[clang] Track function template instantiation from definition (llvm#110387)

This fixes instantiation of definition for friend function templates,
when the declaration found and the one containing the definition
have different template contexts.

In these cases, the the function declaration corresponding to the
definition is not available; it may not even be instantiated at all.

So this patch adds a bit which tracks which function template
declaration was instantiated from the member template.
It's used to find which primary template serves as a context
for the purpose of obtaining the template arguments needed
to instantiate the definition.

Fixes llvm#55509

Author: mizvekov

clang/docs/ReleaseNotes.rst

@@ -474,6 +474,7 @@ Bug Fixes to C++ Support
 - Fixed an assertion failure in debug mode, and potential crashes in release mode, when
   diagnosing a failed cast caused indirectly by a failed implicit conversion to the type of the constructor parameter.
 - Fixed an assertion failure by adjusting integral to boolean vector conversions (#GH108326)
+- Clang is now better at keeping track of friend function template instance contexts. (#GH55509)
 - Fixed an issue deducing non-type template arguments of reference type. (#GH73460)
 - Fixed an issue in constraint evaluation, where type constraints on the lambda expression
   containing outer unexpanded parameters were not correctly expanded. (#GH101754)

clang/include/clang/AST/Decl.h

@@ -2299,6 +2299,13 @@ class FunctionDecl : public DeclaratorDecl,
     FunctionDeclBits.IsLateTemplateParsed = ILT;
   }
 
+  bool isInstantiatedFromMemberTemplate() const {
+    return FunctionDeclBits.IsInstantiatedFromMemberTemplate;
+  }
+  void setInstantiatedFromMemberTemplate(bool Val = true) {
+    FunctionDeclBits.IsInstantiatedFromMemberTemplate = Val;
+  }
+
   /// Whether this function is "trivial" in some specialized C++ senses.
   /// Can only be true for default constructors, copy constructors,
   /// copy assignment operators, and destructors.  Not meaningful until

clang/include/clang/AST/DeclBase.h

@@ -1763,6 +1763,8 @@ class DeclContext {
     uint64_t HasImplicitReturnZero : 1;
     LLVM_PREFERRED_TYPE(bool)
     uint64_t IsLateTemplateParsed : 1;
+    LLVM_PREFERRED_TYPE(bool)
+    uint64_t IsInstantiatedFromMemberTemplate : 1;
 
     /// Kind of contexpr specifier as defined by ConstexprSpecKind.
     LLVM_PREFERRED_TYPE(ConstexprSpecKind)
@@ -1813,7 +1815,7 @@ class DeclContext {
   };
 
   /// Number of inherited and non-inherited bits in FunctionDeclBitfields.
-  enum { NumFunctionDeclBits = NumDeclContextBits + 31 };
+  enum { NumFunctionDeclBits = NumDeclContextBits + 32 };
 
   /// Stores the bits used by CXXConstructorDecl. If modified
   /// NumCXXConstructorDeclBits and the accessor
@@ -1824,12 +1826,12 @@ class DeclContext {
     LLVM_PREFERRED_TYPE(FunctionDeclBitfields)
     uint64_t : NumFunctionDeclBits;
 
-    /// 20 bits to fit in the remaining available space.
+    /// 19 bits to fit in the remaining available space.
     /// Note that this makes CXXConstructorDeclBitfields take
     /// exactly 64 bits and thus the width of NumCtorInitializers
     /// will need to be shrunk if some bit is added to NumDeclContextBitfields,
     /// NumFunctionDeclBitfields or CXXConstructorDeclBitfields.
-    uint64_t NumCtorInitializers : 17;
+    uint64_t NumCtorInitializers : 16;
     LLVM_PREFERRED_TYPE(bool)
     uint64_t IsInheritingConstructor : 1;
 
@@ -1843,7 +1845,7 @@ class DeclContext {
   };
 
   /// Number of inherited and non-inherited bits in CXXConstructorDeclBitfields.
-  enum { NumCXXConstructorDeclBits = NumFunctionDeclBits + 20 };
+  enum { NumCXXConstructorDeclBits = NumFunctionDeclBits + 19 };
 
   /// Stores the bits used by ObjCMethodDecl.
   /// If modified NumObjCMethodDeclBits and the accessor

clang/include/clang/AST/DeclTemplate.h

@@ -1008,6 +1008,15 @@ class FunctionTemplateDecl : public RedeclarableTemplateDecl {
     return getTemplatedDecl()->isThisDeclarationADefinition();
   }
 
+  bool isCompatibleWithDefinition() const {
+    return getTemplatedDecl()->isInstantiatedFromMemberTemplate() ||
+           isThisDeclarationADefinition();
+  }
+  void setInstantiatedFromMemberTemplate(FunctionTemplateDecl *D) {
+    getTemplatedDecl()->setInstantiatedFromMemberTemplate();
+    RedeclarableTemplateDecl::setInstantiatedFromMemberTemplate(D);
+  }
+
   /// Return the specialization with the provided arguments if it exists,
   /// otherwise return the insertion point.
   FunctionDecl *findSpecialization(ArrayRef<TemplateArgument> Args,

clang/include/clang/Sema/Sema.h

@@ -13027,6 +13027,12 @@ class Sema final : public SemaBase {
       std::optional<ArrayRef<TemplateArgument>> Innermost = std::nullopt,
       bool RelativeToPrimary = false, bool ForConstraintInstantiation = false);
 
+  void getTemplateInstantiationArgs(
+      MultiLevelTemplateArgumentList &Result, const NamedDecl *D,
+      const DeclContext *DC = nullptr, bool Final = false,
+      std::optional<ArrayRef<TemplateArgument>> Innermost = std::nullopt,
+      bool RelativeToPrimary = false, bool ForConstraintInstantiation = false);
+
   /// RAII object to handle the state changes required to synthesize
   /// a function body.
   class SynthesizedFunctionScope {

clang/lib/AST/Decl.cpp

@@ -3067,6 +3067,7 @@ FunctionDecl::FunctionDecl(Kind DK, ASTContext &C, DeclContext *DC,
   FunctionDeclBits.IsIneligibleOrNotSelected = false;
   FunctionDeclBits.HasImplicitReturnZero = false;
   FunctionDeclBits.IsLateTemplateParsed = false;
+  FunctionDeclBits.IsInstantiatedFromMemberTemplate = false;
   FunctionDeclBits.ConstexprKind = static_cast<uint64_t>(ConstexprKind);
   FunctionDeclBits.BodyContainsImmediateEscalatingExpression = false;
   FunctionDeclBits.InstantiationIsPending = false;

clang/lib/Sema/SemaTemplateDeduction.cpp

@@ -4007,22 +4007,7 @@ TemplateDeductionResult Sema::FinishTemplateArgumentDeduction(
   if (FunctionTemplate->getFriendObjectKind())
     Owner = FunctionTemplate->getLexicalDeclContext();
   FunctionDecl *FD = FunctionTemplate->getTemplatedDecl();
-  // additional check for inline friend,
-  // ```
-  //   template <class F1> int foo(F1 X);
-  //   template <int A1> struct A {
-  //     template <class F1> friend int foo(F1 X) { return A1; }
-  //   };
-  //   template struct A<1>;
-  //   int a = foo(1.0);
-  // ```
-  const FunctionDecl *FDFriend;
-  if (FD->getFriendObjectKind() == Decl::FriendObjectKind::FOK_None &&
-      FD->isDefined(FDFriend, /*CheckForPendingFriendDefinition*/ true) &&
-      FDFriend->getFriendObjectKind() != Decl::FriendObjectKind::FOK_None) {
-    FD = const_cast<FunctionDecl *>(FDFriend);
-    Owner = FD->getLexicalDeclContext();
-  }
+
   MultiLevelTemplateArgumentList SubstArgs(
       FunctionTemplate, CanonicalDeducedArgumentList->asArray(),
       /*Final=*/false);

clang/lib/Sema/SemaTemplateInstantiate.cpp

@@ -512,13 +512,13 @@ struct TemplateInstantiationArgumentCollecter
 
 } // namespace
 
-MultiLevelTemplateArgumentList Sema::getTemplateInstantiationArgs(
-    const NamedDecl *ND, const DeclContext *DC, bool Final,
+void Sema::getTemplateInstantiationArgs(
+    MultiLevelTemplateArgumentList &Result, const NamedDecl *ND,
+    const DeclContext *DC, bool Final,
     std::optional<ArrayRef<TemplateArgument>> Innermost, bool RelativeToPrimary,
     bool ForConstraintInstantiation) {
   assert((ND || DC) && "Can't find arguments for a decl if one isn't provided");
   // Accumulate the set of template argument lists in this structure.
-  MultiLevelTemplateArgumentList Result;
   const Decl *CurDecl = ND;
 
   if (!CurDecl)
@@ -529,6 +529,17 @@ MultiLevelTemplateArgumentList Sema::getTemplateInstantiationArgs(
   do {
     CurDecl = Collecter.Visit(const_cast<Decl *>(CurDecl));
   } while (CurDecl);
+}
+
+MultiLevelTemplateArgumentList Sema::getTemplateInstantiationArgs(
+    const NamedDecl *ND, const DeclContext *DC, bool Final,
+    std::optional<ArrayRef<TemplateArgument>> Innermost, bool RelativeToPrimary,
+    bool ForConstraintInstantiation) {
+  assert((ND || DC) && "Can't find arguments for a decl if one isn't provided");
+  // Accumulate the set of template argument lists in this structure.
+  MultiLevelTemplateArgumentList Result;
+  getTemplateInstantiationArgs(Result, ND, DC, Final, Innermost,
+                               RelativeToPrimary, ForConstraintInstantiation);
   return Result;
 }
 

clang/lib/Sema/SemaTemplateInstantiateDecl.cpp

@@ -5214,8 +5214,26 @@ void Sema::InstantiateFunctionDefinition(SourceLocation PointOfInstantiation,
     RebuildTypeSourceInfoForDefaultSpecialMembers();
     SetDeclDefaulted(Function, PatternDecl->getLocation());
   } else {
-    MultiLevelTemplateArgumentList TemplateArgs = getTemplateInstantiationArgs(
-        Function, Function->getLexicalDeclContext());
+    DeclContext *DC = Function;
+    MultiLevelTemplateArgumentList TemplateArgs;
+    if (auto *Primary = Function->getPrimaryTemplate();
+        Primary &&
+        !isGenericLambdaCallOperatorOrStaticInvokerSpecialization(Function)) {
+      auto It = llvm::find_if(Primary->redecls(),
+                              [](const RedeclarableTemplateDecl *RTD) {
+                                return cast<FunctionTemplateDecl>(RTD)
+                                    ->isCompatibleWithDefinition();
+                              });
+      assert(It != Primary->redecls().end() &&
+             "Should't get here without a definition");
+      DC = (*It)->getLexicalDeclContext();
+      if (Function->getTemplateSpecializationKind() !=
+          TSK_ExplicitSpecialization)
+        TemplateArgs.addOuterTemplateArguments(
+            Function, Function->getTemplateSpecializationArgs()->asArray(),
+            /*Final=*/false);
+    }
+    getTemplateInstantiationArgs(TemplateArgs, /*D=*/nullptr, DC);
 
     // Substitute into the qualifier; we can get a substitution failure here
     // through evil use of alias templates.

clang/lib/Serialization/ASTReaderDecl.cpp

@@ -1087,6 +1087,7 @@ void ASTDeclReader::VisitFunctionDecl(FunctionDecl *FD) {
   FD->setHasImplicitReturnZero(FunctionDeclBits.getNextBit());
   FD->setIsMultiVersion(FunctionDeclBits.getNextBit());
   FD->setLateTemplateParsed(FunctionDeclBits.getNextBit());
+  FD->setInstantiatedFromMemberTemplate(FunctionDeclBits.getNextBit());
   FD->setFriendConstraintRefersToEnclosingTemplate(
       FunctionDeclBits.getNextBit());
   FD->setUsesSEHTry(FunctionDeclBits.getNextBit());

clang/lib/Serialization/ASTWriterDecl.cpp

@@ -626,7 +626,7 @@ void ASTDeclWriter::VisitDeclaratorDecl(DeclaratorDecl *D) {
 }
 
 void ASTDeclWriter::VisitFunctionDecl(FunctionDecl *D) {
-  static_assert(DeclContext::NumFunctionDeclBits == 44,
+  static_assert(DeclContext::NumFunctionDeclBits == 45,
                 "You need to update the serializer after you change the "
                 "FunctionDeclBits");
 
@@ -732,6 +732,7 @@ void ASTDeclWriter::VisitFunctionDecl(FunctionDecl *D) {
   FunctionDeclBits.addBit(D->hasImplicitReturnZero());
   FunctionDeclBits.addBit(D->isMultiVersion());
   FunctionDeclBits.addBit(D->isLateTemplateParsed());
+  FunctionDeclBits.addBit(D->isInstantiatedFromMemberTemplate());
   FunctionDeclBits.addBit(D->FriendConstraintRefersToEnclosingTemplate());
   FunctionDeclBits.addBit(D->usesSEHTry());
   Record.push_back(FunctionDeclBits);

clang/test/SemaTemplate/GH55509.cpp

@@ -0,0 +1,101 @@
+// RUN: %clang_cc1 -fsyntax-only -verify -std=c++26 %s
+
+namespace t1 {
+  template<int N> struct A {
+    template<class C> friend auto cica(const A<N-1>&, C) {
+      return N;
+    }
+  };
+
+  template<> struct A<0> {
+    template<class C> friend auto cica(const A<0>&, C);
+    // expected-note@-1 {{declared here}}
+  };
+
+  void test() {
+    cica(A<0>{}, 0);
+    // expected-error@-1 {{function 'cica<int>' with deduced return type cannot be used before it is defined}}
+
+    (void)A<1>{};
+    cica(A<0>{}, 0);
+  }
+} // namespace t1
+namespace t2 {
+  template<int N> struct A {
+    template<class C> friend auto cica(const A<N-1>&, C) {
+      return N;
+    }
+  };
+
+  template<> struct A<0> {
+    template<class C> friend auto cica(const A<0>&, C);
+  };
+
+  template <int N, class = decltype(cica(A<N>{}, nullptr))>
+  void MakeCica();
+  // expected-note@-1 {{candidate function}}
+
+  template <int N> void MakeCica(A<N+1> = {});
+  // expected-note@-1 {{candidate function}}
+
+  void test() {
+    MakeCica<0>();
+
+    MakeCica<0>();
+    // expected-error@-1 {{call to 'MakeCica' is ambiguous}}
+  }
+} // namespace t2
+namespace t3 {
+  template<int N> struct A {
+    template<class C> friend auto cica(const A<N-1>&, C) {
+      return N-1;
+    }
+  };
+
+  template<> struct A<0> {
+    template<class C> friend auto cica(const A<0>&, C);
+  };
+
+  template <int N, class AT, class = decltype(cica(AT{}, nullptr))>
+  static constexpr bool MakeCica(int);
+
+  template <int N, class AT>
+  static constexpr bool MakeCica(short, A<N+1> = {});
+
+  template <int N, class AT = A<N>, class Val = decltype(MakeCica<N, AT>(0))>
+  static constexpr bool has_cica = Val{};
+
+  constexpr bool cica2 = has_cica<0> || has_cica<0>;
+} // namespace t3
+namespace t4 {
+  template<int N> struct A {
+    template<class C> friend auto cica(const A<N-1>&, C);
+  };
+
+  template<> struct A<0> {
+    template<class C> friend auto cica(const A<0>&, C) {
+      C a;
+    }
+  };
+
+  template struct A<1>;
+
+  void test() {
+    cica(A<0>{}, 0);
+  }
+} // namespace t4
+namespace regression1 {
+  template <class> class A;
+
+  template <class T> [[gnu::abi_tag("TAG")]] void foo(A<T>);
+
+  template <class> struct A {
+    friend void foo <>(A);
+  };
+
+  template struct A<int>;
+
+  template <class T> [[gnu::abi_tag("TAG")]] void foo(A<T>) {}
+
+  template void foo<int>(A<int>);
+} // namespace regression1