[cfe-commits] r80690 - in /cfe/trunk: include/clang/AST/DeclBase.h lib/Sema/IdentifierResolver.cpp

[Ted Kremenek]

On Sep 1, 2009, at 10:46 AM, Douglas Gregor wrote:

>> On Sep 1, 2009, at 10:31 AM, Douglas Gregor wrote:
>>
>>>> Can we make pointer comparisons of DeclContext objects illegal  
>>>> (e.g., a private operator== method)?
>>>
>>> Sadly, no :(
>>>
>>> An overloaded operator has to have at least on parameter of class  
>>> or enumeration type (or reference to one of those). The signature  
>>> we would need, e.g.,
>>>
>>> private:
>>>   friend bool operator==(DeclContext *, DeclContext *);
>>>
>>> does not qualify.
>>>
>>> - Doug
>>
>> Ah, that's right.  Would it be possible to do this if we were using  
>> DeclContext& instead of DeclContext* in all places where we  
>> referred to DeclContext objects?
>
>
> Yes, although I fear that might be a bit awkward to use. We're often  
> walking parent contexts, lookup contexts, etc., so that DeclContext&  
> wouldn't be appropriate. Now, we could do a whole-sale replacement  
> of DeclContext * for a new smart pointer DeclContextPtr... then we  
> can do whatever we want, including make operator== do the right thing.

Although I think that many would think this was unnecessary overhead,  
I really like the smart pointer idea.  It nicely defines away a  
category of insidious bugs.