[C++20] [Modules] Support to export declarations in language linkage

Close #60405

See the discussion in the above link for the background.

What the patch does:
- Rename `Module::ModuleKind::GlobalModuleFragment` to
  `Module::ModuleKind::ExplicitGlobalModuleFragment`.
- Add another module kind `ImplicitGlobalModuleFragment` to
  `ModuleKind`.
- Create an implicit global module fragment for the language linkage
  declarations inside a module purview.
    - If the language linkage lives inside the scope of an export decl,
      the created modules is marked as exported to outer modules.
- In fact, Sema will only create at most 2 implicit global module
  fragments to avoid creating a lot of unnecessary modules in the edging
case.

Reviewed By: iains

Differential Revision: https://reviews.llvm.org/D144367

clang/docs/ReleaseNotes.rst

@@ -176,6 +176,8 @@ Bug Fixes in This Version
   (`#60268 <https://github.com/llvm/llvm-project/issues/60268>`_)
 - Fix crash when taking the address of a consteval lambda call operator.
   (`#57682 <https://github.com/llvm/llvm-project/issues/57682>`_)
+- Clang now support export declarations in the language linkage.
+  (`#60405 <https://github.com/llvm/llvm-project/issues/60405>`_)
 
 Bug Fixes to Compiler Builtins
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

clang/include/clang/Basic/Module.h

@@ -120,11 +120,17 @@ class alignas(8) Module {
     /// This is a C++ 20 module partition implementation.
     ModulePartitionImplementation,
 
-    /// This is a fragment of the global module within some C++ module.
-    GlobalModuleFragment,
+    /// This is the explicit Global Module Fragment of a modular TU.
+    /// As per C++ [module.global.frag].
+    ExplicitGlobalModuleFragment,
 
     /// This is the private module fragment within some C++ module.
     PrivateModuleFragment,
+
+    /// This is an implicit fragment of the global module which contains
+    /// only language linkage declarations (made in the purview of the
+    /// named module).
+    ImplicitGlobalModuleFragment,
   };
 
   /// The kind of this module.
@@ -170,7 +176,15 @@ class alignas(8) Module {
 
   /// Does this Module scope describe a fragment of the global module within
   /// some C++ module.
-  bool isGlobalModule() const { return Kind == GlobalModuleFragment; }
+  bool isGlobalModule() const {
+    return isExplicitGlobalModule() || isImplicitGlobalModule();
+  }
+  bool isExplicitGlobalModule() const {
+    return Kind == ExplicitGlobalModuleFragment;
+  }
+  bool isImplicitGlobalModule() const {
+    return Kind == ImplicitGlobalModuleFragment;
+  }
 
   bool isPrivateModule() const { return Kind == PrivateModuleFragment; }
 

clang/include/clang/Lex/ModuleMap.h

@@ -553,6 +553,8 @@ class ModuleMap {
   /// parent.
   Module *createGlobalModuleFragmentForModuleUnit(SourceLocation Loc,
                                                   Module *Parent = nullptr);
+  Module *createImplicitGlobalModuleFragmentForModuleUnit(
+      SourceLocation Loc, bool IsExported, Module *Parent = nullptr);
 
   /// Create a global module fragment for a C++ module interface unit.
   Module *createPrivateModuleFragmentForInterfaceUnit(Module *Parent,

clang/include/clang/Sema/Sema.h

@@ -2276,8 +2276,20 @@ class Sema final {
   };
   /// The modules we're currently parsing.
   llvm::SmallVector<ModuleScope, 16> ModuleScopes;
-  /// The global module fragment of the current translation unit.
-  clang::Module *GlobalModuleFragment = nullptr;
+  /// The explicit global module fragment of the current translation unit.
+  /// The explicit Global Module Fragment, as specified in C++
+  /// [module.global.frag].
+  clang::Module *TheGlobalModuleFragment = nullptr;
+
+  /// The implicit global module fragments of the current translation unit.
+  /// We would only create at most two implicit global module fragments to
+  /// avoid performance penalties when there are many language linkage
+  /// exports.
+  ///
+  /// The contents in the implicit global module fragment can't be discarded
+  /// no matter if it is exported or not.
+  clang::Module *TheImplicitGlobalModuleFragment = nullptr;
+  clang::Module *TheExportedImplicitGlobalModuleFragment = nullptr;
 
   /// Namespace definitions that we will export when they finish.
   llvm::SmallPtrSet<const NamespaceDecl*, 8> DeferredExportedNamespaces;
@@ -2293,11 +2305,17 @@ class Sema final {
     return getCurrentModule() ? getCurrentModule()->isModulePurview() : false;
   }
 
-  /// Enter the scope of the global module.
+  /// Enter the scope of the explicit global module fragment.
   Module *PushGlobalModuleFragment(SourceLocation BeginLoc);
-  /// Leave the scope of the global module.
+  /// Leave the scope of the explicit global module fragment.
   void PopGlobalModuleFragment();
 
+  /// Enter the scope of an implicit global module fragment.
+  Module *PushImplicitGlobalModuleFragment(SourceLocation BeginLoc,
+                                           bool IsExported);
+  /// Leave the scope of an implicit global module fragment.
+  void PopImplicitGlobalModuleFragment();
+
   VisibleModuleSet VisibleModules;
 
   /// Cache for module units which is usable for current module.

clang/lib/AST/Decl.cpp

@@ -1590,7 +1590,8 @@ Module *Decl::getOwningModuleForLinkage(bool IgnoreLinkage) const {
     return M;
 
   case Module::ModuleHeaderUnit:
-  case Module::GlobalModuleFragment: {
+  case Module::ExplicitGlobalModuleFragment:
+  case Module::ImplicitGlobalModuleFragment: {
     // External linkage declarations in the global module have no owning module
     // for linkage purposes. But internal linkage declarations in the global
     // module fragment of a particular module are owned by that module for

clang/lib/Frontend/FrontendActions.cpp

@@ -765,8 +765,10 @@ static StringRef ModuleKindName(Module::ModuleKind MK) {
     return "Partition Implementation";
   case Module::ModuleHeaderUnit:
     return "Header Unit";
-  case Module::GlobalModuleFragment:
+  case Module::ExplicitGlobalModuleFragment:
     return "Global Module Fragment";
+  case Module::ImplicitGlobalModuleFragment:
+    return "Implicit Module Fragment";
   case Module::PrivateModuleFragment:
     return "Private Module Fragment";
   }

clang/lib/Lex/ModuleMap.cpp

@@ -855,14 +855,29 @@ Module *ModuleMap::createGlobalModuleFragmentForModuleUnit(SourceLocation Loc,
                                                            Module *Parent) {
   auto *Result = new Module("<global>", Loc, Parent, /*IsFramework*/ false,
                             /*IsExplicit*/ true, NumCreatedModules++);
-  Result->Kind = Module::GlobalModuleFragment;
+  Result->Kind = Module::ExplicitGlobalModuleFragment;
   // If the created module isn't owned by a parent, send it to PendingSubmodules
   // to wait for its parent.
   if (!Result->Parent)
     PendingSubmodules.emplace_back(Result);
   return Result;
 }
 
+Module *ModuleMap::createImplicitGlobalModuleFragmentForModuleUnit(
+    SourceLocation Loc, bool IsExported, Module *Parent) {
+  assert(Parent && "We should only create an implicit global module fragment "
+                   "in a module purview");
+  // Note: Here the `IsExplicit` parameter refers to the semantics in clang
+  // modules. All the non-explicit submodules in clang modules will be exported
+  // too. Here we simplify the implementation by using the concept.
+  auto *Result = new Module(IsExported ? "<exported implicit global>"
+                                       : "<implicit global>",
+                            Loc, Parent, /*IsFramework*/ false,
+                            /*IsExplicit*/ !IsExported, NumCreatedModules++);
+  Result->Kind = Module::ImplicitGlobalModuleFragment;
+  return Result;
+}
+
 Module *
 ModuleMap::createPrivateModuleFragmentForInterfaceUnit(Module *Parent,
                                                        SourceLocation Loc) {

clang/lib/Sema/Sema.cpp

@@ -1202,8 +1202,8 @@ void Sema::ActOnEndOfTranslationUnit() {
 
   // A global-module-fragment is only permitted within a module unit.
   bool DiagnosedMissingModuleDeclaration = false;
-  if (!ModuleScopes.empty() &&
-      ModuleScopes.back().Module->Kind == Module::GlobalModuleFragment) {
+  if (!ModuleScopes.empty() && ModuleScopes.back().Module->Kind ==
+                                   Module::ExplicitGlobalModuleFragment) {
     Diag(ModuleScopes.back().BeginLoc,
          diag::err_module_declaration_missing_after_global_module_introducer);
     DiagnosedMissingModuleDeclaration = true;

clang/lib/Sema/SemaDeclCXX.cpp

@@ -16362,14 +16362,8 @@ Decl *Sema::ActOnStartLinkageSpecification(Scope *S, SourceLocation ExternLoc,
   /// If the declaration is already in global module fragment, we don't
   /// need to attach it again.
   if (getLangOpts().CPlusPlusModules && isCurrentModulePurview()) {
-    Module *GlobalModule =
-        PushGlobalModuleFragment(ExternLoc);
-    /// According to [module.reach]p3.2,
-    /// The declaration in global module fragment is reachable if it is not
-    /// discarded. And the discarded declaration should be deleted. So it
-    /// doesn't matter mark the declaration in global module fragment as
-    /// reachable here.
-    D->setModuleOwnershipKind(Decl::ModuleOwnershipKind::ReachableWhenImported);
+    Module *GlobalModule = PushImplicitGlobalModuleFragment(
+        ExternLoc, /*IsExported=*/D->isInExportDeclContext());
     D->setLocalOwningModule(GlobalModule);
   }
 
@@ -16395,7 +16389,7 @@ Decl *Sema::ActOnFinishLinkageSpecification(Scope *S,
   // need to pop it.
   if (getLangOpts().CPlusPlusModules && getCurrentModule() &&
       getCurrentModule()->isGlobalModule() && getCurrentModule()->Parent)
-    PopGlobalModuleFragment();
+    PopImplicitGlobalModuleFragment();
 
   PopDeclContext();
   return LinkageSpec;

clang/lib/Sema/SemaExprCXX.cpp

@@ -3023,10 +3023,10 @@ void Sema::DeclareGlobalNewDelete() {
 
     // The implicitly declared "std::bad_alloc" should live in global module
     // fragment.
-    if (GlobalModuleFragment) {
+    if (TheGlobalModuleFragment) {
       getStdBadAlloc()->setModuleOwnershipKind(
           Decl::ModuleOwnershipKind::ReachableWhenImported);
-      getStdBadAlloc()->setLocalOwningModule(GlobalModuleFragment);
+      getStdBadAlloc()->setLocalOwningModule(TheGlobalModuleFragment);
     }
   }
   if (!StdAlignValT && getLangOpts().AlignedAllocation) {
@@ -3038,10 +3038,10 @@ void Sema::DeclareGlobalNewDelete() {
 
     // The implicitly declared "std::align_val_t" should live in global module
     // fragment.
-    if (GlobalModuleFragment) {
+    if (TheGlobalModuleFragment) {
       AlignValT->setModuleOwnershipKind(
           Decl::ModuleOwnershipKind::ReachableWhenImported);
-      AlignValT->setLocalOwningModule(GlobalModuleFragment);
+      AlignValT->setLocalOwningModule(TheGlobalModuleFragment);
     }
 
     AlignValT->setIntegerType(Context.getSizeType());
@@ -3170,10 +3170,10 @@ void Sema::DeclareGlobalAllocationFunction(DeclarationName Name,
     // module all the time. But in the implementation, the global module
     // is only meaningful when we're in a module unit. So here we attach
     // these allocation functions to global module conditionally.
-    if (GlobalModuleFragment) {
+    if (TheGlobalModuleFragment) {
       Alloc->setModuleOwnershipKind(
           Decl::ModuleOwnershipKind::ReachableWhenImported);
-      Alloc->setLocalOwningModule(GlobalModuleFragment);
+      Alloc->setLocalOwningModule(TheGlobalModuleFragment);
     }
 
     Alloc->addAttr(VisibilityAttr::CreateImplicit(

clang/lib/Sema/SemaLookup.cpp

@@ -1579,7 +1579,7 @@ bool Sema::isUsableModule(const Module *M) {
   // [module.global.frag]p1:
   //   The global module fragment can be used to provide declarations that are
   //   attached to the global module and usable within the module unit.
-  if (M == GlobalModuleFragment ||
+  if (M == TheGlobalModuleFragment ||
       // If M is the module we're parsing, it should be usable. This covers the
       // private module fragment. The private module fragment is usable only if
       // it is within the current module unit. And it must be the current

clang/lib/Sema/SemaModule.cpp

@@ -233,7 +233,7 @@ Sema::ActOnModuleDecl(SourceLocation StartLoc, SourceLocation ModuleLoc,
   }
 
   assert((!getLangOpts().CPlusPlusModules ||
-          SeenGMF == (bool)this->GlobalModuleFragment) &&
+          SeenGMF == (bool)this->TheGlobalModuleFragment) &&
          "mismatched global module state");
 
   // In C++20, the module-declaration must be the first declaration if there
@@ -358,7 +358,7 @@ Sema::ActOnModuleDecl(SourceLocation StartLoc, SourceLocation ModuleLoc,
     break;
   }
 
-  if (!this->GlobalModuleFragment) {
+  if (!this->TheGlobalModuleFragment) {
     ModuleScopes.push_back({});
     if (getLangOpts().ModulesLocalVisibility)
       ModuleScopes.back().OuterVisibleModules = std::move(VisibleModules);
@@ -408,10 +408,11 @@ Sema::ActOnPrivateModuleFragmentDecl(SourceLocation ModuleLoc,
   // C++20 [basic.link]/2:
   //   A private-module-fragment shall appear only in a primary module
   //   interface unit.
-  switch (ModuleScopes.empty() ? Module::GlobalModuleFragment
+  switch (ModuleScopes.empty() ? Module::ExplicitGlobalModuleFragment
                                : ModuleScopes.back().Module->Kind) {
   case Module::ModuleMapModule:
-  case Module::GlobalModuleFragment:
+  case Module::ExplicitGlobalModuleFragment:
+  case Module::ImplicitGlobalModuleFragment:
   case Module::ModulePartitionImplementation:
   case Module::ModulePartitionInterface:
   case Module::ModuleHeaderUnit:
@@ -958,25 +959,52 @@ Decl *Sema::ActOnFinishExportDecl(Scope *S, Decl *D, SourceLocation RBraceLoc) {
 Module *Sema::PushGlobalModuleFragment(SourceLocation BeginLoc) {
   // We shouldn't create new global module fragment if there is already
   // one.
-  if (!GlobalModuleFragment) {
+  if (!TheGlobalModuleFragment) {
     ModuleMap &Map = PP.getHeaderSearchInfo().getModuleMap();
-    GlobalModuleFragment = Map.createGlobalModuleFragmentForModuleUnit(
+    TheGlobalModuleFragment = Map.createGlobalModuleFragmentForModuleUnit(
         BeginLoc, getCurrentModule());
   }
 
-  assert(GlobalModuleFragment && "module creation should not fail");
+  assert(TheGlobalModuleFragment && "module creation should not fail");
 
   // Enter the scope of the global module.
-  ModuleScopes.push_back({BeginLoc, GlobalModuleFragment,
+  ModuleScopes.push_back({BeginLoc, TheGlobalModuleFragment,
                           /*ModuleInterface=*/false,
                           /*OuterVisibleModules=*/{}});
-  VisibleModules.setVisible(GlobalModuleFragment, BeginLoc);
+  VisibleModules.setVisible(TheGlobalModuleFragment, BeginLoc);
 
-  return GlobalModuleFragment;
+  return TheGlobalModuleFragment;
 }
 
 void Sema::PopGlobalModuleFragment() {
-  assert(!ModuleScopes.empty() && getCurrentModule()->isGlobalModule() &&
+  assert(!ModuleScopes.empty() &&
+         getCurrentModule()->isExplicitGlobalModule() &&
+         "left the wrong module scope, which is not global module fragment");
+  ModuleScopes.pop_back();
+}
+
+Module *Sema::PushImplicitGlobalModuleFragment(SourceLocation BeginLoc,
+                                               bool IsExported) {
+  Module **M = IsExported ? &TheExportedImplicitGlobalModuleFragment
+                          : &TheImplicitGlobalModuleFragment;
+  if (!*M) {
+    ModuleMap &Map = PP.getHeaderSearchInfo().getModuleMap();
+    *M = Map.createImplicitGlobalModuleFragmentForModuleUnit(
+        BeginLoc, IsExported, getCurrentModule());
+  }
+  assert(*M && "module creation should not fail");
+
+  // Enter the scope of the global module.
+  ModuleScopes.push_back({BeginLoc, *M,
+                          /*ModuleInterface=*/false,
+                          /*OuterVisibleModules=*/{}});
+  VisibleModules.setVisible(*M, BeginLoc);
+  return *M;
+}
+
+void Sema::PopImplicitGlobalModuleFragment() {
+  assert(!ModuleScopes.empty() &&
+         getCurrentModule()->isImplicitGlobalModule() &&
          "left the wrong module scope, which is not global module fragment");
   ModuleScopes.pop_back();
 }

clang/test/Modules/export-language-linkage.cppm

@@ -0,0 +1,52 @@
+// RUN: rm -rf %t
+// RUN: split-file %s %t
+// RUN: cd %t
+//
+// RUN: %clang_cc1 -std=c++20 %t/a.cppm -emit-module-interface -o %t/a.pcm
+// RUN: %clang_cc1 -std=c++20 %t/b.cpp -fmodule-file=a=%t/a.pcm -fsyntax-only -verify
+// RUN: %clang_cc1 -std=c++20 %t/c.cppm -fsyntax-only -verify
+// RUN: %clang_cc1 -module-file-info %t/a.pcm | FileCheck %t/a.cppm
+
+//--- a.cppm
+export module a;
+export extern "C++" int foo() { return 43; }
+export extern "C++" {
+    int a();
+    int b();
+    int c();
+}
+
+export {
+    extern "C++" void f1();
+    extern "C++" void f2();
+    extern "C++" void f3();
+}
+
+extern "C++" void unexported();
+
+// CHECK: Sub Modules:
+// CHECK-NEXT: Implicit Module Fragment '<exported implicit global>'
+// CHECK-NEXT: Implicit Module Fragment '<implicit global>'
+
+//--- b.cpp
+import a;
+int use() {
+    a();
+    b();
+    c();
+    f1();
+    f2();
+    f3();
+    unexported(); // expected-error {{missing '#include'; 'unexported' must be declared before it is used}}
+                   // expected-note@a.cppm:15 {{declaration here is not visible}}
+    return foo();
+}
+
+//--- c.cppm
+export module c;
+extern "C++" {
+    // We can't use `export` in an unnamed module.
+    export int f(); // expected-error {{export declaration can only be used within a module purview}}
+}
+
+extern "C++" export int g(); // expected-error {{export declaration can only be used within a module purview}}