[cfe-commits] r75314 - in /cfe/trunk: include/clang/AST/ include/clang/Analysis/PathSensitive/ lib/AST/ lib/Analysis/ lib/CodeGen/ lib/Frontend/ lib/Sema/ test/CodeGenObjC/ test/PCH/ test/SemaObjC/ test/SemaObjCXX/

[steve naroff]

On Jul 13, 2009, at 7:07 PM, Fariborz Jahanian wrote:

>
> On Jul 13, 2009, at 1:18 PM, Chris Lattner wrote:
>
>>>
>>> This following two issues are a result of moving away from the C- 
>>> structure dependency (which is a hack). More info...
>>>
>>> We no longer treat "struct objc_class *" as synonymous with  
>>> "Class". This is a side-effect of removing  
>>> "ASTContext::isObjCClassStructType(T)".
>>>
>>> @interface B1
>>> {
>>>     struct objc_class *isa;
>>>     Int1 *sBase;
>>>     char c;
>>> }
>>> @end
>>>
>>> Since it does effect GCC binary compatibility, we could certainly  
>>> add back a hack to make them synonymous.
>>
>> I think that we should do this for @encode and C++ mangling.   
>> Compatibility is very important at this level because it isn't a  
>> matter of the compiler rejecting or producing a warning, it is a  
>> silent miscompilation.
>
> I want to affirm Chris's point that we just need to hack objc and c+ 
> +'s mangling logic. We don't need the extra complication of making  
> them
> synonymous everywhere.
>

Sounds good to me.

snaroff

> - fariborz
>
>>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20090713/453b325e/attachment.html>