[cfe-commits] C++0x Attributes Patch - INCOMPLETE

[Douglas Gregor]

On Sep 13, 2009, at 5:00 PM, Daniel Dunbar wrote:

> On Sun, Sep 13, 2009 at 12:35 PM, Chris Lattner <clattner at apple.com>  
> wrote:
>> On Sep 13, 2009, at 2:17 AM, Argyris Kyrtzidis wrote:
>>> +  if (After)
>>> +    *After = Tok.getKind();
>>> +
>>> +  return true;
>>> +}
>>>
>>> Why do we need to verify that ']]' follows '[[', isn't it enough for
>>> isCXX0XAttributeSpecifier to just return true if the current tokens
>>> are '[[' ?
>>
>> In ObjC++ we have an ambiguity in an expression context, because we
>> can have this ObjC code:
>>
>> [[obj method] method]
>
> I haven't read the patch, and I don't know the tradeoffs involved, but
> should we only bother with tentative parsing in ObjC++ mode?


 From a performance standpoint, that's the right thing to do. That  
said, I worry a little bit that this code won't get much testing at  
all, given that Objective-C++0x isn't the most common dialect out in  
the wild :)

	- Doug