[webkit-dev] Protecting against stale pointers in DrawingBuffer and GraphicsContext3D
mjs at apple.com
Tue Oct 12 15:46:06 PDT 2010
On Oct 12, 2010, at 2:37 PM, Darin Fisher wrote:
> On Tue, Oct 12, 2010 at 1:20 PM, Maciej Stachowiak <mjs at apple.com> wrote:
> Hmm, I've found weak pointer abstractions to be very useful. The issue with reference counting is that it is "easy" to introduce memory leaks, and has been mentioned, it is sometimes nice to have deterministic object destruction.
We used to have lots of problems with leaking refcounted objects back when most refcounting was done manually, but this has become much less common as we deployed smart pointers.
> It is also nice to avoid having to have explicit clear() functions and then add checks to every other method to assert that they are not called after clear().
Well, if you have a weak pointer you have to check weak pointer validity, so you don't save any checks. I don't think explicit close calls are that bad. I think it's actually a better pattern when you hold significant resources other than just some memory (e.g. VRAM, file handles, sockets, limited kernel resources, window server handles...). That way, cleaning up your external resources does not become dependent on your ownership strategy for the object.
> In the Chromium code base, we have a helper for weak pointers:
> We tend to use this in cases in which:
> 1) there are many consumers interested in holding a back pointer to some shared resource, and
> 2) we'd like the shared resource to die at some predictable time.
> Without a utility like this, the alternative is to make the shared resource notify each of the consumers about the impending destruction of the shared resource.
> It is true that WeakPtr<T> adds a null check in front of each method call made by the consumers, but that runtime cost is often justified in exchange for reduced code complexity (i.e., eliminating the need to notify consumers when the shared resource dies).
The cost isn't just the null checks. You need either storage for all the backpointers, or the smaller storage overhead for a refcounted handle. But then almost by definition you have more overhead than refcounting, since you have a refcounted object plus more work.
I do think weak pointers have their uses, but only in relatively unusual cases (e.g. when you'd have a cycle otherwise). It doesn't sound like a huge win in this case.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the webkit-dev