[cfe-dev] [PATCH]: Sema support for C++ classes

[Chris Lattner]

On Jun 27, 2008, at 5:23 PM, Argiris Kirtzidis wrote:
>>> I'll fix this once Sema support is in place.
>>
>> Ok!  One meta question: the PushDeclContext/pop context logic has  
>> to keep the CurMethodDecl members up to date, and are somewhat  
>> complex.  Would it be better to change direct accesses to  
>> CurMethodDecl to call getCurMethodDecl(), which would just look at  
>> the current context?
>
> Sounds great! Shall I make a separate commit for this ?

Yes, please do it as a separate patch, thanks!

>> Are methods in C++ officially known as "member functions" or  
>> "methods"?
>
> The standard uses "member functions".

Ok!

>> +  else
>> +    NewFD = FieldDecl::Create(Context, Loc, II, T, BitWidth);
>>
>> likewise.
>>
>
> Using plain FieldDecls for simple structs should be easy, I'm not  
> sure about CXX/RecordDecl.
> Can we add this as is so we have the functionality and tests in, and  
> later look for ways to optimize it ?

I'm fine with making it a fixme for now, and addressing it later as  
functionality matures.

>> +Sema::DeclTy *
>> +Sema::ActOnCXXMemberDeclarator(Scope *S, AccessSpecifier AS,  
>> Declarator &D,
>>
>>
>> +  assert(II && "No identifier ?");
>>
>> does this abort on unnamed bitfields like "int : 4;" ?
>
> This requires changes to the parser too. Can we fix it after the  
> patch goes in so I can make one "unified" Parser+Sema+test commit,  
> specific to this case ?

Yep, this patch is definite progress forward.  I'm happy with forward  
progress (even if it doesn't solve all problems) as long as it doesn't  
regress anything that currently works.

Thanks Argiris!

-Chris