
+  case RecoveryExprClass:
   case CXXFoldExprClass:
     llvm_unreachable("shouldn't see dependent / unresolved nodes here");
 
@@ -4467,3 +4469,30 @@ QualType OMPArraySectionExpr::getBaseOriginalType(const Expr *Base) {
   }
   return OriginalTy;
 }
+
+RecoveryExpr::RecoveryExpr(ASTContext &Ctx, SourceLocation BeginLoc,
+                           SourceLocation EndLoc, ArrayRef<Expr *> SubExprs)
+    : Expr(RecoveryExprClass, Ctx.DependentTy, VK_LValue, OK_Ordinary),
+      BeginLoc(BeginLoc), EndLoc(EndLoc), NumExprs(SubExprs.size()) {
+#ifndef NDEBUG
+  for (auto *E : SubExprs)
+    assert(E != nullptr);
+#endif
+
+  llvm::copy(SubExprs, getTrailingObjects<Expr *>());
+  setDependence(computeDependence(this));
+}
+
+RecoveryExpr *RecoveryExpr::Create(ASTContext &Ctx, SourceLocation BeginLoc,
+                                   SourceLocation EndLoc,
+                                   ArrayRef<Expr *> SubExprs) {
+  void *Mem = Ctx.Allocate(totalSizeToAlloc<Expr *>(SubExprs.size()),
+                           alignof(RecoveryExpr));
+  return new (Mem) RecoveryExpr(Ctx, BeginLoc, EndLoc, SubExprs);
+}
+
+RecoveryExpr *RecoveryExpr::CreateEmpty(ASTContext &Ctx, unsigned NumSubExprs) {
+  void *Mem = Ctx.Allocate(totalSizeToAlloc<Expr *>(NumSubExprs),
+                           alignof(RecoveryExpr));
+  return new (Mem) RecoveryExpr(EmptyShell());
+}