More interactive test porting for Linux.

More interactive test porting for Linux.

Also fix a bug in the Windows implementation of SendMouseMoveNotifyWhenDone where the task would never be run if the cursor was already in the destination position before the call.

BUG=19076
BUG=19881
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=24027

[Evan Stade, 2009-08-21 20:49:03]

tested and passing on windows + linux

[tony, 2009-08-21 21:13:39]

http://codereview.chromium.org/174201/diff/14/21
File chrome/common/gtk_util.cc (right):

http://codereview.chromium.org/174201/diff/14/21#newcode152
Line 152: gdk_window_get_origin(widget->window, &x, &y);
x & y still get filled in if there's no window? I thought we depend on the
parent traversal for getting the position of hidden widgets or something (e.g.,
do dragged tabs still look ok?)

[Evan Stade, 2009-08-21 21:43:42]

http://codereview.chromium.org/174201/diff/14/21
File chrome/common/gtk_util.cc (right):

http://codereview.chromium.org/174201/diff/14/21#newcode152
Line 152: gdk_window_get_origin(widget->window, &x, &y);
On 2009/08/21 21:13:39, tony wrote:
> x & y still get filled in if there's no window? I thought we depend on the
> parent traversal for getting the position of hidden widgets or something
(e.g.,
> do dragged tabs still look ok?)

uh, there should always be a window, right? If the widget isn't realized then I
don't think getting the position makes sense. Walking the widget's ancestors is
unnecessary because  the allocation x/y coords are GdkWindow-relative for
no-window widgets (whereas they are parent-widget relative for windowed
widgets).

I've tested tab dragging with and without this patch, and can't notice a
difference. (Although it is noteworthy that in both cases, there are glaring
graphical and functional glitches.)

[tony, 2009-08-21 21:48:51]

http://codereview.chromium.org/174201/diff/14/21
File chrome/common/gtk_util.cc (right):

http://codereview.chromium.org/174201/diff/14/21#newcode152
Line 152: gdk_window_get_origin(widget->window, &x, &y);
If there's always a gdk window, can we add a DCHECK as such?  With that, LGTM.