The `FETCH_ERROR` document is used when we are unable to retrieve a
document (except for 404 errors), which includes when there is no
internet connection. Using the `<current-location>` element in the
document's template to show the path of the page we failed to retrieve
assumes that the element's bundle is available (e.g. cached by the SW)
or can be fetched from the server.
When none of these conditions is met, the `DocViewer` is unable to
prepare the document and fails, never showing the `FETCH_ERROR` page to
the user.
Furthermore, the path we are looking to retrieve via
`<current-location>` is essentially the document ID, which we already
have. Thus, loading and instantiating a whole component just for that is
overkill.
This commit addresses both issues by getting rid of the
`<current-location>` component and directly embedding the document ID
into the `FETCH_ERROR` content.
PR Close#27250
When using relative paths for CSS resources (such as background images),
`@angular/cli` will move them to the root `dist/` directory (and update
the paths in CSS accordingly). This results in the SW being unable to
cache the resource, because it is not where it expects it to be.
This commit fixes this issue for the footer background image, by using
an absolute path for the URL. (It also removes an unused style that
would have been affected by the same issue.)
PR Close#27250
In order for 'Material Icons' to work offline, their `.woff2` file needs
to have been cached by the SW. This file is not requested by the
browser, until an element needs to use the icon font.
In order to speed up the initial page load and avoid FOUC, we use
inlined SVGs for all icons in the app shell. As a result, the `.woff2`
file may not be requested, when a user visits angular.io. If they go
offline before visiting a page that does actually use 'Material Icons',
then such icons will not work correctly (e.g. the `error_outline` icon
used in the error page for failed requests due to network
unanvailability).
This commit fixes this, by adding a non-visible element that needs the
'Material Icons' font on the main component. Thids ensures that the
`.woff2` file will always be loaded, even if the page does not use any
material icons.
(Note: The element is inserted lazily to avoid affecting the initial
rendering.)
PR Close#27250
These icons are part of the app shell and used on every load (on both
desktop and mobile). Inlining them ensures they are rendered asap.
PR Close#27250
Previously, we did not load the Roboto font, instead relying on the user
to have it available on their system and falling back to different fonts
otherwise. This resulted in the page being styled slightly differently
for those people that didn't have the font installed locally.
PR Close#27250
This can help with debugging issues, e.g. with the communication between
the preview server and CI, as it gives a better idea of exactly when was
the preview made available and how long it took.
PR Close#27436
In some applications, developers define a `ts_library` that just consists of `d.ts` files (e.g. to type `module.id`; see: https://github.com/angular/material2/blob/master/src/module-typings.d.ts), and expect the `esm5.bzl` file to not throw an error like:
```
target.typescript.replay_params.outputs
struct' object has no attribute 'outputs'
```
The "replay_parameters" property will exist in that case, but is set to "None" because there is no action that should be replayed in favor of producing ES5 outputs. See: https://github.com/bazelbuild/rules_typescript/pull/326. Notice that this right now breaks similarly because an empty `struct()` is returned that does not have a property called `outputs`. [#326](https://github.com/bazelbuild/rules_typescript/pull/326) fixes that by being explicit that there is no _action_ at all.
PR Close#27401
* Currently when building a `ng_module` with Bazel and having the flat module id option set, the flat module files are not being generated because `@angular/compiler-cli` does not properly determine the entry-point file.
Note that this logic is not necessarily specific to Bazel and the same problem can happen without Bazel if multiple TypeScript input files are specified while the `flatModuleIndex` option has been enabled.
PR Close#27200
Currently when building the `ng_package` multiple times, the old `ng_package` output will be copied over to the new `ng_package` content. Resulting in packages like `src/cdk/npm_package/npm_package/npm_package/AND_MORE`.
This happens because currently all TypeScript definition files are resolved from within the `binDir`. This is just wrong because it could then take up the `d.ts` files from the previous `ng_package` output. All typescript definitions that belong to the target package, should be resolved through Bazel and copied based on that computation.
Also fixes that `esm` files aren't written to the `ng_package` on Windows. This is because we try to flatten paths using the `path.delimiter` while the path is always using Posix delimiters (causing the paths to be incorrect)
PR Close#27200
* Currently when building the ES5 and ES2015 output, `ngc_wrapped` will fail because it tries to write the `fs.openSync` the tsickle output file at the same time. This causes a runtime exception in Windows and can be fixed by just writing the externs for ES5 mode to the proper ES5 "output root".
PR Close#27200
* Fixes that `ng_package` does not work generate UMD bundles on Windows because the `esm5/` files are not written to the output directory. This is because `rootDirs` and `rootDir` are posix paths and cause invalid relative paths when mixed with Windows backslash paths.
PR Close#27200
With this change, we no longer depend on CircleCI to trigger the webhook
(which it sometimes does with considerable delay or not at all).
This has the added benefit that other jobs will not unnecessarily
trigger webhooks and spam the preview server logs. It is only the
`aio_preview` job's webhook that we care about.
Related to #27352.
PR Close#27458
Also build releases into a dedicated output_base so you can't
accidentally publish with outdated version stamp.
Bump the version of rules_nodejs so we don't need to create the
symlink_prefixes for the .publish command to work.
PR Close#27362
BREAKING CHANGES:
Bazel users: rules_angular_dependencies() will no longer install transitive dependencies of build_bazel_rules_nodejs and build_bazel_rules_typescript. User WORKSPACE files will now need to install rules_nodejs and rules_typescript transitive deps directly:
```
load("@build_bazel_rules_typescript//:package.bzl", "rules_typescript_dependencies")
rules_typescript_dependencies()
load("@build_bazel_rules_nodejs//:package.bzl", "rules_nodejs_dependencies")
rules_nodejs_dependencies()
```
PR Close#27264
Reported issue in #18138 is due to EasyList being selected in ABP. This commit fixes both the image at the bottom of the Dynamic Component Loader example, and the Stackblitz demo.
Fixes#18138
PR Close#27212
A recent commit (probably 2c7386c) has changed the import graph of the
DI types in core, and somehow results in the ngc compiler deciding to
re-export core DI types from application factories which tangentially
use inject(). This is not really surprising; ngc's import graph can be
very unstable.
However, this results in a re-export of InjectFlags surviving JS
compilation. InjectFlags was a const enum, akin to an interface in TS,
with no runtime repesentation. This causes a warning to be emitted by
Webpack when it sees the re-export of InjectFlags.
This commit avoids the issue by removing 'const' from the declaration
of InjectFlags, causing it to have a runtime value. This is a temporary
fix. The real fix will be for ngc to no longer write exports of const
enums.
Testing strategy: manually verified. Due to the problem only manifesting
when recompiling after a change and then running Webpack, there is no
existing framework via which this could be easily tested with an
integration test. Additionally, the potential for this issue is gone in
Ivy, so this solution is only temporarily needed.
Fixes#27251.
PR Close#27279
Earlier versions may transitively depend on a malicious version of
`flatmap-stream` (see dominictarr/event-stream#116).
The `aio-builds-setup/` had an older version of `event-stream` (3.3.4),
which did not depend on `flatmap-stream`, but upgraded it anyway.
PR Close#27274
