This lets users continue using runtime-sideeffect Decorators if they choose,
only down-leveling the marked ones to Annotations.
Also remove the "skipTemplateCodegen" option, which is no longer needed
since Angular compiles with tsc-wrapped rather than ngc. The former doesn't
include any codegen.
* feat(I18nExtractor): Add file paths to error messages
relates to #9071
* feat(i18n): allow i18n start comments without meaning
* refactor(i18n): cleanup
* test(HtmlParser): Add depth to expansion forms
Previously these symbols were exposed via platform-browser-dynamic, then we merged then into platform-browser
thinking that tools would know how to shake off the compiler and other dynamic bits not used with the offline
compilation flow. This turned out to be wrong as both webpack and rollup don't have good enough tree-shaking
capabilities to do this today. We think that in the future we'll be able to merge these two entry points into
one, but we need to give tooling some time before we can do it. In the meantime the reintroduction of the -dynamic
package point allows us to separate the compiler dependencies from the rest of the framework.
This change undoes the previous breaking change that removed the platform-browser-dynamic package.
This aligns the configuration of platform pipes / directives with offline compilation.
BREAKING CHANGE:
- `PLATFORM_PIPES` and `PLATFORM_DIRECTIVES` now are fields on `CompilerConfig`.
Instead of providing a binding to these tokens, provide a binding for `CompilerConfig` instead.
This addresses several oversights in assigning security contexts to DOM schema
elements found by our security reviewers (thanks!).
This also adds some more precise unit tests for the interaction between
(Dom)ElementSchemaRegistry and the TemplateParser, and extracts the security
specific parts into dom_security_schema.ts.
Comparison of (potentially) dangerous property names is done case insensitive,
to avoid issues like formAction vs formaction.
Part of issue #8511.
* fix(compiler): throw an error if variable with the same name is already defined. Closes#6492
* fix(compiler): Clean up formatting for issue #6492
* fix(compiler): throw an error if reference with the same name is already defined.
Closes#6492
<!-- xxxx ------->
The issue came from a lack of support for backtracking on string
matching.
The way it is done, if the "end pattern" for consumeRawText starts with
twice the same character, you end up having problem when your string
being parsed has 3 times this char
Example
End string: xxyz
string to parse: aaaaaaxxxyz
calling consumeRawText(false, 'x', attemptParseStr('xyz')) would fail
Closes#7119
Victor Berchet
Pawel Kozlowski
Matias Niemelä
Alex Eagle
Igor Minar
Tobias Bosch
ScottSWu
Martin Probst
laco0416
Suguru Inatomi
Andrii Nechytailov
Igor Minar
Wesley Cho
Grégory Bataille
Tero Parviainen
Ward Bell
Vamsi Varikuti
alexcastillo
Marc Laval