[cfe-commits] r61144 - /cfe/trunk/lib/AST/DeclObjC.cpp

[steve naroff]

On Dec 17, 2008, at 1:25 PM, Fariborz Jahanian wrote:

> Author: fjahanian
> Date: Wed Dec 17 12:25:24 2008
> New Revision: 61144
>
> URL: http://llvm.org/viewvc/llvm-project?rev=61144&view=rev
> Log:
> Added comment to Steve's patch to clarify the case.
>
>
> Modified:
>    cfe/trunk/lib/AST/DeclObjC.cpp
>
> Modified: cfe/trunk/lib/AST/DeclObjC.cpp
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/AST/DeclObjC.cpp?rev=61144&r1=61143&r2=61144&view=diff
>
> =
> =
> =
> =
> =
> =
> =
> =
> ======================================================================
> --- cfe/trunk/lib/AST/DeclObjC.cpp (original)
> +++ cfe/trunk/lib/AST/DeclObjC.cpp Wed Dec 17 12:25:24 2008
> @@ -369,6 +369,9 @@
> ///
> FieldDecl *ObjCInterfaceDecl::lookupFieldDeclForIvar(ASTContext  
> &Context,
>                                                      const  
> ObjCIvarDecl *ivar) {
> +  /* When a super class's ivar is referenced in the subclass method  
> with no ivar
> +     of its own, record for the sub-class is not built yet. Build  
> it lazily
> +     here. */

Fariborz,

Thanks for adding the comment, however I'm hoping it's temporary.

I think it makes sense to always build it lazily (which I believe is  
your plan).

Currently, the only lazy case is when you have no ivars & no curlies:

@interface X : S @end

Note that this obscure case is *not* lazy:

@interface X : S { } @end

 From my perspective, having some cases be lazy and some cases not be  
lazy is confusing (and more likely to be buggy).

No rush on this (since the patch/comment will work fine for now). I  
just wanted to make sure we revisit this...

Thanks,

snaroff

>
>   if (!RecordForDecl)
>     addRecordToClass(Context);
>   assert(RecordForDecl && "lookupFieldDeclForIvar no storage for  
> class");
>
>
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits