Peek at the event queue for wheel events for gtk

Peek at the event queue for wheel events for cros
BUG=NONE
Test=RenderWidgetHostTest.CoalescesWheelEvents (existing)

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=49938

[darin (slow to review), 2010-06-08 18:19:27]

Given this peek-based solution, I think it would be good to remove the previous
PostTask-based solution.  I suspect it would also be good to use this peek-based
solution on other ports that use XWindows.

http://codereview.chromium.org/2679003/diff/2001/3001
File chrome/browser/renderer_host/render_widget_host_view_gtk.cc (right):

http://codereview.chromium.org/2679003/diff/2001/3001#newcode339
chrome/browser/renderer_host/render_widget_host_view_gtk.cc:339: static int
GetScrollEvents(guint current_event_state) {
nit: PopScrollEvents or PopSimilarScrollEvents might be a better name.

http://codereview.chromium.org/2679003/diff/2001/3001#newcode345
chrome/browser/renderer_host/render_widget_host_view_gtk.cc:345: if
(scroll.direction == GDK_SCROLL_UP ||
you don't want to optimize horizontal scrolling in a similar way?

http://codereview.chromium.org/2679003/diff/2001/3001#newcode353
chrome/browser/renderer_host/render_widget_host_view_gtk.cc:353: }
if the event is not a GDK_SCROLL event, then don't you need to free it using
gdk_event_free?  maybe you meant to have a break here?  or maybe the
gdk_event_put should be here?

[Nico, 2010-06-10 18:31:36]

I agree that this is cleaner than the other approach. Can you investigate if
this is reasonable / possible / necessary on osx/win as well?

http://codereview.chromium.org/2679003/diff/2001/3001
File chrome/browser/renderer_host/render_widget_host_view_gtk.cc (right):

http://codereview.chromium.org/2679003/diff/2001/3001#newcode388
chrome/browser/renderer_host/render_widget_host_view_gtk.cc:388: num_clicks +=
GetScrollEvents(event->state);
could this function return a delta instead of a count? then it could be called
in the cross-platform portion and implemented on the other ports too (i think).

[DaveMoore, 2010-06-11 01:51:27]

http://codereview.chromium.org/2679003/diff/2001/3001
File chrome/browser/renderer_host/render_widget_host_view_gtk.cc (right):

http://codereview.chromium.org/2679003/diff/2001/3001#newcode339
chrome/browser/renderer_host/render_widget_host_view_gtk.cc:339: static int
GetScrollEvents(guint current_event_state) {
On 2010/06/08 18:19:30, darin wrote:
> nit: PopScrollEvents or PopSimilarScrollEvents might be a better name.
Changed to GetPendingScrollDelta()

http://codereview.chromium.org/2679003/diff/2001/3001#newcode345
chrome/browser/renderer_host/render_widget_host_view_gtk.cc:345: if
(scroll.direction == GDK_SCROLL_UP ||
On 2010/06/08 18:19:30, darin wrote:
> you don't want to optimize horizontal scrolling in a similar way?

Done.

http://codereview.chromium.org/2679003/diff/2001/3001#newcode353
chrome/browser/renderer_host/render_widget_host_view_gtk.cc:353: }
This was messed up. I've changed the loop logic a bit to ensure we either
coalesce the event or drop out of the loop. In either case we free the event.
On 2010/06/08 18:19:30, darin wrote:
> if the event is not a GDK_SCROLL event, then don't you need to free it using
> gdk_event_free?  maybe you meant to have a break here?  or maybe the
> gdk_event_put should be here?

http://codereview.chromium.org/2679003/diff/2001/3001#newcode388
chrome/browser/renderer_host/render_widget_host_view_gtk.cc:388: num_clicks +=
GetScrollEvents(event->state);
I've changed it to return a delta. It's still all gdk specific and only called
from gdk specific code but it'll be easier to refactor if we decide to do it on
mac / windows.
On 2010/06/10 18:31:36, Nico wrote:
> could this function return a delta instead of a count? then it could be called
> in the cross-platform portion and implemented on the other ports too (i
think).

[darin (slow to review), 2010-06-14 21:17:48]

LGTM