[cfe-commits] [REVIEW] Literal operator semantic improvements

[Douglas Gregor]

On Jan 7, 2010, at 12:53 AM, Sean Hunt wrote:

> This patch actually gets CXX literal operator names into the  
> DeclarationNameTable (oops), and also adds semantic analysis of the  
> literal operator's signature.

A couple of comments:

+/// CheckLiteralOperatorDeclaration - Check whether the declaration
+/// of this literal operator function is well-formed. If so, returns
+/// false; otherwise, emits appropriate diagnostics and returns true.
+bool Sema::CheckLiteralOperatorDeclaration(FunctionDecl *FnDecl) {
+  DeclContext *DC = FnDecl->getDeclContext();
+  if (!isa<NamespaceDecl>(DC) && !isa<TranslationUnitDecl>(DC)) {
+    Diag(FnDecl->getLocation(),  
diag::err_literal_operator_outside_namespace)
+      << FnDecl->getDeclName();
+    return true;
+  }

Can a literal operator be declared in a linkage-specification, e.g.,

	extern "C++" {
		void operator "" good (char16_t);
	}

?

This code would reject that.

+    // The second and final parameter must be an std::size_t
+    if (Context.hasSameType((*Param)->getType(), Context.getSizeType 
()) &&
+        ++Param == FnDecl->param_end())
+      Valid = true;

It can probably be a const-qualified std::size_t, right? We allow this  
for, e.g., operator new's size_t parameter.

Everything else seems reasonable. Please go ahead and commit once  
those issues above are resolved.

	- Doug