Oilpan: fix MediaController crashes after r176358.

Oilpan: fix MediaController crashes after r176358.

MediaController::trace() needs to have the declaration of ExecutionContext
available, so as to determine what NeedsAdjustAndMark<> resolves to for
its (weak) ExecutionContext member.

R=haraken@chromium.org,zerny@chromium.org
BUG=
NOTRY=true

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=176415

[sof, 2014-06-18 13:07:05]

Please take a look.

[haraken, 2014-06-18 13:11:15]

LGTM, thanks for digging into this issue.

However, it looks dangerous that we don't hit a compile error when an unexpected
trace is going to be used.

[zerny-chromium, 2014-06-18 13:11:51]

lgtm. Thanks for catching this! We should consider how to catch when such traits
are resolved incorrectly.

[sof, 2014-06-18 13:15:19]

On 2014/06/18 13:11:51, zerny-chromium wrote:
> lgtm. Thanks for catching this! We should consider how to catch when such
traits
> are resolved incorrectly.

Definitely. FTR, i verified that the other ExecutionContext members do have the
definition in scope.

[sof, 2014-06-18 13:18:46]

The CQ bit was checked by sigbjornf@opera.com

[Mads Ager (chromium), 2014-06-18 13:19:03]

LGTM2

We are testing out ways to check that you have the full definition for tracing
weak members. For member it is not a problem because you need the full
definition to be able to call the trace method. However, you do not call trace
for weak members and therefore it would seem that you do not need the full
declaration. We are experimenting with adding a COMPILE_ASSERT or something that
will require the full T definition for visitor->trace(weak_member_t);

[commit-bot: I haz the power, 2014-06-18 13:19:10]

CQ is trying da patch. Follow status at
 https://chromium-status.appspot.com/cq/sigbjornf@opera.com/344443008/1

[commit-bot: I haz the power, 2014-06-18 13:19:44]

Change committed as 176415

[zerny-chromium, 2014-06-18 13:34:42]

> We are testing out ways to check that you have the full definition for tracing
> weak members. For member it is not a problem because you need the full
> definition to be able to call the trace method. However, you do not call trace
> for weak members and therefore it would seem that you do not need the full
> declaration. We are experimenting with adding a COMPILE_ASSERT or something
that
> will require the full T definition for visitor->trace(weak_member_t);

Compile assert uploaded at: https://codereview.chromium.org/334623006