[cfe-commits] [PATCH] Add -fobjc-direct-class-refs

[Jonathan Schleifer]

Am 19.07.2012 um 21:24 schrieb John McCall:

> Could you explain why you're testing hasSubscripting in one place and isSubscriptPointerArithmetic in another?

Yes, of course:

In SemaExpr, it checks if the subscript is pointer arithmetic. In SemaExprObjC, it checks if subscripting is allowed with the runtime. These are, IMHO, two different things: A runtime could disallow pointer arithmetic _AND_ subscripting. Imagine you target OS X 10.5 x86_64: You have the fragile ABI and thus pointer arithmetic is disallowed. However, subscripting is not supported by the runtime (well, actually the framework).

This is the reason why I decided it makes sense to use these two different runtime properties. They are, currently, the same, but you could change it in the way I just described later. I just wanted to be future-proof, that's all :). If you dislike it, feel free to reduce it to hasSubscripting.

Looking forward to get that integrated, already have two other patches ready :).

PS: I removed the two Davids from CC, as I think they are both subscribed and have no specific interest in this.

--
Jonathan