RefCounted::derefBase() should do the decrement from 1 to 0

RefCounted::derefBase() should do the decrement from 1 to 0

Currently derefBase() does not do the decrement from 1 to 0 because an object that has 0 refcount is going to die immediately (there is no need to do the decrement).

However, in the oilpan world, it's possible that an object that has 0 refcount is resurrected, because the object can still be retained through oilpan handles. Thus we need to manage the refcount correctly even when the refcount is going to be 0.

I'm making this change separately from oilpan changes, since this change might regress performance of something. derefBase() is performance-sensitive.

BUG=340522
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=170256

[haraken, 2014-03-28 02:03:16]

PTAL

[tkent, 2014-03-28 02:10:09]

lgtm.

[haraken, 2014-03-28 02:22:51]

The CQ bit was checked by haraken@chromium.org

[commit-bot: I haz the power, 2014-03-28 02:22:57]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/haraken@chromium.org/214453005/1

[commit-bot: I haz the power, 2014-03-28 03:29:15]

Change committed as 170256

[Mads Ager (chromium), 2014-03-28 06:32:22]

LGTM

If this regresses the non-oilpan build we can do the ref counting manually for
ThreadSafeRefCountedGarbageCollected we already do that for
RefCountedGarbageCollected. :-)

[Mads Ager (chromium), 2014-03-28 06:34:02]

On 2014/03/28 06:32:22, Mads Ager (chromium) wrote:
> LGTM
> 
> If this regresses the non-oilpan build we can do the ref counting manually for
> ThreadSafeRefCountedGarbageCollected we already do that for
> RefCountedGarbageCollected. :-)

We might have to anyway because as you write, with oilpan we do not have the
invariant that the ref count can never become non-zero when it reaches zero. I
think that will trip asserts in the current ref counting code so it might be
cleaner that we do it ourselves.

[haraken, 2014-03-28 07:06:17]

On 2014/03/28 06:34:02, Mads Ager (chromium) wrote:
> On 2014/03/28 06:32:22, Mads Ager (chromium) wrote:
> > LGTM
> > 
> > If this regresses the non-oilpan build we can do the ref counting manually
for
> > ThreadSafeRefCountedGarbageCollected we already do that for
> > RefCountedGarbageCollected. :-)
> 
> We might have to anyway because as you write, with oilpan we do not have the
> invariant that the ref count can never become non-zero when it reaches zero. I
> think that will trip asserts in the current ref counting code so it might be
> cleaner that we do it ourselves.

I'm fine with either way: adding #if ENABLE_OILPAN to the existing RefCounted.h
or duplicating RefCounted.h.