[C++20] [Modules] [Itanium ABI] Generate the vtable in the module unit of dynamic classes (#75912)

Close #70585 and reflect
itanium-cxx-abi/cxx-abi#170.

The significant change of the patch is: for dynamic classes attached to
module units, we generate the vtable to the attached module units
directly and the key functions for such classes is meaningless.

Author: ChuanqiXu9

clang/include/clang/AST/DeclBase.h

@@ -670,6 +670,9 @@ class alignas(8) Decl {
   /// Whether this declaration comes from another module unit.
   bool isInAnotherModuleUnit() const;
 
+  /// Whether this declaration comes from the same module unit being compiled.
+  bool isInCurrentModuleUnit() const;
+
   /// Whether the definition of the declaration should be emitted in external
   /// sources.
   bool shouldEmitInExternalSource() const;

clang/include/clang/Serialization/ASTBitCodes.h

@@ -697,6 +697,9 @@ enum ASTRecordTypes {
 
   /// Record code for \#pragma clang unsafe_buffer_usage begin/end
   PP_UNSAFE_BUFFER_USAGE = 69,
+
+  /// Record code for vtables to emit.
+  VTABLES_TO_EMIT = 70,
 };
 
 /// Record types used within a source manager block.

clang/include/clang/Serialization/ASTReader.h

@@ -805,6 +805,11 @@ class ASTReader
   /// the consumer eagerly.
   SmallVector<GlobalDeclID, 16> EagerlyDeserializedDecls;
 
+  /// The IDs of all vtables to emit. The referenced declarations are passed
+  /// to the consumers's HandleVTable eagerly after passing
+  /// EagerlyDeserializedDecls.
+  SmallVector<GlobalDeclID, 16> VTablesToEmit;
+
   /// The IDs of all tentative definitions stored in the chain.
   ///
   /// Sema keeps track of all tentative definitions in a TU because it has to
@@ -1514,6 +1519,7 @@ class ASTReader
   bool isConsumerInterestedIn(Decl *D);
   void PassInterestingDeclsToConsumer();
   void PassInterestingDeclToConsumer(Decl *D);
+  void PassVTableToConsumer(CXXRecordDecl *RD);
 
   void finishPendingActions();
   void diagnoseOdrViolations();

clang/include/clang/Serialization/ASTWriter.h

@@ -495,6 +495,10 @@ class ASTWriter : public ASTDeserializationListener,
   std::vector<SourceRange> NonAffectingRanges;
   std::vector<SourceLocation::UIntTy> NonAffectingOffsetAdjustments;
 
+  /// A list of classes which need to emit the VTable in the corresponding
+  /// object file.
+  llvm::SmallVector<CXXRecordDecl *> PendingEmittingVTables;
+
   /// Computes input files that didn't affect compilation of the current module,
   /// and initializes data structures necessary for leaving those files out
   /// during \c SourceManager serialization.
@@ -849,6 +853,8 @@ class ASTWriter : public ASTDeserializationListener,
 
   bool getDoneWritingDeclsAndTypes() const { return DoneWritingDeclsAndTypes; }
 
+  void handleVTable(CXXRecordDecl *RD);
+
 private:
   // ASTDeserializationListener implementation
   void ReaderInitialized(ASTReader *Reader) override;
@@ -943,6 +949,7 @@ class PCHGenerator : public SemaConsumer {
 
   void InitializeSema(Sema &S) override { SemaPtr = &S; }
   void HandleTranslationUnit(ASTContext &Ctx) override;
+  void HandleVTable(CXXRecordDecl *RD) override { Writer.handleVTable(RD); }
   ASTMutationListener *GetASTMutationListener() override;
   ASTDeserializationListener *GetASTDeserializationListener() override;
   bool hasEmittedPCH() const { return Buffer->IsComplete; }

clang/lib/AST/DeclBase.cpp

@@ -1128,6 +1128,15 @@ bool Decl::isInAnotherModuleUnit() const {
   return M != getASTContext().getCurrentNamedModule();
 }
 
+bool Decl::isInCurrentModuleUnit() const {
+  auto *M = getOwningModule();
+
+  if (!M || !M->isNamedModule())
+    return false;
+
+  return M == getASTContext().getCurrentNamedModule();
+}
+
 bool Decl::shouldEmitInExternalSource() const {
   ExternalASTSource *Source = getASTContext().getExternalSource();
   if (!Source)

clang/lib/CodeGen/CGVTables.cpp

@@ -1051,6 +1051,11 @@ CodeGenModule::getVTableLinkage(const CXXRecordDecl *RD) {
   if (!RD->isExternallyVisible())
     return llvm::GlobalVariable::InternalLinkage;
 
+  // V-tables for non-template classes with an owning module are always
+  // uniquely emitted in that module.
+  if (RD->isInNamedModule())
+    return llvm::GlobalVariable::ExternalLinkage;
+
   // We're at the end of the translation unit, so the current key
   // function is fully correct.
   const CXXMethodDecl *keyFunction = Context.getCurrentKeyFunction(RD);
@@ -1185,6 +1190,21 @@ bool CodeGenVTables::isVTableExternal(const CXXRecordDecl *RD) {
       TSK == TSK_ExplicitInstantiationDefinition)
     return false;
 
+  // Itanium C++ ABI [5.2.3]:
+  // Virtual tables for dynamic classes are emitted as follows:
+  //
+  // - If the class is templated, the tables are emitted in every object that
+  // references any of them.
+  // - Otherwise, if the class is attached to a module, the tables are uniquely
+  // emitted in the object for the module unit in which it is defined.
+  // - Otherwise, if the class has a key function (see below), the tables are
+  // emitted in the object for the translation unit containing the definition of
+  // the key function. This is unique if the key function is not inline.
+  // - Otherwise, the tables are emitted in every object that references any of
+  // them.
+  if (RD->isInNamedModule())
+    return RD->shouldEmitInExternalSource();
+
   // Otherwise, if the class doesn't have a key function (possibly
   // anymore), the vtable must be defined here.
   const CXXMethodDecl *keyFunction = CGM.getContext().getCurrentKeyFunction(RD);
@@ -1194,13 +1214,7 @@ bool CodeGenVTables::isVTableExternal(const CXXRecordDecl *RD) {
   const FunctionDecl *Def;
   // Otherwise, if we don't have a definition of the key function, the
   // vtable must be defined somewhere else.
-  if (!keyFunction->hasBody(Def))
-    return true;
-
-  assert(Def && "The body of the key function is not assigned to Def?");
-  // If the non-inline key function comes from another module unit, the vtable
-  // must be defined there.
-  return Def->shouldEmitInExternalSource() && !Def->isInlineSpecified();
+  return !keyFunction->hasBody(Def);
 }
 
 /// Given that we're currently at the end of the translation unit, and

clang/lib/CodeGen/ItaniumCXXABI.cpp

@@ -2130,6 +2130,9 @@ bool ItaniumCXXABI::canSpeculativelyEmitVTable(const CXXRecordDecl *RD) const {
   if (!canSpeculativelyEmitVTableAsBaseClass(RD))
     return false;
 
+  if (RD->shouldEmitInExternalSource())
+    return false;
+
   // For a complete-object vtable (or more specifically, for the VTT), we need
   // to be able to speculatively emit the vtables of all dynamic virtual bases.
   for (const auto &B : RD->vbases()) {

clang/lib/Sema/SemaDecl.cpp

@@ -18333,6 +18333,15 @@ void Sema::ActOnTagFinishDefinition(Scope *S, Decl *TagD,
       if (NumInitMethods > 1 || !Def->hasInitMethod())
         Diag(RD->getLocation(), diag::err_sycl_special_type_num_init_method);
     }
+
+    // If we're defining a dynamic class in a module interface unit, we always
+    // need to produce the vtable for it even if the vtable is not used in the
+    // current TU.
+    //
+    // The case that the current class is not dynamic is handled in
+    // MarkVTableUsed.
+    if (getCurrentModule() && getCurrentModule()->isInterfaceOrPartition())
+      MarkVTableUsed(RD->getLocation(), RD, /*DefinitionRequired=*/true);
   }
 
   // Exit this scope of this tag's definition.

clang/lib/Sema/SemaDeclCXX.cpp

@@ -18706,11 +18706,15 @@ bool Sema::DefineUsedVTables() {
 
     bool DefineVTable = true;
 
-    // If this class has a key function, but that key function is
-    // defined in another translation unit, we don't need to emit the
-    // vtable even though we're using it.
     const CXXMethodDecl *KeyFunction = Context.getCurrentKeyFunction(Class);
-    if (KeyFunction && !KeyFunction->hasBody()) {
+    // V-tables for non-template classes with an owning module are always
+    // uniquely emitted in that module.
+    if (Class->isInCurrentModuleUnit())
+      DefineVTable = true;
+    else if (KeyFunction && !KeyFunction->hasBody()) {
+      // If this class has a key function, but that key function is
+      // defined in another translation unit, we don't need to emit the
+      // vtable even though we're using it.
       // The key function is in another translation unit.
       DefineVTable = false;
       TemplateSpecializationKind TSK =
@@ -18755,7 +18759,7 @@ bool Sema::DefineUsedVTables() {
     DefinedAnything = true;
     MarkVirtualMembersReferenced(Loc, Class);
     CXXRecordDecl *Canonical = Class->getCanonicalDecl();
-    if (VTablesUsed[Canonical])
+    if (VTablesUsed[Canonical] && !Class->shouldEmitInExternalSource())
       Consumer.HandleVTable(Class);
 
     // Warn if we're emitting a weak vtable. The vtable will be weak if there is

clang/lib/Serialization/ASTReader.cpp

@@ -3903,6 +3903,13 @@ llvm::Error ASTReader::ReadASTBlock(ModuleFile &F,
       }
       break;
 
+    case VTABLES_TO_EMIT:
+      if (F.Kind == MK_MainFile ||
+          getContext().getLangOpts().BuildingPCHWithObjectFile)
+        for (unsigned I = 0, N = Record.size(); I != N;)
+          VTablesToEmit.push_back(getGlobalDeclID(F, LocalDeclID(Record[I++])));
+      break;
+
     case IMPORTED_MODULES:
       if (!F.isModule()) {
         // If we aren't loading a module (which has its own exports), make
@@ -8067,6 +8074,10 @@ void ASTReader::PassInterestingDeclToConsumer(Decl *D) {
     Consumer->HandleInterestingDecl(DeclGroupRef(D));
 }
 
+void ASTReader::PassVTableToConsumer(CXXRecordDecl *RD) {
+  Consumer->HandleVTable(RD);
+}
+
 void ASTReader::StartTranslationUnit(ASTConsumer *Consumer) {
   this->Consumer = Consumer;