This code all got dumped here during the port. This change deletes tests
that have already **moved** to the aspnet/AspNetCore repo.
Also fixed the markup rewriter tests for components.
\n\nCommit migrated from 86cc904004
Fixes: aspnet/AspNetCoredotnet/aspnetcore-tooling#6184
This gives us parity with the current behavior. Since comments aren't
meaningful in a component/render-tree context, we ignore everything
inside of them for codegen/lowering.
\n\nCommit migrated from 7250dd2624
We need to be able to wire up these features from outside of the Razor
repo. For layering reasons this can't be done in the main Razor
assembly, so it can't be done by default.
\n\nCommit migrated from 386e2b3797
The issue here is that the compiler generates a different syntax tree
for `data-` attributes :(. I've added a note to dotnet/aspnetcore-tooling#6112 to follow up on
the compiler behavior. We used to parse data- attributes differently for
components, but I really don't think having the compiler special case
them makes sense.
For now I've added a workaround here to lower `data-` attributes in a
fashion similar to other dynamic attribute values.
\n\nCommit migrated from f5109fbfec
The bug here is that we have extra whitespace being written out when the
component renders. This happens because parser adds an empty C#
statement node to wrap all of the whitespace that precedes `@functions`.
This is tracked by aspnet/AspNetCoredotnet/aspnetcore-tooling#6207.
To fix this issue I'm allowing the whitespace pass to skip over an empty
C# statement to remove whitespace. This is sound because an empty C#
statement produces no output - we allow removing whitespace before and
after all nodes that produce no output.
\n\nCommit migrated from e68d93f458
This change contains the enabling features to use Razor Components in a
class library. Currently we require a few workarounds, see the test
project file. This is good enough to get things unblocked.
One part that was needed was to register the correct component features
in the rzc. This is a good example, of why we don't like to add new
features that get registered conditionally, it's error-prone :)
The other part that was needed was to make some of the MVC-related
features for assembly attributes conditional on the TFM. We need to be
able to use the 3.0 (all inclusive) SDK, but without the MVC-related
features. This isn't the right heuristic, but it gets us unblocked.
\n\nCommit migrated from 8dbf02076b
This change ports the existing Blazor component compiler features to
work on top of the Razor compiler. These new features are on once Razor
Language version is set to V3 or higher.
The tests and code for the old Blazor compiler were added to this repo
in previous changes. This change adds the necessary enabling features,
and gets the tests working.
Code generation for tests tests is identical with a few exceptions.
Design time directives now include line mappings (as they should). The
infrastructure seems to prefer windows style line-endings whereas the
Blazor code did the opposite. I plan to leave this alone unless it's a
problem for testing on *nix.
IR Generation will be cosmetically different due to different node names
and formatting. Structure should generally be the same.
----
In order to implement this support, I had to add the following core
features.
Configuring passes/engine-features based on the langauge version. This
required revisiting a *lot* of old tests and forcing them to use a
version appropriate Razor engine. A lot of our tests for legacy things
were using new versions of the language :(.
Configuring passes/engine-features to switch on/off depending on the
document kind.
Configuring directives depending on the file-kind. Directives are a
parser feature, so this decision needs to be made before the IR phase.
Directives that are specific to a file kind will override directives
that are not - this allows components to customize templates for
instance.
Adding functions/inherits/implements directive as mainline features that
are always on in V3. I'm not quite motivated to do this yet for
typeparam, but we might end up there. Remember that directives are
reserved words in Razor.
\n\nCommit migrated from a04addb86f
This change cleans up the intermediate node types that we have so far
for components and changes the naming, visibility, etc to match our
existing nodes.
The next change will integrate these types with the IR lowering process.
\n\nCommit migrated from 34cf523c0d
This change adds all of the code from the Blazor/Components compiler into
the Razor language assembly. Minimal refactoring or integration work has
been done, that is next :)
All of the code compiles and all unit tests pass, except the one I had
to skip until integration work is done.
\n\nCommit migrated from aa5c82ca5e
This is a bit of a rework of how we initially set this up, but with more
forethought to how this will work in the project system. I have not yet
surfaced this through VS.
My immediate next step is to light up the component integration tests
and something like this is on the critical path to get that work since
we need a way to specify in tests that a document should be treated as
a component.
\n\nCommit migrated from 6b81da3f02
This is a dump of the existing component codegen tests from
the Blazor effort.
I'm also adding the minimal Component types as a new shim to get things
to compile. This is similar to how our shims work already for MVC.
The next step will be add port functionality until the tests pass.
\n\nCommit migrated from 49f49dd9f0
- This happens because a completion session starts when we've rendered a marker directive token without a line pragma and then we return our parse results which provide a C# buffer with a line pragma. This pull request fixes that.
- Updated tests and verified in VS d16.0stg.
aspnet/AspNetCoredotnet/aspnetcore-tooling#4810
\n\nCommit migrated from ca593c80f7
- Prior to this we were unconditionally adding the `DotNetCoreRazorConfiguration` capability resulting in 1.0 projects looking like 2.1+ projects.
- Added a project capability test that verifies we're maintaining project capabilities for netcoreapp3.0 projects.
- Added a project capability test that verifies we don't muck with 1.1 project capabilities.
aspnet/Razordotnet/aspnetcore-tooling#2751
\n\nCommit migrated from 8d733ecc04
- When we detect a design time build we force tooling to use 2.1
- In runtime builds (i.e. rzc) it still uses the 3.0 configuration for 3.0 apps.
(cherry picked from commit dotnet/aspnetcore-tooling@0326abe530)
\n\nCommit migrated from a6dc88e455
When a user builds a 2.1 or 2.2 project with a 3.0 they can get an error
due to invalid command line options from rzc. What happens is that the
3.0 tasks (which are a singleton) are talking to the 2.X build tool
(which is NOT a singleton).
Recently some changes were make to the 3.0 tasks which caused this bug
because it did not properly accomodate downlevel versions of rzc. We
have an integration test for this scenario for but the regression was
not detected, so obviously something is wrong with the test setup. I'll
investigate that separately.
(cherry picked from commit dotnet/aspnetcore-tooling@1aace2b6c7)
\n\nCommit migrated from d2654d0847
Ajay Bhargav Baaskaran
Ryan Nowak
Ajay Bhargav Baaskaran
N. Taylor Mullen
Ryan Nowak