[PATCH] D65510: [clangd] Fix implicit template instatiations appearing as topLevelDecls.

[Haojian Wu via Phabricator via cfe-commits]

hokein added inline comments.


================
Comment at: clang-tools-extra/clangd/ClangdUnit.cpp:68
+  if (const auto *TD = dyn_cast<T>(D))
+    return TD->getTemplateSpecializationKind() == TSK_ImplicitInstantiation;
+  return false;
----------------
ilya-biryukov wrote:
> We also want to skip `TSK_ExplicitInstantiationDeclaration` and  `TSK_ExplicitInstantiationDefinition` here.
> This covers cases like (not sure which one of the two enum values we get, though):
> ```
> template <class T>
> int foo(T) { ... }
> 
> template int foo(int); // we'd rather not traverse these, highlightings will run into the same problems.
> ```
> 
> Semantics I'm describing are roughly similar to `isImplicitInstatiation(D) == !isExplicitInstantion(D)`, where `isExplicitInstantiation` is taken from `CodeComplete.cpp`. (If we ignore `TSK_Undeclared`, which, I believe, should never be encountered in decls passed to HandleTopLevelDecl).
> 
> Please extract the helper from code complete and this one into a separate file (e.g. `clangd/AST.h`) and possibly implement one through the other
> Semantics I'm describing are roughly similar to isImplicitInstatiation(D) == !isExplicitInstantion(D), 

I think there is a typo here, I believe you mean `isImplicitInstantiation(D) == !isExplicitSpecialization(D) ` (in CodeComplete.cpp, it checks whether a Decl is an explicit **specialization**).


================
Comment at: clang-tools-extra/clangd/unittests/SemanticHighlightingTests.cpp:219
+    R"cpp(
+      class $Class[[Foo]] {
+      public:
----------------
instead adding a highlighting test, could we add this testcase to `ClangdUnitTests.cpp` instead? 


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D65510/new/

https://reviews.llvm.org/D65510