Earlier, when the —enable-vulkan flag was specified, the out directory name was determined by the other flags specified to `flutter/tools/gn`. But, the build itself was Vulkan enabled. Now that we are going to add the vulkan target to buildbot, we want to clearly delineate a Vulkan enabled build from a non-Vulkan enabled one.
This updates our compilers, gn and ninja. Also, we now explicitly specify the buildtools artifacts we actually use instead of downloading everything from the catalog.
The build error was due to unused lambda captures being reported as warnings (which we turn into errors).
This updates our compilers, gn and ninja. Also, we now explicitly specify the buildtools artifacts we actually use instead of downloading everything from the catalog.
Ensure that both self.text and self.selectedTextRange are updated before
triggering textDidChange: or selectionDidChange: on the input delegate
(which then re-computes positions). This prevents inconsistencies in
selection vs text position/offset lookups triggered by didChange
notifications.
Ensure that rangeOfComposedCharacterSequenceAtIndex: is only ever called
for index values > 0 && < self.text.length. This prevents such calls for
position with index == self.text.length.
Also reduces unnecessary textDidChange/selectionDidChange notifications:
replaceRange:withText: and setMarkedText:selectedRange: are only ever
called by UIKit and methods called by UIKit (insertText,
deleteBackward), so notification is unnecessary.
* Remove the Point class.
* Add Size.contains and Rect.translate for consistency
* Make Offset and Size compare == with subclasses.
Without this we would break the _DebugSize hack.
* Fix copy/pasta
* Fix the examples in the engine repo
textContentType was added in iOS 10, and we target iOS 8. Either way the
framework doesn't (yet) include support for this; keyboard type covers
most bases.
Supports:
* autocorrect, suggestions view/selection.
* Support for Chinese, Japanese, Korean and other languages using
multi-stage input with candidate selection.
* styling the composing (mark) region for multi-stage input
* key-repeat when backspace held down
* physical and soft keyboard support for arrow key cursor movement and
selection.
* improved third-party keyboard support
Does not yet support:
* force-touch drag cursor positioning
* auto-positioning the candidates view for multi-stage input methods
when used with a physical keyboard (e.g., iPad Pro) currently
positioned at 0,0.
* voice dictation
This version does not recycle buffers. It is meant only as an initial solution,
and the proper buffer recycling logic has been broken out into a separate change
