aura-x11: Fix touch-calibration for multi-monitor setups.

aura-x11: Fix touch-calibration for multi-monitor setups.

The fix is:
 * Remember the size of the X root window.
 * Listen for touch events in the X root window.
 * Convert the touch-event coordinates from the X root-window to the aura root-window coordinate.
 * If the event belongs to the aura root-window after the calibration, then process the event, otherwise discard.

BUG=133721
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=155931

[Daniel Erat, 2012-09-10 17:27:31]

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
File ui/aura/root_window_host_linux.cc (right):

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
ui/aura/root_window_host_linux.cc:54: const int kXRootWindowPaddingLeft = 40;
nit: add a comment documenting what these are, and where the numbers come from

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
ui/aura/root_window_host_linux.cc:570: XGetWindowAttributes(xdisplay_,
x_root_window_, &attrs);
i was going to suggest XGetGeometry() instead, but it'd probably end up being
more lines of code because of all the stupid temporary variables you'd need to
pass in :-(

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
ui/aura/root_window_host_linux.cc:585: XSelectInput(xdisplay_, x_root_window_,
StructureNotifyMask);
do this before you query the bounds; otherwise there's a race

can you just move this into SelectEventsForRootWindow()?  it certainly fits the
description.

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
File ui/aura/root_window_host_linux.h (right):

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
ui/aura/root_window_host_linux.h:94: gfx::Rect root_bounds_;
nit: "x_root_bounds_", or something including the word "host"?  "root" is a bit
ambiguous within this class

http://codereview.chromium.org/10905163/diff/1007/ui/base/events/event.cc
File ui/base/events/event.cc (right):

http://codereview.chromium.org/10905163/diff/1007/ui/base/events/event.cc#new...
ui/base/events/event.cc:295: int calibrated_x = x() * to.width() / from.width();
nit: cast to floating-point and round?

[sadrul, 2012-09-10 18:19:32]

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
File ui/aura/root_window_host_linux.cc (right):

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
ui/aura/root_window_host_linux.cc:54: const int kXRootWindowPaddingLeft = 40;
On 2012/09/10 17:27:31, Daniel Erat wrote:
> nit: add a comment documenting what these are, and where the numbers come from

Done (to make sure this gets fixed properly, I have filed a bug as well)

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
ui/aura/root_window_host_linux.cc:570: XGetWindowAttributes(xdisplay_,
x_root_window_, &attrs);
On 2012/09/10 17:27:31, Daniel Erat wrote:
> i was going to suggest XGetGeometry() instead, but it'd probably end up being
> more lines of code because of all the stupid temporary variables you'd need to
> pass in :-(

If XGetGeometry is less expensive than XGetWindowAttributes, then I am happy to
make this change.

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
ui/aura/root_window_host_linux.cc:585: XSelectInput(xdisplay_, x_root_window_,
StructureNotifyMask);
On 2012/09/10 17:27:31, Daniel Erat wrote:
> do this before you query the bounds; otherwise there's a race
> 
> can you just move this into SelectEventsForRootWindow()?  it certainly fits
the
> description.

Done.

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
File ui/aura/root_window_host_linux.h (right):

http://codereview.chromium.org/10905163/diff/1007/ui/aura/root_window_host_li...
ui/aura/root_window_host_linux.h:94: gfx::Rect root_bounds_;
On 2012/09/10 17:27:31, Daniel Erat wrote:
> nit: "x_root_bounds_", or something including the word "host"?  "root" is a
bit
> ambiguous within this class

Done (changed to |x_root_bounds_|)

http://codereview.chromium.org/10905163/diff/1007/ui/base/events/event.cc
File ui/base/events/event.cc (right):

http://codereview.chromium.org/10905163/diff/1007/ui/base/events/event.cc#new...
ui/base/events/event.cc:295: int calibrated_x = x() * to.width() / from.width();
On 2012/09/10 17:27:31, Daniel Erat wrote:
> nit: cast to floating-point and round?

Done.

[Daniel Erat, 2012-09-10 18:43:57]

lgtm

[sky, 2012-09-10 21:13:30]

LGTM