Don't send mouse move related events to Cocoa plugins in background windows

Don't send mouse move related events to Cocoa plugins in background windows


BUG=111275
TEST=Open a new window while hovering over a Flash instance that hides the cursor; the cursor should re-appear as soon as the mouse moves.


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=134691

[stuartmorgan, 2012-04-20 10:59:22]

avi: for review
ananta: for owners

See the bug for background; ideally I'd like to see the weird events not being
sent to the plugin host in the first place, but this is a pretty bad bug and the
workaround is trivial.

[Avi (use Gerrit), 2012-04-20 14:44:01]

LGTM with nits; do we know where to look in WebKit to track down the weird
events?

https://chromiumcodereview.appspot.com/10169003/diff/1/webkit/plugins/npapi/w...
File webkit/plugins/npapi/webplugin_delegate_impl_mac.mm (right):

https://chromiumcodereview.appspot.com/10169003/diff/1/webkit/plugins/npapi/w...
webkit/plugins/npapi/webplugin_delegate_impl_mac.mm:524: // have focus; Cocoa
event model plugins don't expect to recieve mouse move
s/recieve/receive/

https://chromiumcodereview.appspot.com/10169003/diff/1/webkit/plugins/npapi/w...
webkit/plugins/npapi/webplugin_delegate_impl_mac.mm:527:
instance()->event_model() == NPEventModelCocoa &&
Only the Cocoa event model?

[stuartmorgan, 2012-04-20 15:14:57]

I could trace them if I had time. It's just a lot of tedious work setting the
right breakpoints in the renderer process where the plugin proxy gets the event,
figuring out the ordering of events with the window focus, looking at stacks,
etc. I don't have the time to dig into it, but someone else could down the road
and this could eventually be removed.

https://chromiumcodereview.appspot.com/10169003/diff/1/webkit/plugins/npapi/w...
File webkit/plugins/npapi/webplugin_delegate_impl_mac.mm (right):

https://chromiumcodereview.appspot.com/10169003/diff/1/webkit/plugins/npapi/w...
webkit/plugins/npapi/webplugin_delegate_impl_mac.mm:524: // have focus; Cocoa
event model plugins don't expect to recieve mouse move
On 2012/04/20 14:44:01, Avi wrote:
> s/recieve/receive/

Done.

https://chromiumcodereview.appspot.com/10169003/diff/1/webkit/plugins/npapi/w...
webkit/plugins/npapi/webplugin_delegate_impl_mac.mm:527:
instance()->event_model() == NPEventModelCocoa &&
On 2012/04/20 14:44:01, Avi wrote:
> Only the Cocoa event model?

The Carbon event model is a cesspool. There are no formal mouse move events,
just the constant idle callbacks which include the current mouse location. Those
fire regardless of foreground/background state. Enter and Leave don't exist at
all, there's just a "hey, you might want to check your cursor state now" event.

[stuartmorgan, 2012-04-25 08:59:39]

ananta: ping for owners sign-off

[stuartmorgan, 2012-04-30 10:25:27]

On 2012/04/25 08:59:39, stuartmorgan wrote:
> ananta: ping for owners sign-off

Ping again. This is something Adobe has flagged as a significant issue, so we
really need to get this into the next branch.

[ananta, 2012-04-30 17:14:07]

LGTM

[commit-bot: I haz the power, 2012-05-01 06:45:57]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/stuartmorgan@chromium.org/10169003/4001

[commit-bot: I haz the power, 2012-05-01 13:42:00]

Change committed as 134691