Bring back ref to element in ClassList and RelList

Bring back ref to element in ClassList and RelList

During https://codereview.chromium.org/1629403003/ the non-Oilpan ref forwarding towards element from ClassList and RelList was unrightfully removed. This CL brings it back.

BUG=

[Yoav Weiss, 2016-02-18 14:59:40]

Description was changed from

==========
Bring back ref to element in ClassList and RelList

BUG=
==========

to

==========
Bring back ref to element in ClassList and RelList

During https://codereview.chromium.org/1629403003/ the non-Oilpan ref forwarding
towards element from ClassList and RelList was unrightfully removed. This CL
brings it back.

BUG=
==========

[Yoav Weiss, 2016-02-18 14:59:40]

yoav@yoav.ws changed reviewers:
+ sigbjornf@opera.com

[sof, 2016-02-18 15:35:42]

https://codereview.chromium.org/1709033002/diff/1/third_party/WebKit/Source/c...
File third_party/WebKit/Source/core/html/ClassList.h (right):

https://codereview.chromium.org/1709033002/diff/1/third_party/WebKit/Source/c...
third_party/WebKit/Source/core/html/ClassList.h:56: virtual void ref() {
m_element->ref(); }
A virtual on a final class doesn't add a lot, did you intend to override here?

I think you need to test & debug this with an enable_oilpan:0 build, as the
initial ref() on a e.g. a ClassList won't happen (RefCountedBase sets an initial
ref count to avoid that call), which looks like what you're relying on here. So
you're ref/deref calls are unbalanced -- something that will assert on shutdown,
I'm guessing.

[sof, 2016-02-18 15:39:16]

sigbjornf@opera.com changed reviewers:
+ philipj@opera.com, tkent@chromium.org

[sof, 2016-02-18 15:39:16]

+ reviewers on the original CL.

[Yoav Weiss, 2016-02-19 08:34:31]

On 2016/02/18 15:35:42, sof wrote:
>
https://codereview.chromium.org/1709033002/diff/1/third_party/WebKit/Source/c...
> File third_party/WebKit/Source/core/html/ClassList.h (right):
> 
>
https://codereview.chromium.org/1709033002/diff/1/third_party/WebKit/Source/c...
> third_party/WebKit/Source/core/html/ClassList.h:56: virtual void ref() {
> m_element->ref(); }
> A virtual on a final class doesn't add a lot, did you intend to override here?
> 
> I think you need to test & debug this with an enable_oilpan:0 build, as the
> initial ref() on a e.g. a ClassList won't happen (RefCountedBase sets an
initial
> ref count to avoid that call), which looks like what you're relying on here.
So
> you're ref/deref calls are unbalanced -- something that will assert on
shutdown,
> I'm guessing.

I'm having trouble debugging the non-oilpan build as it keeps crashing on me
(unrelated to these changes. crashes on master)

[Yoav Weiss, 2016-02-19 08:41:50]

On 2016/02/19 08:34:31, Yoav Weiss wrote:
> On 2016/02/18 15:35:42, sof wrote:
> >
>
https://codereview.chromium.org/1709033002/diff/1/third_party/WebKit/Source/c...
> > File third_party/WebKit/Source/core/html/ClassList.h (right):
> > 
> >
>
https://codereview.chromium.org/1709033002/diff/1/third_party/WebKit/Source/c...
> > third_party/WebKit/Source/core/html/ClassList.h:56: virtual void ref() {
> > m_element->ref(); }
> > A virtual on a final class doesn't add a lot, did you intend to override
here?
> > 
> > I think you need to test & debug this with an enable_oilpan:0 build, as the
> > initial ref() on a e.g. a ClassList won't happen (RefCountedBase sets an
> initial
> > ref count to avoid that call), which looks like what you're relying on here.
> So
> > you're ref/deref calls are unbalanced -- something that will assert on
> shutdown,
> > I'm guessing.
> 
> I'm having trouble debugging the non-oilpan build as it keeps crashing on me
> (unrelated to these changes. crashes on master)

I'll try my linux setup, synced to ToT and see what it gives me.

[sof, 2016-02-19 09:49:12]

On 2016/02/19 08:41:50, Yoav Weiss wrote:
> On 2016/02/19 08:34:31, Yoav Weiss wrote:
> > On 2016/02/18 15:35:42, sof wrote:
> > >
> >
>
https://codereview.chromium.org/1709033002/diff/1/third_party/WebKit/Source/c...
> > > File third_party/WebKit/Source/core/html/ClassList.h (right):
> > > 
> > >
> >
>
https://codereview.chromium.org/1709033002/diff/1/third_party/WebKit/Source/c...
> > > third_party/WebKit/Source/core/html/ClassList.h:56: virtual void ref() {
> > > m_element->ref(); }
> > > A virtual on a final class doesn't add a lot, did you intend to override
> here?
> > > 
> > > I think you need to test & debug this with an enable_oilpan:0 build, as
the
> > > initial ref() on a e.g. a ClassList won't happen (RefCountedBase sets an
> > initial
> > > ref count to avoid that call), which looks like what you're relying on
here.
> > So
> > > you're ref/deref calls are unbalanced -- something that will assert on
> > shutdown,
> > > I'm guessing.
> > 
> > I'm having trouble debugging the non-oilpan build as it keeps crashing on me
> > (unrelated to these changes. crashes on master)
> 
> I'll try my linux setup, synced to ToT and see what it gives me.

fwiw, https://codereview.chromium.org/1715653002/ takes care of it for me.