Expands the decorator validation rule to support blacklisting properties whose values match a pattern. This allows disabling things like inline styles.
* The ``MatTextareaAutosize` directive misses a type for the `resizeToFitContent` method. The `gulp api-docs` task showed that the type is missing.
* Improves the `normalizeMethodParameters` method in the Dgeni setup, to also add type information to parameters that have been parsed properly by Dgeni, but still miss a type.
* Fixes the `OverlayKeyboardDispatcher` tests leaking `.cdk-overlay-container` instances which can throw off any subsequent tests. This seems to be what is causing the current test flakes in master that happen if the tests are run in a minimized browser window.
* Gets rid of the explicit cleanup of overlay containers in the overlay container tests. This can cause tests to be even flakier if there are any leaks, in addition to being unnecessary because there shouldn't be any containers to begin with.
* Adds more logging messages to the publish artifacts script (better debugging)
* Updates the `find` command for the version `sed` call to no longer match files inside of the `.git/` directory (might lead to a broken .git/ setup)
Fixes not being able to close a datepicker's calendar in IE11, if the datepicker's trigger opens it on focus. The issue comes down to the fact that all browsers focus elements synchronously, whereas IE does so asynchronously. Since our logic depends on everything firing in sequence, when IE focuses at a later point, the datepicker is already considered as closed which causes the logic that restores focus to the trigger to reopen the calendar.
Fixes#8914.
Currently the expansion panel renders all of its content on load which can be expensive, considering that all of it won't be visible. These changes allow for some of the content to be rendered lazily via the `matExpansionPanelContent` directive.
Fixes#8230.
Adds the `threshold` option to the `CloseScrollStrategy` that allows for the consumer to specify the amount of pixels that the user has to scroll before the overlay is closed.
Fixes not being able to open a closed `mat-select` in single selection mode using the alt + up/down arrow combo. It seems like this was introduced when we added the ability go through the values using arrow keys on a closed select.
Switches all of the ripple-related event listeners to be passive for improved performance and to prevent Chrome from logging warnings for almost every Material component.
As of https://github.com/angular/material2/commit/004e0fef2fe5172b6e6ed2c4d368d21dab051355 the drawer supports a two-way binding on the `opened` property, however because the `openedChange` emits at the end of the animation and two-way bindings invoke their setter which then trigger an animation which in turn invokes `openedChange`, there is the potential for an infinite loop if the user changes the `opened` state mid-animation. These changes avoid the issue by turning the `openedChanged` into an async stream and emitting the current value at the end of the animation, rather than determining the value based on the animation state.
Fixes#8869.
Currently the datepicker re-subscribes to the same `backdropClicked` stream whenever the popup is attached, which causes it to stack up multiple subscriptions for every time the calendar is opened. These changes move the subscription so it only happens once.
Fixes plain inputs (without a `matInput`) that has a `formControl` and an autocomplete attached not being disabled via `FormControl.disable`. The issue was due to the `MatAutocompleteTrigger` not implementing the `setDisableState`. The current implementation is in line with the approach in the `DefaultValueAccessor`.
Fixes#8735.
Fixes clicking on the chip remove button propagating up and potentially triggering its parent control. This is usually handled by the chip itself, however clicking on the button could remove the chip before it has the chance to stop propagation.
Fixes#8771.
Fixes the select's `change` event emitting before the `value` binding has been updated, causing consumers that check the data-bound output to get a stale value.
Fixes#8739.
Joey Perrott