This option was introduced out of caution as a way for developers to opt out of
the new behavior in v18 which scheduled change detection even when
events happened outside the NgZone. After monitoring the results post-release, we
have determined that this feature is working as desired and do not believe it
should ever be disabled by setting this option to `true`.
This option was deprecated in v18.2, less than 3 months after it was
introduced in v18. We do not really expect it to be used.
BREAKING CHANGE: `ignoreChangesOutsideZone` is no longer available as an
option for configuring ZoneJS change detection behavior.
PR Close#62700
The internal error handler in TestBed rethrows errors to prevent them
from being silently ignored in tests. Prior to this commit, tests which
used `provideZoneChangeDetection` in the providers would override the
internal error handler of TestBed and prevent these errors from being
rethrown.
BREAKING CHANGE: (test only) - Using `provideZoneChangeDetection` in the
TestBed providers would previously prevent `TestBed` from rethrowing
errors as it should. Errors in the test will now be rethrown, regardless
of the usage of `provideZoneChangeDetection`. Tests should be adjusted to
prevent or account for these errors. As in previous major versions,
this behavior can be disabled with `rethrowApplicationErrors: false` in
`configureTestingModule` as a last resort.
PR Close#63404
This fixes a bug introduced in #62994 where `toPromise` was used. This
doesn't work in all situations here because some observables don't
complete. This only affected the redirect path, since the others are
already behind other rxjs code which takes the first value from the user
guards.
Note: This is marked as a refactor rather than fix because the commit above is not in
any release yet.
PR Close#63485
To aid in hitting external breaking change deadlines without pressure
of fixing everything in g3 first, add an internal opt out flag.
This also adds a privately exported provider to revert to the old
rxjs-based behavior, which can be synchronous, until any issues that
come up are addressed.
PR Close#62994
This is effectively a revert of 72e6a948bb.
Debugging the recognize stage is considerably easier with async/await
stacks compared to rxjs. This also improves maintainability and is a
better 1:1 with server-side logic that has been implemented to match
and can be more easily kept in sync.
This also ensures that the recognize step is always async, whereas it
can sometimes be synchronous with rxjs.
BREAKING CHANGE: Router navigations may take several additional
microtasks to complete. Tests have been found to often be highly
dependent on the exact timing of navigation completions with respect to
the microtask queue. The most common fix for tests is to ensure all
navigations have been completed before making assertions. On rare
occasions, this can also affect production applications. This can be
caused by multiple subscriptions to router state throughout the application,
both of which trigger navigations that happened to not conflict with the
previous timing.
PR Close#62994
This commit updates provider definitions that relied on the `deps` array
with `new Optional()` and `new SkipSelf()` to instead use the modern
`inject(..., { optional: true, skipSelf: true })` API.
Previously:
deps: [[KeyValueDiffers, new SkipSelf(), new Optional()]]
Now:
const parent = inject(KeyValueDiffers, { optional: true, skipSelf: true });
**Bundle size reduction**: `Optional` and `SkipSelf` are runtime values
created by `makeParamDecorator()`. Even in production builds, esbuild
and other bundlers must keep their factory code because they are
referenced with `new Optional()` / `new SkipSelf()`. With `inject()`,
those classes are no longer referenced, allowing them and the
`makeParamDecorator` scaffolding to be tree-shaken when unused.
As a result, production bundles can drop both `Optional`, `SkipSelf`, and
their supporting factory code when not used elsewhere, reducing code size
while keeping the same behavior.
PR Close#63386
This commit drops the `X-Request-URL` header. It was a non-standard HTTP response header, deprecated long ago and never part of any official specification. Modern browsers now expose the final URL via the `XMLHttpRequest.responseURL` property, as defined in the WHATWG spec.
PR Close#61943
This fixes the rare case that someone uses binding syntax with `animate.leave` providing a value with a string that has spaces in it. For example:
```
<example `[animate.leave]="'class-a class-b"` />
```
fixes: #63365
PR Close#63366
Change direct deps in bazel targets and import specifiers within files to maintain strict deps requirements ahead of enabling strict deps tests in the repo
PR Close#63323
There's special logic in place to prevent duplicate nodes from showing up in the case when an `@if` toggles a view quickly. This had the unfortunate side effect of causing `@for` leave animations to get cancelled when an add and remove happened simultaneously, even if it was a different index. This fix prevents that from happening in the `@for` loop case.
fixes: #63307
PR Close#63328
The `FormArrayDirective` will allow to have a `FormArray` as a top-level form object.
* `NgControlStatusGroup` directive will be applied to the `FormArrayDirective`
* `NgForm` will still create a `FormGroup`
Fixes angular#30264
BREAKING CHANGE: This new directive will conflict with existing FormArray directives or formArray inputs on the same element.
PR Close#55880
In the case that someone wants to disable animations via selector specificity, for example by adding an `.animate-disabled` class to a parent node, we need to make sure the animate instructions don't misbehave. Now we detect if animations exist in the provided classes and react accordingly.
fixes: #63161
PR Close#63242
Ensure consistency in error message wording by aligning the NG0303
error with other Angular error strings. This improves clarity and
maintains a uniform developer experience
PR Close#63222
Host bindings for `(animate.enter)` and `(animate.leave)` were not firing properly. This fixes the compiler ingest to make sure they do fire.
fixes: #63199
PR Close#63217
This applies the same fix that fixed the class version to the event binding version. It prevents duplicate elements from being on screen when animations have been toggled too fast.
fixes: #63127
PR Close#63216
Fixes that the pipeline wasn't processing the fallback content of `ng-content` for i18n which resulted in a compiler error further down the line.
Fixes#63065.
PR Close#63156
The `pnpm install` command was failing for the `packages/core/test/bundling` project with the error:
`ERR_PNPM_NO_MATCHING_VERSION No matching version found for @angular/compiler-cli@0.0.0-PLACEHOLDER`
This happenes only in renovate updates.
PR Close#63105
In #62758 we started loading the component resources during bootstrap in JIT mode to ensure that they're in place by the time we create the component. This won't work for lazy-loaded components in the router, because they don't exist at bootstrap time.
These changes add similar logic when the router loads a component.
PR Close#63062
In the case that a leave animation is running and someone toggles an `@if`, a new node would be inserted. For a brief moment, there may be two of the same nodes visible at once. While this is expected with native CSS, it's not ideal. Instead, we retain a reference to the leaving element and can remove that node when the new node is entering.
fixes: #63020
PR Close#63048
Any time something causes styles to recalculate,
`element.getAnimations()` will be empty. This updates `animate.enter`
and `animate.leave` to rely on `getComputedStyles` to determine the
longest animation instead.
fixes: #63006
PR Close#63007
Currently when loading external resources in JIT, when `fetch` fails,
the `text` is empty and the component is loading. This hides the actual
underlying fetch error. We should properly detect this and error out.
PR Close#62992
In the case when composing animation classes with `animate.enter` on the
element itself and also with host bindings, the removal would only
have context for one of the classes added: the last one added. This
allows for tracking of the classes added by `animate.enter` via a
WeakMap so we know the exact classes added and which to remove.
Also shores up the tests to make sure we are fully testing animate.enter.
PR Close#62981
Space separated strings, e.g. `class-1 class-2`, should work with both enter and leave animations. `animate.leave` lost that functionality in a refactor. Tests are now added to catch this.
fixes: #62964
PR Close#62979
When animate.leave is used, stylesheet pruning causes issues. Stylesheets with the appropriate animations get pruned before the animations can run. This will delay the removal in the case that the registry is present.
fixes: #62942
PR Close#62943
This adds a test module configuration to define whether animations should be enabled or disabled in test. By default, they are disabled.
PR Close#62764
Currently the HTML parser will stop parsing as soon as it hits an end character in the name of an attribute (e.g. `/` or `>`). This ends up being problematic with some third-party packages like Tailwind which uses a wider range of characters for its class names. While the characters are fine when inside the `class` attribute, our current parser behavior prevents users from setting those classes conditionally through `[class.]` bindings.
These changes adjust the parser to handle such cases.
Fixes#61671.
PR Close#62742
Currently when testing a component using `TestBed.createComponent`, we always create the component as a `div` which isn't aligned with the runtime. The runtime tries to parse out the tag name from the first selector in `@Component` and only falls back to `div` if there isn't one. This behavior difference can cause components to not behave like they would in production which reduces the usefulness of the tests.
These changes add the `inferTagName` option to `TestBed.createComponent` and `TestBed.configureTestingModule` that allows apps to opt into inferring the tag name from the selector in the same way as the runtime. Currently the new option is set to `false`, but we intend to change it to `true` in a future version.
PR Close#62283
