Implement surface-based browser process hit testing for Mac and Aura

Implement surface-based browser process hit testing for Mac and Aura

For efficient input handling with out-of-process iframes, events need to be routed directly to the correct renderer process that will end up handling the event (rather than forwarding from renderer process to renderer process, as we have been doing previously). This patch causes RenderWidgetHostViewAura and RenderWidgetHostViewMac to send mouse and mousewheel events to a routing class that performs a hit test base on the event's coordinates and forwards to the correct RenderWidgetHostView.

This should only have an affect when running in Site-per-process mode. Otherwise the hit test gets short circuited because there is only one RenderWidgetHostView per tab.

BUG=491334
Committed: https://crrev.com/2a565f883aa938587ed745a077e1dd734c389c55
Cr-Commit-Position: refs/heads/master@{#347008}

[kenrb, 2015-08-31 19:00:20]

kenrb@chromium.org changed reviewers:
+ piman@chromium.org

[kenrb, 2015-08-31 19:02:23]

piman: PTAL?

There still seems to be a problem with this working on Mac, with the coordinates
being skewed when passed to an OOPIF, but I am investigating if it might
actually be due to a known scale factor bug
(https://code.google.com/p/chromium/issues/detail?id=522237).

[kenrb, 2015-08-31 19:02:26]

piman: PTAL?

There still seems to be a problem with this working on Mac, with the coordinates
being skewed when passed to an OOPIF, but I am investigating if it might
actually be due to a known scale factor bug
(https://code.google.com/p/chromium/issues/detail?id=522237).

[piman, 2015-08-31 23:45:14]

lgtm

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
File content/browser/renderer_host/render_widget_host_input_event_router.h
(right):

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
content/browser/renderer_host/render_widget_host_input_event_router.h:50:
typedef std::map<uint32_t, RenderWidgetHostViewBase*>
nit: hash_map?

[dcheng, 2015-08-31 23:45:54]

Some random drive-bys.

https://codereview.chromium.org/1255483004/diff/240001/content/browser/compos...
File content/browser/compositor/delegated_frame_host.h (right):

https://codereview.chromium.org/1255483004/diff/240001/content/browser/compos...
content/browser/compositor/delegated_frame_host.h:156: // Returns a null
SurfaceId if a surface has not yet been created.
which surface? The one associated with this DelegatedFrameHost

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
File content/browser/renderer_host/render_widget_host_input_event_router.cc
(right):

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
content/browser/renderer_host/render_widget_host_input_event_router.cc:75: const
SurfaceIdNamespaceOwnerMap::iterator iter = owner_map_.find(id);
owner_map_.erase(id) is a shorter way to write this.

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
File content/browser/renderer_host/render_widget_host_input_event_router.h
(right):

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
content/browser/renderer_host/render_widget_host_input_event_router.h:35:
virtual ~RenderWidgetHostInputEventRouter();
Why virtual? I don't see any overrides.

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
content/browser/renderer_host/render_widget_host_input_event_router.h:53: };
DISALLOW_COPY_AND_ASSIGN?

[kenrb, 2015-09-01 15:56:16]

https://codereview.chromium.org/1255483004/diff/240001/content/browser/compos...
File content/browser/compositor/delegated_frame_host.h (right):

https://codereview.chromium.org/1255483004/diff/240001/content/browser/compos...
content/browser/compositor/delegated_frame_host.h:156: // Returns a null
SurfaceId if a surface has not yet been created.
On 2015/08/31 23:45:54, dcheng wrote:
> which surface? The one associated with this DelegatedFrameHost

Yes, clarified.

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
File content/browser/renderer_host/render_widget_host_input_event_router.cc
(right):

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
content/browser/renderer_host/render_widget_host_input_event_router.cc:75: const
SurfaceIdNamespaceOwnerMap::iterator iter = owner_map_.find(id);
On 2015/08/31 23:45:54, dcheng wrote:
> owner_map_.erase(id) is a shorter way to write this.

Done.

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
File content/browser/renderer_host/render_widget_host_input_event_router.h
(right):

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
content/browser/renderer_host/render_widget_host_input_event_router.h:35:
virtual ~RenderWidgetHostInputEventRouter();
On 2015/08/31 23:45:54, dcheng wrote:
> Why virtual? I don't see any overrides.

I thought I might need it for unit tests, but didn't end up writing any so it is
now removed.

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
content/browser/renderer_host/render_widget_host_input_event_router.h:50:
typedef std::map<uint32_t, RenderWidgetHostViewBase*>
On 2015/08/31 23:45:14, piman (OOO back 2015-08-31) wrote:
> nit: hash_map?

Done.

https://codereview.chromium.org/1255483004/diff/240001/content/browser/render...
content/browser/renderer_host/render_widget_host_input_event_router.h:53: };
On 2015/08/31 23:45:54, dcheng wrote:
> DISALLOW_COPY_AND_ASSIGN?

Done.

[lazyboy, 2015-09-01 21:29:43]

https://chromiumcodereview.appspot.com/1255483004/diff/280001/content/browser...
File content/browser/renderer_host/render_widget_host_view_mac.mm (right):

https://chromiumcodereview.appspot.com/1255483004/diff/280001/content/browser...
content/browser/renderer_host/render_widget_host_view_mac.mm:1158:
render_widget_host_->delegate()->GetInputEventRouter()) {
FYI, the mac failures seems to be b/c render_widget_host_->delegate() is NULL

[kenrb, 2015-09-02 16:35:06]

https://codereview.chromium.org/1255483004/diff/280001/content/browser/render...
File content/browser/renderer_host/render_widget_host_view_mac.mm (right):

https://codereview.chromium.org/1255483004/diff/280001/content/browser/render...
content/browser/renderer_host/render_widget_host_view_mac.mm:1158:
render_widget_host_->delegate()->GetInputEventRouter()) {
On 2015/09/01 21:29:42, lazyboy wrote:
> FYI, the mac failures seems to be b/c render_widget_host_->delegate() is NULL

Thanks for helping with debugging this! I didn't realize that WebContents can be
destroyed before RenderWidgetHosts, and have added null checks to cover that.

[kenrb, 2015-09-02 19:19:18]

The CQ bit was checked by kenrb@chromium.org

[kenrb, 2015-09-02 19:19:19]

The patchset sent to the CQ was uploaded after l-g-t-m from piman@chromium.org
Link to the patchset: https://codereview.chromium.org/1255483004/#ps320001
(title: "Disabled test on Android (not implemented there)")

[commit-bot: I haz the power, 2015-09-02 19:19:50]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1255483004/320001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1255483004/320001

[commit-bot: I haz the power, 2015-09-02 20:25:07]

Committed patchset #17 (id:320001)

[commit-bot: I haz the power, 2015-09-02 20:25:49]

Patchset 17 (id:??) landed as
https://crrev.com/2a565f883aa938587ed745a077e1dd734c389c55
Cr-Commit-Position: refs/heads/master@{#347008}