[cfe-dev] extern "C" support

[Chris Lattner]

On Jan 10, 2008, at 1:51 PM, Mike Stump wrote:
> On Jan 9, 2008, at 1:36 PM, Chris Lattner wrote:
>> This is looking very nice.  Some last remaining nit-picky details:
>
> Time will tell if these are truly the last of the nit-pics...  :-)

Hehe, the first significant patch is always the hardest.

>> +void LinkageSpecDecl::EmitInRec(Serializer& S) const {
>> +  Decl::EmitInRec(S);
>> +  S.EmitInt(getLanguage());
>> +  S.EmitPtr(D);
>> +}
>>
>> Emitting the language enum as part of the serialization format  
>> implicitly makes the enum value part of the serialized encoding.   
>> I'm fine with this, but please add a comment about this issue above  
>> the enum definition:
>
> Done.  Hum, yeah, would be nice to have a nice clean separation  
> between the two.  For an abi, I think I'd rather have the value  
> chosen by the dwarf standard, I think they both enumerate languages  
> and select values.  [ pause ]  Ok, I fixed up the code.

Yep.  At this point, we really don't care about 100% stability  
(because things are changing and immature), but when we do start  
caring we want to know what pieces implicitly define the abi.

>> +  if (strncmp (Lang, "\"C\"", StrSize) == 0)
>> +    Language = LinkageSpecDecl::lang_c;
>> +  else if (strncmp (Lang, "\"C++\"", StrSize) == 0)
>> +    Language = LinkageSpecDecl::lang_cxx;
>>
>> I'm not sure if strncmp is the right thing here.  Will it work if  
>> "Lang" is not null terminated?
>
> Uhm, this is why strncmp was invented!?  Yes.  strncmp only fails  
> when languages are defined to have '\0' in their names.  When that  
> happens, we have bigger problems.

Ah ok, this works because you're comparing StrSize bytes, not StrSize 
+1, silly me :)

>> Finally, should 'extern "c"' be allowed?
>
> No.

Ok!

> Ok, We'll try this one...

Very nice, applied:
http://lists.cs.uiuc.edu/pipermail/cfe-commits/Week-of-Mon-20080107/003743.html

-Chris