[cfe-commits] r71405 - in /cfe/trunk: include/clang/AST/ include/clang/Frontend/ include/clang/Parse/ lib/AST/ lib/CodeGen/ lib/Frontend/ lib/Parse/ lib/Sema/ test/SemaCXX/ www/

[Sebastian Redl]

Chris Lattner wrote:
> On May 10, 2009, at 11:38 AM, Sebastian Redl wrote:
> Implement C++0x nullptr.
>
>> //===----------------------------------------------------------------------===//
>>
>> @@ -431,6 +434,9 @@
>>       Width = Target.getLongDoubleWidth();
>>       Align = Target.getLongDoubleAlign();
>>       break;
>> +    case BuiltinType::NullPtr:
>> +      Width = Target.getPointerWidth(0); // C++ 3.9.1p11:
>> sizeof(nullptr_t)
>> +      Align = Target.getPointerAlign(0); //   == sizeof(void*)
>>     }
>>     break;
>
> Hi Sebastian,
>
> Though not strictly needed, please do add an explicit break to this case.
>
Sure, was only an oversight.

> +++ cfe/trunk/lib/Sema/SemaExpr.cpp Sun May 10 13:38:11 2009
> @@ -3811,6 +3811,20 @@
>     ImpCastExprToType(rex, lType); // promote the pointer to pointer
>     return ResultTy;
>   }
> +  // C++ allows comparison of pointers with null pointer constants.
> +  if (getLangOptions().CPlusPlus) {
> +    if (lType->isPointerType() && RHSIsNull) {
> +      ImpCastExprToType(rex, lType);
> +      return ResultTy;
> +    }
> +    if (rType->isPointerType() && LHSIsNull) {
> +      ImpCastExprToType(lex, rType);
> +      return ResultTy;
> +    }
> +    // And comparison of nullptr_t with itself.
> +    if (lType->isNullPtrType() && rType->isNullPtrType())
> +      return ResultTy;
> +  }
>
> Can't this be handled by making isNullPointerConstant handle this?  If
> so, LHSIsNull/RHSIsNull are already computed.

As Eli pointed out, that would allow the illegal nullptr == 0.

Sebastian