Build artifacts will be published differently for different branches. For example, commits in `master` will be published on the `master` branch on the `*-builds` repository. `5.2.x` commits will be published on the `5.2.x` branch on the `*-builds` repository.
Tags on the build repositories will be explicit to indicate from what branch they come. The version placeholders (in the build output) will be similar as before, and won't contain any information about the branch. e.g. `5.2.x-5.2.1-43s435s` will look bad.
Ensures that versions are following a consistent format before publishing. e.g. versions like `6.0.0-beta-0` is not valid, because the last dash should be actually a dot.
Closes#9910.
Scopes the logic, that skips the build if only Markdown files, only to PR builds. This avoids issues where we merge in a lot of PRs to master, followed by a docs PR which ends up silencing potential build errors.
Currently if something throws an error right after Karma has initialized, the log will look something like the following:
```
HeadlessChrome 0.0.0 (Windows 10 0.0.0) ERROR
{
"originalErr": {},
"__zone_symbol__currentTask": {
"type": "microTask",
"state": "notScheduled",
"source": "Promise.then",
"zone": "<root>",
"cancelFn": null,
"runCount": 0
}
}
```
These changes log out the stack trace explicitly so we have some more info to work with if it happens.
* Fixes a minor typo in the `MergeInheritedProperties` processor.
* Fixes an incorrectly placed comment inside of `decorators.ts`.
* Cleans up the `dgeni/index.ts` file
Since the `MatListOption` component requires a parent `MatSelectionList`, the DI of the selection list shouldn't be optional. Otherwise there will be runtime exceptions due to an undefined selection list property.
Adds a method overload to the `launch` method of the `MatRipple` class. This should make it possible for developers to just pass in a configuration without coordinates that will be ignored (e.g. `launch(0, 0, {centered: true});`).
This was the initial use-case for introducing the locked positioning to the `ConnectedPositionStrategy`. Since `mat-select` has some special positioning requirements, it would look weird if we kept repositioning it as the user scrolls. These changes add an input `CdkConnectedOverlay` that allows for the position to be locked and locks it for `mat-select`.
Something that came up while doing the docs. Adds the ability for the consumer to pass back a result from a bottom sheet. Also adds a test for `afterDismissed` since it didn't have one.
* Makes the `added` and `removed` properties on the `SelectionChange` event required since they're guaranteed to be an empty array. This makes it more convenient to consume, because it avoids having to null check them every time.
* Turns the `SelectionChange` into an interface to reduce the amount of code being generated.
Fixes some errors being thrown by the `ViewportRuler` when attempting to measure the viewport or to get the scroll position while rendering on the server.
Relates to #9066.
Currently we only remove the event listeners and CSS classes when we stop monitoring an element, however any subscriptions are still live. These changes complete the subject since it won't be reused if we decide to start monitoring again.
* Adds ripples to the `MatButtonToggle` component
* Fixes that the focus overlay for button toggles shows on mouse/touch press.
* Properly stops monitoring through the `FocusMonitor` on component destroy.
* Removes unnecessary mixin for focus-overlay color. Color can be always set, because the overlay will be toggled through `opacity` (performance improvement)
Closes#9442
Adds support for theming the datepicker using the `color` attribute, as well as inheriting the theme color from the form field around a datepicker input.
This adds bazel test rules for everything under cdk/ and sets CircleCI Overlay tests don't work because there's not yet a way to pass through flags to the browser. These issues will be addressed in a follow-up PR.
* Fixes the `GlobalPositionStrategy` tests leaking `.cdk-overlay-container` instances which can cause flakes in other tests.
* Fixes a wrong method name in the bottom sheet overview.
Bumps to the latest minor version of Angular now that a compilation error, that was blocking us previously, has been resolved. Also fixes a couple of AoT errors in the demo app.
* Creating a custom form field docs -- updated ngControl section to reflect errors injecting NgControl
If the component already implements ControlValueAccessor, injecting NgControl in the constructor (as described in the docs) may fail. Added an alternate way.
* Update creating-a-custom-form-field-control.md
* added word "provider" for clarity
* Hides the warnings about unresolved TypeScript symbols for the mixin base classes. Those warnings are not fixable and are polluting the Dgeni output.
* Adds missing `@deprecated` tags to entries that already have the `@deletion-target` set.
* Updates to the latest version of dgeni-packages because https://github.com/angular/dgeni-packages/commit/3a4e5685589b99b85dfb88e8fd8dbf46aab24088 landed.
Since we have some cases where we pass in positions verbatim to the connected position strategy (the connected position directive), we could end up in a situation where the consumer passed in a set of invalid positions that don't break necessarily, but don't work correctly either. These changes add some sanity checks at runtime to make debugging easier.
* Reworks the tooltip to reuse the `OverlayRef` after the first time it is opened, rather than destroying it and creating a new one for every subsequent open. This reduces the overhead from having the create the new overlays all the time.
* Changes the signature of the `ConnectedPositionStrategy.withScrollableContainers` method to return the position strategy. This is primarily for consistency with the other methods.
In some situations, the `disable()` method of a form control is called while the datepicker input is focused. This causes a changed after checked error.
Normally, native input elements automatically blur if they turn disabled. This behavior is problematic, because it would mean that it triggers another change detection cycle, which then causes a changed after checked error if the input element was focused before.
Fixes#9559
Adds a simple "aria-describedby" input in the manner of "aria-labelledby."
The ability to set "aria-describedby" is essential for accessible radio buttons,
as the attribute allows further description of a radio button to be read
after mentioning the button's label and whether or not it is checked.
Since there's a timeout when announcing something through the `LiveAnnouncer.announce` method, it may be hard/prone to race conditions for the consumer to coordinate it with some other action. These changes expose a promise that will be resolved once the message has been added to the DOM.
Paul Gschwendtner