The popup menu in the stocks app is supposed to size its width to the max
intrinsic width of the menu. This CL teaches it how to do that.

It's a shame that we need to ceilToDouble the output of RenderParagraph. If we
don't do that, we run into floating point layout trouble and the menu triggers
a line break. The correct fix is to do layout in fixed point.

R=ianh@google.com

Review URL: https://codereview.chromium.org/1168113005

I don't think this is the final API we're going to use.

I suspect we'll add a TextStyle struct to hold
color and size, etc. which back-ends into CSS like
this does today.

Currently no one uses this color even though it exists.

This is one step towards fixing:
https://github.com/domokit/mojo/issues/213

R=ianh@google.com

Review URL: https://codereview.chromium.org/1162573007

This CL splits getIntrinsicDimensions into getMinIntrinsicWidth,
getMaxIntrinsicWidth, getMinIntrinsicHeight, and getMaxIntrinsicHeight so that
we can properly shrink-wrap the stocks app menu. This CL just contains the
refactoring. The use in stocks app will come in a later CL.

R=ianh@google.com

Review URL: https://codereview.chromium.org/1167293003

Also change SizedBox.

This makes the code that uses Container() and SizedBox() much more readable.

The underlying RenderSizedBox is not affected by this change.

R=abarth@chromium.org, eseidel@chromium.org

Review URL: https://codereview.chromium.org/1166203002

R=abarth@chromium.org, abarth

Review URL: https://codereview.chromium.org/1164363002

R=eseidel@chromium.org, ianh@google.com

Review URL: https://codereview.chromium.org/1166183002

The popup menu is displayed in the wrong location with the wrong width, but it
does draw. I've also removed the checkbox because the flex container was
causing me trouble.  I'll add it back in a later CL.

R=ianh@google.com

Review URL: https://codereview.chromium.org/1166153002

TBR=abarth

Review URL: https://codereview.chromium.org/1170783002

This CL is the first step towards getting PopupMenuItem working in fn2. We
introduce the ability to add a minWidth to a Container by creating a
RenderConstrainedBox class to apply the new constraints.

R=ianh@google.com

Review URL: https://codereview.chromium.org/1155683011

This includes making SizedBox default to infinite size, and making
Container default to containing one infinite-sized SizedBox. That way,
you can use an empty Container as a spacer in a flex box.

The rendering doesn't quite work, because of a bug in flex whereby it
doesn't shrink-wrap its contents in the cross-direction, it fills the
parent. Collin is on that.

R=jackson@google.com

Review URL: https://codereview.chromium.org/1163633003

And deploy it in a few places.

R=ianh@google.com

Review URL: https://codereview.chromium.org/1164303002

TBR=abarth

Review URL: https://codereview.chromium.org/1169543009

Changes:
- Remove all the old tracing logic since that was a lot of code for
  just random printf debugging and nobody seems to use it
- Convert "defunct", a permanent state transition, to "mounted", a
  state that can be entered and exitted arbitrarily
- Coallesce mount/unmount notifications so there's only one per frame
- Remove handleRemoved() since we don't use it
- Rename a local variable 'parent' to 'ancestor' to avoid shadowing
  the field
- Replace uses of forEach() with for-in

Effects:
- Reusing a node provided to a constructor in a new place in a tree
  (especially at a different depth) no longer fails (it used to fail
  because the node would be marked defunct then remounted later, at
  which point it would fail)

R=abarth@chromium.org

Review URL: https://codereview.chromium.org/1169673002

R=abarth@chromium.org, abarth

Review URL: https://codereview.chromium.org/1161813005

R=abarth@chromium.org

Review URL: https://codereview.chromium.org/1156383006

We were hitting an assert in RenderCustomPaint that the callback wasn't null.

R=eseidel@chromium.org, jackson@chromium.org

Review URL: https://codereview.chromium.org/1162023010

Export Point, Size, Rect, Color, Paint, Path, BoxDecoration, Border, BorderSide, EdgeDims, and FlexDirection from fn2.dart
That way most of our libraries don't have to use sky.Foo all over the place, and we can reduce the number of imports in most Dart components.

R=abarth@chromium.org

Review URL: https://codereview.chromium.org/1161323004

R=abarth@chromium.org, abarth

Review URL: https://codereview.chromium.org/1171543002

The + isn't quite centered properly, but the button is a circle in the right
place.

R=eseidel@chromium.org, ianh@google.com

Review URL: https://codereview.chromium.org/1156383004

 - several places where we use a getter knowing that it will return a
   subclass of its declared type, even asserting that it does, but
   where the analyzer is worried that the getter might start returning
   a new value unexpectedly, solved by having a temporary local
   variable shadow the getter and asserting that it doesn't change
   value
 - many many places where we do this with parentData specifically,
   solved by type-erasing parentData.
 - a place where a mixin wants to be subclassing another class, and
   uses its methods, solved by saying that the mixin is abstract but
   implements the superclass

R=abarth@chromium.org

Review URL: https://codereview.chromium.org/1160923008

Turns out Scaffold needs to hook _remove, so we need to rename it to remove to
expose it outside fn2.dart.

R=ianh@google.com

Review URL: https://codereview.chromium.org/1160013007

R=abarth@chromium.org, eseidel@chromium.org, abarth

Review URL: https://codereview.chromium.org/1157573012

Fixed the mojo deploy script to no longer deploy mojo.
mojo now uses a more sophisticated CDN, etc.

Fixed old fn.dart to work again. :)  Looks like it was
victim of a mass rename.

Fixed engine.cc to be able to handle navigating from
WebView content to SkyView content.  We can't handle
the other direction yet, but we don't need to.
/sky_home is written in the .sky system but all of our
new examples are .dart (and thus SkyView).

R=abarth@chromium.org

Review URL: https://codereview.chromium.org/1165003006

Rename "BorderSide.None" to "BorderSide.none", since it's a constant, and change the sky.Size.infinite() constructor to a constant.

R=abarth@chromium.org

Review URL: https://codereview.chromium.org/1158263005

Include DETACHED, NEEDS-LAYOUT, and relayoutSubtreeRoot info in the debug output for the render tree

R=chinmaygarde@google.com

Review URL: https://codereview.chromium.org/1149083011

RenderScaffold needs to attach and detach its children when asked.

R=eseidel@chromium.org, ianh@google.com

Review URL: https://codereview.chromium.org/1149893025

Improve the RenderObject.toString() logic so that there's not blank lines everywhere and to generally be better.
Adds sizes, positions, transforms, improves RenderParagraph's output.

R=abarth@chromium.org

Review URL: https://codereview.chromium.org/1160233004

Does not work yet.

R=ianh@google.com

Review URL: https://codereview.chromium.org/1163973005

R=eseidel@chromium.org, ianh@google.com

Review URL: https://codereview.chromium.org/1166953005

We should be asserting that old.root is a ContainerRenderObjectMixin because
we're able to manipulate its child list.

R=ianh@google.com

Review URL: https://codereview.chromium.org/1169593002

The goal of this is to have a Renderer-layer abstraction
around the Sky Engine c++ apis for drawing a shadow.

R=mpcomplete@chromium.org

Review URL: https://codereview.chromium.org/1146893004

R=abarth@chromium.org

Review URL: https://codereview.chromium.org/1151533006

R=eseidel@chromium.org

Review URL: https://codereview.chromium.org/1162623011

This makes debugging the render tree a lot easier. Just print the node
you care about, and you get an indented tree view of its subtree,
including settings. New subclasses should implement the new virtual
method debugDescribeSettings() to expose new settings.

R=eseidel@chromium.org

Review URL: https://codereview.chromium.org/1157993005

This CL makes FixedHeightScrollable able to measure heights. In the DOM-based
version, FixedHeightScrollable forced a synchronous layout in a microtask in
order to measure the size of itself and its children. In the DOM-less version,
we now require the subclass to tell us the item height and we measure our own
height using a layout callback.

R=ianh@google.com

Review URL: https://codereview.chromium.org/1161063005

R=abarth@chromium.org, abarth

Review URL: https://codereview.chromium.org/1163913004

https://www.dartlang.org/articles/style-guide/#names

R=jackson@google.com

Review URL: https://codereview.chromium.org/1158813004

R=abarth@chromium.org

Review URL: https://codereview.chromium.org/1169583002

R=abarth@chromium.org

Review URL: https://codereview.chromium.org/1144933007

In the future, we might want to extend this class to clip to a configurable
rect in addition to clipping to its bounds.

R=eseidel@chromium.org, ianh@google.com

Review URL: https://codereview.chromium.org/1145263006