Objc: Extended setter property not synthesized

[llvmbot]

Bugzilla Link	2836
Resolution	FIXED
Resolved on	Apr 09, 2009 11:55
Version	unspecified
OS	MacOS X
Attachments	test case
Reporter	LLVM Bugzilla Contributor

Extended Description

The extended setter property is not synthesized by the code generation.
The attached test case is actually taken from the 'Objective-C 2 Programming language' guide.

Thanks,
Thomas

[lattner]

To verify, this is a bug with -emit-llvm, not with -fsyntax-only right?

[llvmbot]

Unsure, I need to investigate how this is getting represented. It may just been looking in a different place for the properties.

[llvmbot]

Yes, this is using -emit-llvm.

The getter is correctly generated:
-(NSString *)language

but the setter is not generated:

• (void)setLanguage:(NSString *)

Therefore when executing code such as:
object.language = @"en"

an exception is thrown by the objc runtime: "selector setLanguage: not recognized".

[llvmbot]

In DeclObjC.cpp:ObjCInterfaceDecl::FindPropertyDeclaration (which is called from SemaDeclObjC.cpp:Sema::ActOnPropertyImplDecl), the code looks through the interface and returns a result without looking through the categories.
I guess this is why my re-declaration is ignored.

Not sure what the correct fix is though.

Here's what the Objective-C 2.0 Programming Language has to say about re-declaration:
"If you declare a property in one class as readonly, you can redeclare it as readwrite in a class extension, a protocol, or a subclass. In the case of a class extension redeclaration, the fact that the property was redeclared prior to any @​synthesize statement will cause the setter to be synthesized."

[llvmbot]

proposed patch
Here is a proposed patch which moves the interface lookup after looking through categories and protocols.
Again, I am not sure this is the correct fix.

[lattner]

Steve, Daniel, can you please look at this patch?

[llvmbot]

Looking at bug/patch now...

[llvmbot]

The proposed patch doesn't look correct. The current lookup order in ObjCInterfaceDecl::FindPropertyDeclaration() is identical to GCC 4.2's lookup_property(). From a front-end perspective, both clang and GCC appear to be doing exactly the same thing (accepting the code without error). Looks like a codegen issue to me...I'll discuss with Daniel when he gets in.

[llvmbot]

I suspect this is a bug in codegen. Will investigate.

[llvmbot]

Fariborz fixed this.