[PATCH] Cleaning up ownership attribute semantics

[Aaron Ballman]

It does have to be an identifier, because if it's a string literal,
"isArgIdent" will return false.  Do we have existing code that is
using string literals?

~Aaron

On Mon, Sep 16, 2013 at 1:46 PM, Anna Zaks <ganna at apple.com> wrote:
> -      << AL.getName()->getName() << 1 << AANT_ArgumentString;
> +      << AL.getName() << 1 << AANT_ArgumentIdentifier;
>
> I do think the first argument does not have to be an identifier. (I might have been a bit vague in one of the previous emails - I was describing a hypothetical case.)
>
> Otherwise, looks good tome.
>
> Anna.
> On Sep 16, 2013, at 10:38 AM, Aaron Ballman <aaron at aaronballman.com> wrote:
>
>> I noticed that the ownership attributes have no tests in sema, and
>> that some of the functionality was not diagnosing errors as expected.
>> This patch refactors ownership attribute semantic handling, and adds
>> test cases for the functionality so future refactorings do not
>> accidentally regress the behavior further.
>>
>> ~Aaron
>> <Ownership.patch>
>