[cfe-dev] Objective-C tidy up

[David Chisnall]

On 3 Jun 2008, at 18:27, Chris Lattner wrote:

> I fail to see the point.  "this" in C++ has the exact same behavior  
> and should also be modeled with PredefinedExpr.

There is one important difference with this.  In C++, this is a  
keyword, so the following is an invalid C++ program:

int main(void)
{
     int this = 0;
     return this;
}

In contrast, the following is a valid Objective-C program:

int main(void)
{
     int self = 0;
     return self;
}

Any function in Objective-C can use self and _cmd as identifier  
names.  Any occurrence of the 'this' token in a C++ program refers to  
an instance of an object with a type defined by the context.   
Occurrences of self / _cmd in an Objective-C program refer to some  
variable defined in some way in some scope, with an implicit  
declaration if the scope happens to be a method.  This means that,  
unless you want to complicate parsing a lot, self and _cmd want to be  
some kind of decl refs, when encountered in the source code (this was  
originally done with super, where it was a decl ref with an implicit  
cast, but I had to change it because super really does require special  
handling).  By the way, it is not uncommon to use self and _cmd  
symbols in Objective-C programs when you write method-like functions  
that you want to call directly, without going via the dynamic dispatch  
mechanism (the equivalent of C++ non-virtual member functions).

So, the question is not what should occurrences of the self and _cmd  
tokens be turned into (although that might be a separate question),  
it's what the entries in the decl map should be populated with.  As  
far as my code goes, this can be pretty much any sort of decl.  It  
needs to have a type associated with it for Sema to be able to be able  
to do type checking (for self, this is a pointer to an instance of the  
current class, e.g. NSObject*, for _cmd it is SEL).  For CodeGen, it  
needs to have a name.

Steve - I don't intend to make any requirements on the calling  
convention.  For the Étoilé runtime, _cmd is a field of a structure  
passed as the second argument, not the second argument.  As long as  
CodeGen can get the argument from somewhere and create %self.addr and  
%_cmd.addr in the function preamble, it won't be a problem (currently  
I do this in the runtime-agnostic part.  I plan on moving this into  
the CGObjCRuntime class, so that all of the traditional runtimes can  
do it this way, but the Étoilé runtime and potentially COLA or a  
future Apple runtime can do it a different way).

David