[cfe-dev] Smart pointer usage in the Parser/Actions

[Douglas Gregor]

On Dec 19, 2008, at 3:39 PM, steve naroff wrote:
> On Dec 19, 2008, at 3:09 PM, Fariborz Jahanian wrote:
> A variation on the FIXME is to have a subclass of ParmVarDecl (e.g.  
> ObjCParmVarDecl) that contains the original type (to support method  
> encoding). This is still wasteful, however the waste is isolated to  
> ObjC. Note that C++ is already wasting space (with the DefaultArg  
> slot). Conceptually, I like the idea of having ObjC and C++  
> subclasses of ParmVarDecl.

I'm not a fan of having separate ObjC/C++ subclasses for Decls unless  
there are differences in the data we need to store in the AST. In this  
case, the difference isn't whether we're in Objective-C or not; it's  
whether or not the original type is different from the converted type.  
So, why not have ParmVarWithOriginalTypeDecl be a subclass of  
ParmVarDecl, and create ParmVarWithOriginalTypeDecls whenever we've  
actually performed some kind of promotion of the declaration's type?  
That will capture all of the information in the source code while not  
special-casing Objective-C or introducing an extra slot into  
ParmVarDecls that aren't promoted.

I've been meaning to do the same thing with FieldDecl, by creating a  
subclass BitFieldDecl that's has the extra Stmt* containing the  
bitfield size. The FieldDecl class's interface won't change... but it  
will perform a dyn_cast<BitFieldDecl> in FieldDecl::getBitWidth(), so  
that clients of FieldDecl's won't ever have to know about BitFieldDecl.

	- Doug