Fix DCHECK in infobar animation.

Fix DCHECK in infobar animation.

Solved in http://codereview.chromium.org/6788014/

Reason was traced down to OnBoundsChanged not getting called when the animation
fired even when the infobar's preferred size should have changed.  Thus Paint
could get a different animation value from that of the previous animation value
without the height being adjusted by Layout.  I think this may be in the case
where one infobar is opening when another is closing and the overall size of the
infobar container is not changing, hence squelches layout propagation to its
children.

In addition, when extensions load, they call set_target_height, which doesn't
cause Layout.

As a hopefully less fragile alternative to plugging all of the places where
Layout should be invalidated, I cache the height of the tab/bar along with the
paths, so that whenever Paint occurs, it always goes against the last SetBounds
values and doesn't have to recalculate based on animation, which may have
progressed without causing a Layout and subsequent SetBounds.

BUG=75451, 75450, 60990
TEST=browser_tests --gtest_filter=ExtensionApiTest.Infobars --gtest_repeat=100

[Sheridan Rawlins, 2011-03-30 01:18:20]

Peter, can you take a look?  Also see the temporary debugging logging
http://code.google.com/p/chromium/issues/detail?id=75451#c13 which lead me to
the RCA and fix.

[Sheridan Rawlins, 2011-03-30 02:43:27]

There must be another case on the try servers...  I'll add some logging there
and keep trying - I'll let you know when the CL is ready for review.

[Sheridan Rawlins, 2011-03-30 19:58:08]

Peter,

I like this fix better than the previous one - rather than trying to find all
places where Layout is required, I save/cache the tab and bar heights when my
bounds change.  Thus if my bounds don't change as expected (for whatever reason,
be it InvalidateLayout not called or set_target_height called but not causing a
Layout, or something else I haven't considered) then I don't have to
re-calculate them when painting based upon the then state of the world and end
up with different sizes than what I'm currently set to.

trybots were run against this patch and FWIW, the chromeos one failed on
unrelated tests after passing the browser_tests.

-SCR

[Peter Kasting, 2011-03-30 20:07:23]

On 2011/03/30 19:58:08, Sheridan Rawlins wrote:
> I like this fix better than the previous one - rather than trying to find all
> places where Layout is required, I save/cache the tab and bar heights when my
> bounds change.  Thus if my bounds don't change as expected (for whatever
reason,
> be it InvalidateLayout not called or set_target_height called but not causing
a
> Layout, or something else I haven't considered) then I don't have to
> re-calculate them

This worries me, because we don't want any places where we're not calling
InvalidateLayout() where we should be.  We need at least your previous change,
plus fixes for any other places you've missed, regardless of whether you do this
current change.

[Sheridan Rawlins, 2011-03-30 20:29:43]

Right - this DCHECK reveals places which were probably janky because they were
not being laid out when our preferred size changed.

I'll add the Invalidate calls back in.

-SCR

On 2011/03/30 20:07:23, Peter Kasting wrote:
> On 2011/03/30 19:58:08, Sheridan Rawlins wrote:
> > I like this fix better than the previous one - rather than trying to find
all
> > places where Layout is required, I save/cache the tab and bar heights when
my
> > bounds change.  Thus if my bounds don't change as expected (for whatever
> reason,
> > be it InvalidateLayout not called or set_target_height called but not
causing
> a
> > Layout, or something else I haven't considered) then I don't have to
> > re-calculate them
> 
> This worries me, because we don't want any places where we're not calling
> InvalidateLayout() where we should be.  We need at least your previous change,
> plus fixes for any other places you've missed, regardless of whether you do
this
> current change.

[Sheridan Rawlins, 2011-03-30 23:48:27]

Added the InvalidateLayout in the SetTargetHeight, and AnimationProgressed
methods to induce layout when our size is changing.

[Peter Kasting, 2011-04-01 17:06:22]

As promised, my version of this is in http://codereview.chromium.org/6788014/ . 
Feel free to close this if you like that better.