* fix(input): apply readonly attribute when readonly
Fixes issue where the readonly attribute was no longer applied because it was made to be an `@Input`
* Update input.ts
* Now re-runs the given deploy mode a second time if failed the first time. This ensure that deployment works fine most of the time and should also avoid conflicts between different Travis CI jobs that are deplyoing at the same time.
* Currently the TypeScript source files are being copied to the package output. This is only necessary for the AOT task and therefore shouldn't happen if someone tries to run the devapp.
* Adds the selection list component to the kitchen sink.
* Piggy-backing off of the smaller fix, this is something I've meant to do for a while, but it's so minor that it doesn't make sense to open a separate PR. Currently there are a few unit tests that will cause Karma to hang until it times out, if they fail. This seems to be because Karma fails to stringify the `activeOption`. These changes switch the tests to cast the expected result to a boolean first.
* Adds a fallback position to ensure that tooltips can't go outside the viewport.
* Refactors the `transform-origin` logic to depend on the overlay position, instead of the tooltip position.
Fixes#5428.
Currently the checkbox only allows changing the tabIndex using the tabIndex binding. Using the native tabindex attribute doesn't have any affect. With this change the native tabindex property will be respected.
References #6465
Prevents the doctype warning from being logged when rendering server-side in development mode. Previously, while we did have a `document`, it was not the same as the client-side `document` and thus didn't have a `doctype`. We didn't notice this earlier, because we run the server-side rendering check in production mode where the sanity checks are disabled.
Relates to #6292.
* feat(selection-model): de/select multiple values at the same time
Adds support for passing multiple values to the SelectionModel at the same time. This feature is described in the JSDoc already, but just wasn't implemented yet.
This functionality gives us more control about the `onChange` event, which will otherwise fire every time if for example multiple values need to be selected.
* Address comments
* fix(common-module): check if computed styles are available
Due to a Firefox Bug (https://bugzilla.mozilla.org/show_bug.cgi?id=548397) the `getComputedStyle` call in the `MdCommonModule` will return `null` and cause a runtime exception because normally the computed styles are never `null` for the test element.
Since this it's very uncommon that a developer places a Angular Material application inside of a hidden iframe, a simple null check should be enough to get the application running.
Fixes#7000
* Add comment about Firefox case
Don't emit a closed event when another event will be emitted. Previously, if one clicked on a menu item, one would get two events: `undefined` and `click` in that order. One would see similar behavior for `keydown` or clicking the backdrop. Unit tests were updated to prevent a regression.
Whenever the value of the slide-toggle changes through interaction, a `change` event will be emitted. The value change then will be reported to the `ControlValueAccessor` which fires the `ngModelChange` output.
Right now the `ngModelChange` output fires after the normal `change` output. This should be the other way around because developers expect the two-way data binding to already reflect the change.
Fixes#7074
Fixes an error that was thrown when an autocomplete trigger is initialized without a panel. Note that an error will still be thrown, but only if the user attempts to open the panel.
Fixes#7069.
* Fixes subsequent snack bars not showing up if they are opened while another snack bar is open. This was due to the latter snack bar getting closed immediately, because one of the animation events coming in is `void->void`.
* Simplifies the logic for determining the animation state. Using the getter ended up passing returning an invalid state in some cases (e.g. `undefined-top` or `undefined-bottom`).
Fixes#7063.
Since the `direction` option from the overlay only sets the `dir` attribute, it means that any components that depend on the `Directionality` service (e.g. `md-slider`) won't be able to pick it up. These changes add the proper direction to the custom dialog injector.
* Removes the `destroyed` and `onFocus` observables and instead just calls methods on the injected `MdSelectionList` instance.
* Fixes that the active item is not updating if the active item is being destroyed.
Fixes the calendar not re-rendering when the `minDate`, `maxDate` or `dateFilter` change. The issue was due to the fact that the `minDate`, `maxDate` and `dateFilter` weren't being passed down to the views via `@Input`.
Fixes#7202.
Fixes an issue that caused the autocomplete not to be cleaned up when the user tabs away, if there were no results when the autocomplete was open.
Fixes#7268.
Prevents readonly inputs from being highlighted when they're focused. This is consistent with the native input behavior.
Note: This was introduced initially through #5776, but it looks like it didn't make it through the transition to `mat-form-field`.
* In some situations, the change event of the `<mat-slider>` is not being emitted if the user drags the thumb of the slider. This is because the `slidestart` event fires a second time before the sliding ends. This causes the slider to think that there was no value change.
Fixes#7207
Aligns the stepper appearance closer to the spec by:
* Removing the focused background when focused by mouse. Now it only shows up for programmatic and keyboard focus.
* Adding ripples to the step header.
* Using a slightly heavier font font the active step.
* Using `cursor: pointer` to show that the step is clickable.
Fixes#7260.
Andrew Seguin