Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Clang][Sema] Diagnose unexpanded packs in the template argument lists of function template specializations #76677

Merged
merged 1 commit into from
Jan 3, 2024

Conversation

sdkrystian
Copy link
Member

@sdkrystian sdkrystian commented Jan 1, 2024

This diagnoses unexpanded packs in the unqualified-id of a function template specialization's declarator-id. For example:

template<typename... Ts>
struct A
{
    template<typename U>
    void f();

    template<>
    void f<Ts>(); // error: explicit specialization contains unexpanded parameter pack 'Ts'
};

I moved the handling of template-id's so it happens right after we determine whether we are declaring a function template/function template specialization so diagnostics are issued in lexical order.

@llvmbot llvmbot added clang Clang issues not falling into any other category clang:frontend Language frontend issues, e.g. anything involving "Sema" labels Jan 1, 2024
@llvmbot
Copy link
Collaborator

llvmbot commented Jan 1, 2024

@llvm/pr-subscribers-clang

Author: Krystian Stasiowski (sdkrystian)

Changes

This diagnoses unexpanded packs in the unqualified-id of a function template specialization's declarator-id, e.g.:

template&lt;typename... Ts&gt;
struct A
{
    template&lt;typename U&gt;
    void f();

    template&lt;&gt;
    void f&lt;Ts&gt;(); // error: explicit specialization contains unexpanded parameter pack 'Ts'
};

I moved the handling of template-id's so it happens right after we determine whether we are declaring a function template/function template specialization so diagnostics are issued in lexical order.


Full diff: https://github.com/llvm/llvm-project/pull/76677.diff

2 Files Affected:

  • (modified) clang/lib/Sema/SemaDecl.cpp (+43-46)
  • (modified) clang/test/CXX/temp/temp.decls/temp.variadic/p5.cpp (+12)
diff --git a/clang/lib/Sema/SemaDecl.cpp b/clang/lib/Sema/SemaDecl.cpp
index ffbe317d559995..e94637b4f053e0 100644
--- a/clang/lib/Sema/SemaDecl.cpp
+++ b/clang/lib/Sema/SemaDecl.cpp
@@ -9900,15 +9900,15 @@ Sema::ActOnFunctionDeclarator(Scope *S, Declarator &D, DeclContext *DC,
     // Match up the template parameter lists with the scope specifier, then
     // determine whether we have a template or a template specialization.
     bool Invalid = false;
+    TemplateIdAnnotation *TemplateId =
+        D.getName().getKind() == UnqualifiedIdKind::IK_TemplateId
+            ? D.getName().TemplateId
+            : nullptr;
     TemplateParameterList *TemplateParams =
         MatchTemplateParametersToScopeSpecifier(
             D.getDeclSpec().getBeginLoc(), D.getIdentifierLoc(),
-            D.getCXXScopeSpec(),
-            D.getName().getKind() == UnqualifiedIdKind::IK_TemplateId
-                ? D.getName().TemplateId
-                : nullptr,
-            TemplateParamLists, isFriend, isMemberSpecialization,
-            Invalid);
+            D.getCXXScopeSpec(), TemplateId, TemplateParamLists, isFriend,
+            isMemberSpecialization, Invalid);
     if (TemplateParams) {
       // Check that we can declare a template here.
       if (CheckTemplateDeclScope(S, TemplateParams))
@@ -9921,6 +9921,11 @@ Sema::ActOnFunctionDeclarator(Scope *S, Declarator &D, DeclContext *DC,
         if (Name.getNameKind() == DeclarationName::CXXDestructorName) {
           Diag(NewFD->getLocation(), diag::err_destructor_template);
           NewFD->setInvalidDecl();
+          // Function template with explicit template arguments.
+        } else if (TemplateId) {
+          Diag(D.getIdentifierLoc(), diag::err_function_template_partial_spec)
+              << SourceRange(TemplateId->LAngleLoc, TemplateId->RAngleLoc);
+          NewFD->setInvalidDecl();
         }
 
         // If we're adding a template to a dependent context, we may need to
@@ -9973,6 +9978,11 @@ Sema::ActOnFunctionDeclarator(Scope *S, Declarator &D, DeclContext *DC,
             << FixItHint::CreateRemoval(RemoveRange)
             << FixItHint::CreateInsertion(InsertLoc, "<>");
           Invalid = true;
+
+          // Recover by faking up an empty template argument list.
+          HasExplicitTemplateArgs = true;
+          TemplateArgs.setLAngleLoc(InsertLoc);
+          TemplateArgs.setRAngleLoc(InsertLoc);
         }
       }
     } else {
@@ -9986,6 +9996,33 @@ Sema::ActOnFunctionDeclarator(Scope *S, Declarator &D, DeclContext *DC,
       if (TemplateParamLists.size() > 0)
         // For source fidelity, store all the template param lists.
         NewFD->setTemplateParameterListsInfo(Context, TemplateParamLists);
+
+      // "friend void foo<>(int);" is an implicit specialization decl.
+      if (isFriend && TemplateId)
+        isFunctionTemplateSpecialization = true;
+    }
+
+    // If this is a function template specialization and the unqualified-id of
+    // the declarator-id is a template-id, convert the template argument list
+    // into our AST format and check for unexpanded packs.
+    if (isFunctionTemplateSpecialization && TemplateId) {
+      HasExplicitTemplateArgs = true;
+
+      TemplateArgs.setLAngleLoc(TemplateId->LAngleLoc);
+      TemplateArgs.setRAngleLoc(TemplateId->RAngleLoc);
+      ASTTemplateArgsPtr TemplateArgsPtr(TemplateId->getTemplateArgs(),
+                                         TemplateId->NumArgs);
+      translateTemplateArguments(TemplateArgsPtr, TemplateArgs);
+
+      // FIXME: Should we check for unexpanded packs if this was an (invalid)
+      // declaration of a function template partial specialization? Should we
+      // consider the unexpanded pack context to be a partial specialization?
+      for (const TemplateArgumentLoc &ArgLoc : TemplateArgs.arguments()) {
+        if (DiagnoseUnexpandedParameterPack(
+                ArgLoc, isFriend ? UPPC_FriendDeclaration
+                                 : UPPC_ExplicitSpecialization))
+          NewFD->setInvalidDecl();
+      }
     }
 
     if (Invalid) {
@@ -10438,46 +10475,6 @@ Sema::ActOnFunctionDeclarator(Scope *S, Declarator &D, DeclContext *DC,
            diag::ext_operator_new_delete_declared_inline)
         << NewFD->getDeclName();
 
-    // If the declarator is a template-id, translate the parser's template
-    // argument list into our AST format.
-    if (D.getName().getKind() == UnqualifiedIdKind::IK_TemplateId) {
-      TemplateIdAnnotation *TemplateId = D.getName().TemplateId;
-      TemplateArgs.setLAngleLoc(TemplateId->LAngleLoc);
-      TemplateArgs.setRAngleLoc(TemplateId->RAngleLoc);
-      ASTTemplateArgsPtr TemplateArgsPtr(TemplateId->getTemplateArgs(),
-                                         TemplateId->NumArgs);
-      translateTemplateArguments(TemplateArgsPtr,
-                                 TemplateArgs);
-
-      HasExplicitTemplateArgs = true;
-
-      if (NewFD->isInvalidDecl()) {
-        HasExplicitTemplateArgs = false;
-      } else if (FunctionTemplate) {
-        // Function template with explicit template arguments.
-        Diag(D.getIdentifierLoc(), diag::err_function_template_partial_spec)
-          << SourceRange(TemplateId->LAngleLoc, TemplateId->RAngleLoc);
-
-        HasExplicitTemplateArgs = false;
-      } else if (isFriend) {
-        // "friend void foo<>(int);" is an implicit specialization decl.
-        isFunctionTemplateSpecialization = true;
-      } else {
-        assert(isFunctionTemplateSpecialization &&
-               "should have a 'template<>' for this decl");
-      }
-    } else if (isFriend && isFunctionTemplateSpecialization) {
-      // This combination is only possible in a recovery case;  the user
-      // wrote something like:
-      //   template <> friend void foo(int);
-      // which we're recovering from as if the user had written:
-      //   friend void foo<>(int);
-      // Go ahead and fake up a template id.
-      HasExplicitTemplateArgs = true;
-      TemplateArgs.setLAngleLoc(D.getIdentifierLoc());
-      TemplateArgs.setRAngleLoc(D.getIdentifierLoc());
-    }
-
     // We do not add HD attributes to specializations here because
     // they may have different constexpr-ness compared to their
     // templates and, after maybeAddCUDAHostDeviceAttrs() is applied,
diff --git a/clang/test/CXX/temp/temp.decls/temp.variadic/p5.cpp b/clang/test/CXX/temp/temp.decls/temp.variadic/p5.cpp
index 30ce6b40e1fb5f..3c500c2c4dc4a7 100644
--- a/clang/test/CXX/temp/temp.decls/temp.variadic/p5.cpp
+++ b/clang/test/CXX/temp/temp.decls/temp.variadic/p5.cpp
@@ -376,6 +376,11 @@ namespace Specializations {
   template<typename... Ts>
   struct PrimaryClass<Ts>; // expected-error{{partial specialization contains unexpanded parameter pack 'Ts'}}
 
+  template<typename T, typename... Ts>
+  void PrimaryFunction();
+  template<typename T, typename... Ts>
+  void PrimaryFunction<Ts>(); // expected-error{{function template partial specialization is not allowed}}
+
 #if __cplusplus >= 201402L
   template<typename T, typename... Ts>
   constexpr int PrimaryVar = 0;
@@ -392,6 +397,13 @@ namespace Specializations {
     template<typename U>
     struct InnerClass<U, Ts>; // expected-error{{partial specialization contains unexpanded parameter pack 'Ts'}}
 
+    template<typename... Us>
+    void InnerFunction();
+    template<>
+    void InnerFunction<Ts>(); // expected-error{{explicit specialization contains unexpanded parameter pack 'Ts'}}
+
+    friend void PrimaryFunction<Ts>(); // expected-error{{friend declaration contains unexpanded parameter pack 'Ts'}}
+
 #if __cplusplus >= 201402L
     template<typename... Us>
     constexpr static int InnerVar = 0;

@sdkrystian
Copy link
Member Author

Ping @erichkeane

@cor3ntin
Copy link
Contributor

cor3ntin commented Jan 3, 2024

Can you add a release note?
Otherwise, I think the changes are reasonable. I'd like a second pair of eyes though @erichkeane

Copy link
Collaborator

@erichkeane erichkeane left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

As Corentin said, needs a release note, else LGTM.

@@ -518,6 +518,7 @@ Improvements to Clang's diagnostics
- Clang now diagnoses definitions of friend function specializations, e.g. ``friend void f<>(int) {}``.
- Clang now diagnoses narrowing conversions involving const references.
(`#63151: <https://github.com/llvm/llvm-project/issues/63151>`_).
- Clang now diagnoses unexpanded packs within the template argument lists of function template specializations.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Do you have a github issue that this fixes, or is this just something that never had an issue? If it has one, please include here, else, this is fine.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@erichkeane I just looked -- I don't see an open issue.

@sdkrystian
Copy link
Member Author

Added release note & rebased. I'll need someone to merge this for me :)

@erichkeane erichkeane merged commit 7fbc1de into llvm:main Jan 3, 2024
3 of 4 checks passed
@nico
Copy link
Contributor

nico commented Jan 3, 2024

This broke check-clang-tools: http://45.33.8.238/linux/127018/step_8.txt

Please take a look and revert for now if it takes a while to fix.

erichkeane added a commit that referenced this pull request Jan 3, 2024
…ent lists of function template specializations (#76677)"

This reverts commit 7fbc1de.
erichkeane added a commit that referenced this pull request Jan 3, 2024
…ent lists of function template specializations" (#76876)

Reverts #76677

See discussion here: #76677
@erichkeane
Copy link
Collaborator

Reverted here: #76876

@sdkrystian : please take a look at the regression that nico pointed out and re-submit with considerations for that.

@sdkrystian sdkrystian deleted the fct-unexpanded-pack branch January 5, 2024 13:46
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
clang:frontend Language frontend issues, e.g. anything involving "Sema" clang Clang issues not falling into any other category
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

5 participants