BREAKING CHANGE:
- Renderer:
* renderComponent method is removed form `Renderer`, only present on `RootRenderer`
* Renderer.setDebugInfo is removed. Renderer.createElement / createText / createTemplateAnchor
now take the DebugInfo directly.
- Query semantics:
* Queries don't work with dynamically loaded components.
* e.g. for router-outlet: loaded components can't be queries via @ViewQuery,
but router-outlet emits an event `activate` now that emits the activated component
- Exception classes and the context inside changed (renamed fields)
- DebugElement.attributes is an Object and not a Map in JS any more
- ChangeDetectorGenConfig was renamed into CompilerConfig
- AppViewManager.createEmbeddedViewInContainer / AppViewManager.createHostViewInContainer
are removed, use the methods in ViewContainerRef instead
- Change detection order changed:
* 1. dirty check component inputs
* 2. dirty check content children
* 3. update render nodes
Closes#6301Closes#6567
To workaround https://github.com/Microsoft/TypeScript/issues/7573
we must remove the readonly keyword from generated .d.ts files.
This solution will not scale, but will probably buy enough time to require our users move to a 2.0 beta.
Closes#8003
Also enable DDC checks across all non-web worker playground apps. We are
now down to 2 DDC errors across all of them. The remaining two need to be
fixed in package:analyzer, not in angular.
BREAKING CHANGE:
- there's a chance of breakage as router's Instruction constructor
signature changed.
Closes#6693
Previously we grepped all hand-written Dart code and ran analyzer in strong mode against it.
Now we run it against transformed playground apps, which:
1. does not analyze unnecessary code (we primarily care about stuff that runs in the browser)
2. analyzes generated code, which does run in the browser and which we failed to analyze in the previous version of the build
Closes#6436
Originally we ran gulp enforce-format at the beginning of the build.
This was annoying because you came back from lunch to find that no tests
ran so you have to start your PR over.
Then we changed it to run the linters at the end. This is annoying because
you might be ready to merge to master, and could have fixed the lint
issues immediately, but now much wait for another PR.
The solution is to run the lint checks in another build. This marks
your PR red very early, but you still get the feedback of whether the
tests are passing.