834659 - Consider dropping refcounting in AsyncPanZoomController and maybe other layers/ipc classes

Status: RESOLVED INVALID

(Firefox OS Graveyard :: General, defect)

Description

[Benoit Jacob [:bjacob] (mostly away)]

The lifetime of AsyncPanZoomController doesn't seem to be controlled by refcounting (all the more since https://hg.mozilla.org/integration/mozilla-inbound/rev/b7c845b323d0) yet AsyncPanZoomController is refcounted.

Having unused refcounting seems dangerous, because if now someone arrived here, saw APZC being refcounted, and made code holding a nsRefPtr<APZC>, then the APZC would get destroyed as soon as that nsRefPtr<APZC> would go away (because nothing else addref'd it), which would be surprising, right?

Comment 1

[(no longer active)]

Attachment: Patch (v1)

Comment 2

[(no longer active)]

Comment on attachment 706417 [details] [diff] [review]
Patch (v1)

Nevermind... <http://mxr.mozilla.org/mozilla-central/source/gfx/layers/ipc/CompositorParent.cpp#1247> and friends seem to rely on this being refcounted.

Comment 3

[Benoit Jacob [:bjacob] (mostly away)]

Sorry about that. Shall we RESOLVE INVALID?

Comment 4

[(no longer active)]

(In reply to comment #3)
> Sorry about that. Shall we RESOLVE INVALID?

Well, if we're sure that the lifetime of this object is not managed manually anywhere, sure.

Comment 5

[Benoit Jacob [:bjacob] (mostly away)]

It looks complicated. If I understand correctly, the lifetime is managed by refcount, but the refcount is tweaked manually. So I think that this bug is INVALID.