gtk: Hide the status bubble when the mouse nears it.

gtk: Hide the status bubble when the mouse nears it.

This isn't as slick as other platforms, in that the bubble
just slides down and gets cropped instead of sliding down out
of the browser window -- doing the latter will probably
require reparenting the bubble into its own window when we
want to move it.

BUG=18311
TEST=ran "nc -l -p 8080", went to localhost:8080, and made sure that the bubble ran away with both LTR and RTL languages

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=31984

[Daniel Erat, 2009-11-12 02:25:21]

This still needs testing on other platforms, but if you don't mind taking an
early look to let me know if it seems reasonable...

http://codereview.chromium.org/392007/diff/1001/1006
File chrome/browser/gtk/status_bubble_gtk.cc (right):

http://codereview.chromium.org/392007/diff/1001/1006#newcode173
Line 173: gtk_widget_queue_resize_no_redraw(parent);
There's probably some room for improvement here in avoiding this when the status
bubble's not visible -- I think we'd still need to recompute y_offset_ in
response to the mouse motion events, but we can defer the resize until the
bubble becomes visible.  Let me know if you think that it's worth doing this.

[tony, 2009-11-12 20:01:22]

http://codereview.chromium.org/392007/diff/1001/1002
File chrome/browser/browser.cc (right):

http://codereview.chromium.org/392007/diff/1001/1002#newcode2028
Line 2028: GetStatusBubble()->MouseMoved(location, !motion);
Can we get by with not passing motion through?  You can reset the position when
the status bubble is cleared.  Alternately, should passing MouseMoved handle
clearing the status bubble on all platforms?

http://codereview.chromium.org/392007/diff/1001/1003
File chrome/browser/browser.h (right):

http://codereview.chromium.org/392007/diff/1001/1003#newcode13
Line 13: #include "base/gfx/point.h"
Nit: You can forward declare this.

http://codereview.chromium.org/392007/diff/1001/1005
File chrome/browser/cocoa/status_bubble_mac.mm (right):

http://codereview.chromium.org/392007/diff/1001/1005#newcode218
Line 218: // TODO: Use 'location' here instead of NSEvent.
Nit: Give this TODO an owner.

http://codereview.chromium.org/392007/diff/1001/1006
File chrome/browser/gtk/status_bubble_gtk.cc (right):

http://codereview.chromium.org/392007/diff/1001/1006#newcode173
Line 173: gtk_widget_queue_resize_no_redraw(parent);
On 2009/11/12 02:25:22, Daniel Erat wrote:
> There's probably some room for improvement here in avoiding this when the
status
> bubble's not visible -- I think we'd still need to recompute y_offset_ in
> response to the mouse motion events, but we can defer the resize until the
> bubble becomes visible.  Let me know if you think that it's worth doing this.

This seems fine for now.  We can optimize if it's slow.

http://codereview.chromium.org/392007/diff/1001/1009
File chrome/browser/status_bubble.h (right):

http://codereview.chromium.org/392007/diff/1001/1009#newcode12
Line 12: class Point;
Nit: No indent.

http://codereview.chromium.org/392007/diff/1001/1014
File chrome/browser/views/status_bubble_views.h (right):

http://codereview.chromium.org/392007/diff/1001/1014#newcode8
Line 8: #include "base/gfx/point.h"
Nit: forward declare

[Nico, 2009-11-12 20:04:51]

FWIW, on win and os x the bubble moves to the bottom right corner when the mouse
gets near and there's no room below the window to slide down.

[use derat at chromium.org, 2009-11-12 23:35:10]

Yeah, this is just a quick first attempt.  The status bubble gets
clipped instead of sliding out of the window in this change, so we
might want to always move it to the bottom right instead of sliding it
down.

On Thu, Nov 12, 2009 at 12:04 PM,  <thakis@chromium.org> wrote:
> FWIW, on win and os x the bubble moves to the bottom right corner when the
> mouse
> gets near and there's no room below the window to slide down.
>
> http://codereview.chromium.org/392007
>

[Daniel Erat, 2009-11-13 22:13:59]

Another look?  I ended up fixing the behavior when the download shelf is visible
since it looked funny otherwise (the bubble was getting drawn on top of the
shelf when it slid down).

http://codereview.chromium.org/392007/diff/4024/2013
File chrome/browser/gtk/status_bubble_gtk.cc (right):

http://codereview.chromium.org/392007/diff/4024/2013#newcode159
Line 159: (ltr ? 0 : parent->allocation.width - requisition.width),
I'm not thrilled about repeating the container's calculations here but can't
think of a cleaner way to do this.

[tony, 2009-11-13 22:36:00]

Adding estade so he's aware of some of the download shelf changes.

http://codereview.chromium.org/392007/diff/4024/2012
File chrome/browser/gtk/rounded_window.cc (right):

http://codereview.chromium.org/392007/diff/4024/2012#newcode188
Line 188: RoundedWindowData* data = static_cast<RoundedWindowData*>(
I see, so previously we would add two callbacks and clobber data.  Now we just
clobber data.  I guess I would rather we DCHECK in this case (i.e., require
people to first call StopActingAsRoundedWindow).

http://codereview.chromium.org/392007/diff/4024/2013
File chrome/browser/gtk/status_bubble_gtk.cc (right):

http://codereview.chromium.org/392007/diff/4024/2013#newcode171
Line 171: if ((ltr && location.x() < right_threshold ||
I'm surprised that gcc doesn't complain about your use of || and && without
parens.

http://codereview.chromium.org/392007/diff/4024/2015
File chrome/browser/gtk/tab_contents_container_gtk.cc (right):

http://codereview.chromium.org/392007/diff/4024/2015#newcode202
Line 202: if (ltr ^ status->flip_horizontally())
Ok, but you should comment this.

[tony, 2009-11-13 23:01:24]

LGTM

http://codereview.chromium.org/392007/diff/2026/2032
File chrome/browser/gtk/rounded_window.cc (right):

http://codereview.chromium.org/392007/diff/2026/2032#newcode214
Line 214: g_object_steal_data(G_OBJECT(widget), kRoundedData);
Nit: Do we have to delete the data after we steal it?

[Daniel Erat, 2009-11-13 23:09:54]

Thanks!

On Fri, Nov 13, 2009 at 3:01 PM,  <tony@chromium.org> wrote:
> LGTM
>
>
> http://codereview.chromium.org/392007/diff/2026/2032
> File chrome/browser/gtk/rounded_window.cc (right):
>
> http://codereview.chromium.org/392007/diff/2026/2032#newcode214
> Line 214: g_object_steal_data(G_OBJECT(widget), kRoundedData);
> Nit: Do we have to delete the data after we steal it?

Yikes, nice catch.  Yes.

> http://codereview.chromium.org/392007
>