Dedupe BaseTexture dimension updating (#3680) ...

* dedupe update dimensions

* lint

Added recursive children hitTest tests (#3665) ...

* added recursive children hitTest tests

* remove superfluous parameter

lint

trigger initial transform calc

Merge remote-tracking branch 'upstream/dev' into container_recalc_transform_tests

Merge branch 'dev' into dev-geometry-update ...

# Conflicts:
#	test/core/WebGLRenderer.js

Added legacy option to webGL renderer (#3630) ...

* Added legacy option to webGL renderer

Setting ```legacy``` to true now forces VertexArrayObjects and
MultiTexture Batching to be disabled. We also store the variable on the
renderer just incase we want to use it to force settings in anything
else.

also added a test

* remove WebGLTests as they don't work on server

* Added extra param to Application

* Adds WebGL test for legacy

* Better check for WebGL in tests

Better respecting of the scale mode (#3678)

Added nine slice back in fixed filter merge issue

added Graphics.containsPoint with no fill expectation (#3608)

Prepare docs fix (#3624) ...

* Fixing prepare docs
Putting into right namespace, and updating the list of things you can upload for clarity

* Lint fix

Able to set the alpha on Text drop shadows (#3651)

Add missing 'flag updating' (#3640) ...

* Add missing 'flag updating'

When add or remove child , the flags should be changed.

* Update Container.js

* Update Container.js

fix doc comments (#3673)

Merge branch 'dev-legacy-option' into dev-geometry-update ...

# Conflicts:
#	src/core/renderers/webgl/managers/FilterManager.js
#	src/extras/cacheAsBitmap.js

Merga all but interaction

Merge remote-tracking branch 'origin/dev-legacy-option' into dev-legacy-option

Added extra param to Application

Merge branch 'dev' into dev-legacy-option

Minor tweaks to filter

Program now extracts information from an actual shader rather than the source

added container recalc transform tests

Linty lint

FilterManager shaderCache reset as object (#3664) ...

The shaderCache is initially created as an object. When the Filter Manager gets destroyed, it should be reset to an empty object, not an array.

Refactored shader to hava a program + uniforms

Doc update (#3661)

Improve error message for out-of-bounds Textures (#3659)

Rebasing on dev to get MockPointer changes for #3375 (#3546) ...

* Fix issue #2662 Added a touch counter instead of flag to ensure the correct touch count

* See #2662 --- bending to the will of eslint

* See #2662 --- Added _touchCount to interactiveTarget.js, and removed an extraneous line in InteractionManager.js

* Removed an unnecessary check for _touchCount === undefined, since the mixin for interactiveTarget will always set this

* Fix issue #2662 Added a touch counter instead of flag to ensure the correct touch count

* See #2662 --- bending to the will of eslint

* See #2662 --- Added _touchCount to interactiveTarget.js, and removed an extraneous line in InteractionManager.js

* Removed an unnecessary check for _touchCount === undefined, since the mixin for interactiveTarget will always set this

* Adding cancel events to interaction manager

* Refactoring InteractionManager to normalize all events to PointerEvents. Also updated InteractiveTarget to handle multiple touch or pointer events by recording pointerId for all tracked events, rather than just counting events. Finally, InteractionManager now updates on all tracked events, rather than just mouse events, so it should send move events when UI objects move underneath any held event, not just the mouse

* Typo: mouse is not a function

* InteractionManager normalizeToPointerData now returns an array of normalized events for TouchEvents with multiple changedtouches

* Typo: no such thing as onPointerStart

* Fixing some more typos: trying to read the pointerId of un-normalized events

* Typos: un-normalized events don't have a pointerType

* Added identifier property to InteractionData, and got rid of some typos---trying to read pointerId from un-normalized events

* Forgot to bind onPointerCancel

* Refactoring interactiveTarget.js a bit, so that over, left down and right down are pointer properties, not mouse properties.

* Forgot an export default

* Changed the mixin pattern for Pixi.interaction.interactiveTarget, so that this is bound correctly, so that DisplayObjects can each get a unique object for tracking pointers

* Fixed some typos, moved interactiveTarget back to being an object, and released InteractionData after touch up/cancelled events

* Small fixed

* Safari compatibility

* hasOwnProperty fix

* Safer to use a string for mouse pointer than a number

* InteractionManager should now handle native DOM pointer events with pointerType mouse or touch by emitting mouse and touch events, respectively

* InteractionManager test now respects the refactored spec for InteractionManager

* Pointer events exclusive

* String.startsWith is not available in all browsers esp. older version of Safari. As it's just used in one place we can use a RegExp here too

* Moving from regex, slightly faster

* See #3149 --- Add support for touchcancel and pointercancel (#3375)

* Fix issue #2662 Added a touch counter instead of flag to ensure the correct touch count

* See #2662 --- bending to the will of eslint

* See #2662 --- Added _touchCount to interactiveTarget.js, and removed an extraneous line in InteractionManager.js

* Removed an unnecessary check for _touchCount === undefined, since the mixin for interactiveTarget will always set this

* Fix issue #2662 Added a touch counter instead of flag to ensure the correct touch count

* See #2662 --- bending to the will of eslint

* See #2662 --- Added _touchCount to interactiveTarget.js, and removed an extraneous line in InteractionManager.js

* Removed an unnecessary check for _touchCount === undefined, since the mixin for interactiveTarget will always set this

* Adding cancel events to interaction manager

* Refactoring InteractionManager to normalize all events to PointerEvents. Also updated InteractiveTarget to handle multiple touch or pointer events by recording pointerId for all tracked events, rather than just counting events. Finally, InteractionManager now updates on all tracked events, rather than just mouse events, so it should send move events when UI objects move underneath any held event, not just the mouse

* Typo: mouse is not a function

* InteractionManager normalizeToPointerData now returns an array of normalized events for TouchEvents with multiple changedtouches

* Typo: no such thing as onPointerStart

* Fixing some more typos: trying to read the pointerId of un-normalized events

* Typos: un-normalized events don't have a pointerType

* Added identifier property to InteractionData, and got rid of some typos---trying to read pointerId from un-normalized events

* Forgot to bind onPointerCancel

* Refactoring interactiveTarget.js a bit, so that over, left down and right down are pointer properties, not mouse properties.

* Forgot an export default

* Changed the mixin pattern for Pixi.interaction.interactiveTarget, so that this is bound correctly, so that DisplayObjects can each get a unique object for tracking pointers

* Fixed some typos, moved interactiveTarget back to being an object, and released InteractionData after touch up/cancelled events

* Small fixed

* Safari compatibility

* hasOwnProperty fix

* Safer to use a string for mouse pointer than a number

* InteractionManager should now handle native DOM pointer events with pointerType mouse or touch by emitting mouse and touch events, respectively

* InteractionManager test now respects the refactored spec for InteractionManager

* Pointer events exclusive

* String.startsWith is not available in all browsers esp. older version of Safari. As it's just used in one place we can use a RegExp here too

* Moving from regex, slightly faster

* Fixed interaction tests for the new world of pointer events

Added a fix for the Chrome on android (#3656)

fix unexpected 0 text behaviour (#3650) ...

We got a weird glitch upgrading from pixi 3 to 4, setting the text in a PIXI.Text to 0 results in the label actually saying ' '. This seems like unexpected behaviour considering all the other numbers work :)