This CL also refactors common logic for RenderPhysicalModel and
RenderPhysicalShape into a base class _RenderPhysicalModelBase.

Apple encodes syslog entries using a 7-bit encoding where input UTF-8 bytes
are encoded as follows:
1. 0x00 to 0x19: non-printing range. Some ignored, some encoded as <...>.
2. 0x20 to 0x7f: as-is, with the exception of 0x5c (backslash).
3. 0x5c (backslash): octal representation \134.
4. 0x80 to 0x9f: \M^x (using control-character notation for range 0x00 to 0x40).
5. 0xa0: octal representation \240.
6. 0xa1 to 0xf7: \M-x (where x is the input byte stripped of its high-order bit).
7. 0xf8 to 0xff: unused in 4-byte UTF-8.

As there doesn't appear to be a system tool to decode these strings, we
implement here in Dart. If we're unable to decode a string (e.g.
decoding results in an invalid UTF-8 string), we fall back to emitting
the log line as-is.

Just download a Flutter ZIP bundle, unzip, and double-click on `flutter_console.bat` to get started. No PATH modifications required.

And add a test that catches these problems.

This gets us back to where we were a few minutes ago.

This reverts commit 592f2979.

* Revive "Push physical layers with paths instead of rects in RenderPhysicalModel. (#13942)"

That commit was reverted due to a performance problem which has now been
fixed.

* Roll engine 05fe72d068e19c7886e8d27f9b004201d5ad1300

Fixes https://github.com/flutter/flutter/issues/14058.

* Reland 9534082f with fix for incremental compilation.

When in incremental mode, awaiting exitCode won't work because compiler is not expected to exit after compilation.
Instead listen for stdout stream closing and report error if outputFilename has not been received.

* Fix lints

* Fix handling --preview-dart-2 for ios

* final var

* send 'reloadRecommended' info to IDEs

* rename hint ==> hintMessage

* Remove mention of --assert-initializer from CONTRIBUTING.md
* Remove options variable as well

This reverts commit 9534082f.

Causes hot_mode_dev_cycle__preview_dart_2_benchmark test timeout.

* Check frontend_server exit code.

When frontend_server completes abnormally, flutter tools has to stop and let user know.

* Add tests

In the packaged ZIP files for Flutter we bundle mingit on Windows to make setup for our users easier (they don't have to download and install git anymore). This change makes Flutter use the bundled mingit if it is available and if git is not installed on the system.

This change will still require that users add `flutter\bin` to their path or they'll have to execute `flutter\bin\flutter.bat` directly. A follow-up PR will add a pre-configured Flutter Console that people can start with a double-click to minimize setup even further.

Also in this PR:
* If 7Zip is not available to unpack the Dart SDK try 7Zip's standalone version.

This provides a way to draw colors, images, and general decorations on Material widgets, without interfering with InkWells that are further descendants of the widget.

This thus provides a cleaner way to solve the issue of FlatButtons and InkWells not working when placed over Image widgets than the old hack of introducing a transparency Material.

Fixes #3782.

Also, some fixes to documentation, and remove a redundant property on the Image widget.

Rolling the Dart SDK to 2.0, with associated package updates.  

https://groups.google.com/forum/#!topic/flutter-dev/E22RdKAYjs8

* Revert "Reverting package changes until I can figure out how to fix Windows. (#14007)"

This reverts commit 6fda8ee8.

* Make prepare_package run on Windows

* Add hot_mode preview_dart_2 tests for ios.

* Explicitly request ios os in the test.

* 2018

* libtxt now calculates selection rectangles reflecting the heights of
differently styled text spans within a line (matching Blink)
* libtxt does not round the coordinates of selection rectangles (unlike Blink)

* Do not crash when toggeling Semantics on, off, on

* review comments

Fixes https://github.com/flutter/flutter/issues/12553

The root cause of https://github.com/flutter/flutter/issues/12553 was that when the refresh indicator went away, which happened after the avatars had started their image loads but before the avatars had loaded, the DecoratedBoxes, along with the rest of the list, got GlobalKey-reparented, which caused the RenderDecoratedBox objects to unregister from the DecorationImagePainters, but they never re-registered since the whole subtree was in a RepaintBoundary and was therefore not explicitly repainted when the tree got reattached.

This fixes the bug by explicitly requiring any RenderDecoratedBox to repaint when it's reattached. This is probably a little more aggressive than required; we could probably expose a flag on Decoration that says whether or not the onChanged handler will ever be invoked, and only call markNeedsPaint if that's true, but we'll do that if it turns out that there's a performance issue here.

(This patch also introduces a bunch of improved debugging information that I used to track down the bug.)