[cfe-commits] r68128 - in /cfe/trunk: include/clang/AST/TemplateName.h include/clang/Basic/DiagnosticSemaKinds.td lib/AST/TemplateName.cpp lib/Sema/Sema.h lib/Sema/SemaTemplateInstantiate.cpp test/SemaTemplate/instantiate-template-template-parm.cpp test/SemaTemplate/metafun-apply.cpp

[Douglas Gregor]

On Mar 31, 2009, at 12:43 PM, Chris Lattner wrote:

> On Mar 31, 2009, at 11:38 AM, Douglas Gregor wrote:
>> URL: http://llvm.org/viewvc/llvm-project?rev=68128&view=rev
>> Log:
>> Implement template instantiation for template names, including both
>> template template parameters and dependent template names. For
>> example, the oft-mentioned
>>
>> typename MetaFun::template apply<T1, T2>::type
>>
>> can now be instantiated, with the appropriate name lookup for  
>> "apply".
>
> Nice!  Some nits:
>
>> +++ cfe/trunk/include/clang/AST/TemplateName.h Tue Mar 31 13:38:02  
>> 2009
>> @@ -99,6 +99,10 @@
>>  /// \brief Print the template name.
>>  void Print(llvm::raw_ostream &OS) const;
>>
>> +  /// \brief Debugging aid that dumps the template name to standard
>> +  /// error.
>> +  void Dump() const;
>
> Most of the other dump/print methods are lower case.  Should these  
> be lower case also?

Sure. Same fix in NestedNameSpecifier, too.

>>
>> +++ cfe/trunk/lib/AST/TemplateName.cpp Tue Mar 31 13:38:02 2009
>> @@ -14,6 +14,7 @@
>> #include "clang/AST/DeclTemplate.h"
>> #include "clang/AST/NestedNameSpecifier.h"
>> #include "llvm/Support/raw_ostream.h"
>> +#include <stdio.h>
>
> <cstdio>?
>
>
>> +void TemplateName::Dump() const {
>> +  std::string Result;
>> +  {
>> +    llvm::raw_string_ostream OS(Result);
>> +    Print(OS);
>> +  }
>> +  fprintf(stderr, "%s", Result.c_str());
>> +}
>
> Please just use Print(llvm::errs()), which eliminates the need for  
> <stdio>

Much better! Thanks.

>> +TemplateName
>> +Sema::InstantiateTemplateName(TemplateName Name, SourceLocation Loc,
>> +                              const TemplateArgument *TemplateArgs,
>> +                              unsigned NumTemplateArgs) {
>> +  if (TemplateTemplateParmDecl *TTP
>> +        = dyn_cast_or_null<TemplateTemplateParmDecl>(
>> +                                                  
>> Name.getAsTemplateDecl())) {
>> +    assert(TTP->getDepth() == 0 &&
>> +           "Cannot reduce depth of a template template parameter");
>> +    assert(TTP->getPosition() < NumTemplateArgs && "Wrong # of  
>> template args");
>> +    assert(dyn_cast_or_null<ClassTemplateDecl>(
>
> Maybe Casting.h should grow an isa_or_null?


I'm doing something ugly there anyway. It's cleaner, now.

Thanks for the review!

	- Doug