[cfe-commits] r49058 - in /cfe/trunk: include/clang/AST/Decl.h include/clang/AST/DeclObjC.h include/clang/Basic/DiagnosticKinds.def lib/AST/Decl.cpp lib/Sema/Sema.h lib/Sema/SemaDecl.cpp lib/Sema/SemaDeclObjC.cpp lib/Sema/SemaExpr.cpp lib/Sema/SemaExprObjC.cpp test/Sema/alias-test-1.m

[Chris Lattner]

On Apr 1, 2008, at 5:38 PM, Steve Naroff wrote:

>>
>> Also, can 'isTypeName' just call LookupDecl?
>>
>
> I considered this, however LookupDecl calls LazilyCreateBuiltin  
> (which I didn't think made sense when being called from isTypeName).

Good point.

> If I am wrong (and this is harmless), then I can make the change.
> If not, we could consider complicating the API a bit.

You're right, isTypeName shouldn't trigger this.  However, the code  
can still be factored somehow.  Maybe LookupDecl can get a bool  
(defaulting to false) to say whether it should create the decl?

> Yet another solution I considered was to have LookupDecl/isTypeName  
> call a helper to walk the scope change (this would give us the  
> single point of control we'd like).

This is another good solution.  This is probably even cleaner than the  
bool.

>>>
> Agreed. I plan on tweaking this later...

Understood, thanks Steve,

-Chris