Yea, I introduced this. Will fix. Thanks.

I'm also seeing this when looking into views_mus_unittests

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/462ef8c3894d3dc1eceb4a2327006b372737f749

commit 462ef8c3894d3dc1eceb4a2327006b372737f749
Author: fsamuel <fsamuel@chromium.org>
Date: Fri Mar 03 19:12:11 2017

Decouple FrameSink Hierarchy registration from SurfaceFactoryClient

Prior to this patch, FrameSink hierarchy registration and
SurfaceFactoryClient registration were tightly coupled in
SurfaceManager.

In SurfaceManager's destructor, it verified that the
frame_sink_source_map_ is empty. This proved problematic in Mus+Ash
because the Mus window server tries to unregister the framesink
hierarchy at shutdown by issuing IPCs to Mus-GPU. Thus, there is
no guarantee these IPCs will arrive in the gpu process prior to
SurfaceManager destruction.

By design, Mojo provides no guarantees on process shutdown order, and
even if a process tries to override the behavior, Mojo may someday
decide to force quit processes and so we cannot guarantee Mus-WS will
outlive Mus-GPU.

Currently, in Mus+Ash, the BeginFrame hierarchy can be much larger
than the set of FrameSinkIds with an associated SurfaceFactoryClient
(an associated CompositorFrameSink), in order to facilitate reparenting
windows in an environment where not all windows submit
CompositorFrames (a single client may embed only a single
CompositorFrame that embeds a bunch of local window concepts for
performance reasons). It seems to make sense to decouple the BeginFrame
hierarchy from the SurfaceFactoryClient hierarchy, in that case.

Furthermore, the DCHECK in the SurfaceManager destructor exists to
verify that there are no lingering clients and we haven't been
wastefully ticking BeginFrames. By decoupling client registration and
BeginFrame registration, we can continue to guard against that
potential bug while allowing Mus+Ash to have a clean shutdown.

BUG= 697116 
CQ_INCLUDE_TRYBOTS=master.tryserver.blink:linux_trusty_blink_rel

Review-Url: https://codereview.chromium.org/2731743002
Cr-Commit-Position: refs/heads/master@{#454641}

[modify] https://crrev.com/462ef8c3894d3dc1eceb4a2327006b372737f749/cc/surfaces/surface_manager.cc
[modify] https://crrev.com/462ef8c3894d3dc1eceb4a2327006b372737f749/cc/surfaces/surface_manager.h
[modify] https://crrev.com/462ef8c3894d3dc1eceb4a2327006b372737f749/cc/surfaces/surface_manager_unittest.cc