Initialize GC mixin bases only when leftmost vtable has been initialized.

Initialize GC mixin bases only when leftmost vtable has been initialized.

For GC mixin constructors that allocate on the Oilpan heap, extra care
is required. It is unsafe to fully construct these until the vptr for the
leftmost object that implement/provide the mixin has been initialized.

Should a (conservative) GC kick in while in that mixin's constructor, the
object reference to the object instance will be located through stack
scanning, but it will not as-yet have its trace method set up. The result
is failure to trace the object being constructed (including its mixin
base).

R=haraken
BUG=420515
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=188695

[haraken, 2015-01-20 15:40:31]

haraken@chromium.org changed reviewers:
+ haraken@chromium.org, oilpan-reviews@chromium.org

[haraken, 2015-01-20 15:40:32]

Can we have a FIXME on those initialize methods to explain why the initialize()
is needed?

LGTM.

https://codereview.chromium.org/860063002/diff/1/Source/core/svg/SVGTests.cpp
File Source/core/svg/SVGTests.cpp (right):

https://codereview.chromium.org/860063002/diff/1/Source/core/svg/SVGTests.cpp...
Source/core/svg/SVGTests.cpp:41: ASSERT(!m_systemLanguage.get());

.get() won't be needed.

https://codereview.chromium.org/860063002/diff/1/Source/core/svg/SVGURIRefere...
File Source/core/svg/SVGURIReference.cpp (right):

https://codereview.chromium.org/860063002/diff/1/Source/core/svg/SVGURIRefere...
Source/core/svg/SVGURIReference.cpp:36: ASSERT(!m_href.get());

.get() won't be needed.

[sof, 2015-01-20 19:02:50]

haraken wrote:
> Can we have a FIXME on those initialize methods to explain why the
initialize()
> is needed?

Added explanatory comments.

https://codereview.chromium.org/860063002/diff/1/Source/core/svg/SVGTests.cpp
File Source/core/svg/SVGTests.cpp (right):

https://codereview.chromium.org/860063002/diff/1/Source/core/svg/SVGTests.cpp...
Source/core/svg/SVGTests.cpp:41: ASSERT(!m_systemLanguage.get());
On 2015/01/20 15:40:32, haraken wrote:
> 
> .get() won't be needed.

Done, dropped them from the asserts here.

[sof, 2015-01-20 19:02:56]

The CQ bit was checked by sigbjornf@opera.com

[commit-bot: I haz the power, 2015-01-20 19:03:55]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/860063002/20001

[commit-bot: I haz the power, 2015-01-20 21:09:23]

Committed patchset #2 (id:20001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=188695