Aura: Fix window resizing for large drags.

Aura: Fix window resizing for large drags.

* Fix a DCHECK due to a negative width when resizing the right edge past the left one, and vice versa.
* Fix the window moving when resizing its left edge past the minimum size, ditto for top edge.
* Add unit tests for above.
* Add GetMinimizeSize() method to aura::WindowDelegate interface.

BUG=104245
TEST=aura_shell_unittests, manual

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=111244

[James Cook, 2011-11-21 23:39:04]

Oshima/Xiyuan - just FYI.

Ben, PTAL.  If a window resize operation is clamped due to hitting the minimum
height/width, we may need to adjust changes to the window's origin.  This code
computes a corrected delta_x/delta_y drag amount, rather than just blocking the
origin change, to prevent a single-frame "popping" effect when you reshape down
to the limit and back.

I tried writing the resize code with a +1/-1 multiplier for left/right and
top/bottom size changes, but I found it more confusing than just writing out the
various cases explicitly.

[Ben Goodger (Google), 2011-11-22 16:01:58]

Also, how does this interact with the existing minsize stuff xiyuan did?

http://codereview.chromium.org/8618009/diff/1/ui/aura_shell/toplevel_window_e...
File ui/aura_shell/toplevel_window_event_filter.cc (right):

http://codereview.chromium.org/8618009/diff/1/ui/aura_shell/toplevel_window_e...
ui/aura_shell/toplevel_window_event_filter.cc:247: if (!(bounds_change &
kBoundsChange_Resizes))
I hate these sort of d&d/sizing functions. They always seem a bit "write-only"
to me. Can you describe the problem? I thought I had fixed all these weird
issues when I originally wrote this filter. I suspect we've had a regression if
it's malfunctioning.

[James Cook, 2011-11-22 19:45:36]

Ben, please take another look.  I think this is as simple as I can write it.

I removed the OnBoundsChanging() method from WindowDelegate, since all it was
doing was enforcing a minimum size, and I need to do that enforcement in
aura_shell anyhow.  I don't feel strongly about removing it -- I could put it
back if you foresee another use.

[xiyuan, 2011-11-22 20:25:41]

On 2011/11/22 19:45:36, James Cook (Chromium) wrote:
> Ben, please take another look.  I think this is as simple as I can write it.
> 
> I removed the OnBoundsChanging() method from WindowDelegate, since all it was
> doing was enforcing a minimum size, and I need to do that enforcement in
> aura_shell anyhow.  I don't feel strongly about removing it -- I could put it
> back if you foresee another use.

Please try HtmlDialogBrowserTest.SizeWindow. The last test in the test case is
to ensure a HTMLDialog never goes to 0x0.

[James Cook, 2011-11-22 21:02:43]

On 2011/11/22 20:25:41, xiyuan wrote:
> On 2011/11/22 19:45:36, James Cook (Chromium) wrote:
> > Ben, please take another look.  I think this is as simple as I can write it.
> > 
> > I removed the OnBoundsChanging() method from WindowDelegate, since all it
was
> > doing was enforcing a minimum size, and I need to do that enforcement in
> > aura_shell anyhow.  I don't feel strongly about removing it -- I could put
it
> > back if you foresee another use.
> 
> Please try HtmlDialogBrowserTest.SizeWindow. The last test in the test case is
> to ensure a HTMLDialog never goes to 0x0.

OK, I fixed HtmlDialogBrowserTest for linux_aura by adding the minimum size
check to aura::Window::SetBoundsInternal().  Again, I don't feel strongly about
the WindowDelegate::OnBoundsChanging() method -- I felt like I should remove it
because I added a new method to the interface and it felt a little redundant as
it's currently used.

[Ben Goodger (Google), 2011-11-22 21:21:50]

lgtm