[cfe-commits] r64969 - /cfe/trunk/lib/CodeGen/CodeGenModule.cpp

[Douglas Gregor]

On Feb 18, 2009, at 2:52 PM, Daniel Dunbar wrote:

> Author: ddunbar
> Date: Wed Feb 18 16:52:09 2009
> New Revision: 64969
>
> URL: http://llvm.org/viewvc/llvm-project?rev=64969&view=rev
> Log:
> Add anti-FIXME.
>
> Modified:
>    cfe/trunk/lib/CodeGen/CodeGenModule.cpp
>
> Modified: cfe/trunk/lib/CodeGen/CodeGenModule.cpp
> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/CodeGenModule.cpp?rev=64969&r1=64968&r2=64969&view=diff
>
> =
> =
> =
> =
> =
> =
> =
> =
> ======================================================================
> --- cfe/trunk/lib/CodeGen/CodeGenModule.cpp (original)
> +++ cfe/trunk/lib/CodeGen/CodeGenModule.cpp Wed Feb 18 16:52:09 2009
> @@ -162,7 +162,10 @@
> /// FIXME: Returning an IdentifierInfo* here is a total hack. We
> /// really need some kind of string abstraction that either stores a
> /// mangled name or stores an IdentifierInfo*. This will require
> -/// changes to the GlobalDeclMap, too.
> +/// changes to the GlobalDeclMap, too. (I disagree, I think what we
> +/// actually need is for Sema to provide some notion of which Decls
> +/// refer to the same semantic decl. We shouldn't need to mangle the
> +/// names and see what comes out the same to figure this out. - DWD)


If you have any thoughts on how to handle this notion of the semantic  
decl, I'd be interested to hear them. We need precisely the same thing  
to handle class template specializations (although I've been referring  
to it as the "canonical" declaration) and improve our diagnostics in  
cases where two declarations for the same thing occur in two different  
scopes (where neither is visible from the other).

	- Doug