- Removed all design time descriptors and put their API surface into their corresponding descriptor. Part of removing the design time API surface was removing the tracking of `<Remarks>`, it wasn't used so there's on need to track it until we need it.
- Removed the Type requirement from `TagHelperDescriptor`. With this separation we'll be able to have abstract `TagHelper`s that aren't based on a class implementation.
- Removed Prefix from the `TagHelperDescriptor` API surface. It was a legacy requirement based on how the Razor parser was put together. We can work around this now.
- Stripped correlation information from the immediate `TagHelperDescriptor` API surface. Instead this information is now tracked in `TagMatchingRule`s. This change means that you will not have multiple `TagHelperDescriptor`s per `TagHelper`; instead it's all tracked in a single descriptor. A side effect of this change was the transformation of `IsIndexer` => 3 new properties.
- Renamed many descriptor types and property names.
- Added builder APIs to construct TagHelpers since they're inherently immutable in their API surface.
- Added `ITagHelperDescriptorBuilder` to represent `TagHelper`s that are built from an `ITagHelper` implementing class. It re-introduces the `TypeName` association of a `TagHelper`.
- Added `ITagHelperBoundAttributeDescriptorBuilder` to represent that an attribute was associated with a property.
- Added validation methods to the descriptor builders to enable consumers to validate the current state of the builder and add diagnostics as necessary.
- Moved descriptors away from RazorError.
- Updated the various comparers to understand the descriptors new API.
- Added a new `RazorDiagnosticFactory` abstraction to handle `RazorDiagnostic`s and their corresponding errors/ids etc. This new API should allow for easy addition of new `RazorDiagnostic` errors.
- Updated the `DefaultTagHelperDescriptorFactory` to construct `TagHelperDescriptor`s using the new builder APIs and in the new descriptor format (1 descriptor per type).
- Updated `ViewComponentTagHelperDescriptorFactory` to construct `TagHelperDescriptor`s with the builder API.
- With both factory implementations code was duplicated because the ViewComponent work will be moving outside of Razor once we have the proper hooks.
- Updated `TagHelper` binding bits to capture a binding result in order to query which rules appy to a given tag name.
Addressed feedback
- Update tests to react to new `TagHelperDescriptor` API.
- Remove case sensitive comparers and some cleanup
- Added TagHelperDescriptorJsonConverter, RazorDiagnosticJsonConverter and added serialization tests
We had a startTask we would await in DisposeAsync and channel completion continuation. This task would be initially set to a completed task and then once StartAsync was invoked it would be replaced with the actual task representing StartAsyncInternal. However if a transport failed immediately after starting the channel completion continuation could have been called before the StartAsyncInternal method completed. In this case we would await the inital completed task and then very likely would fail trying to await _receiveLoop task because it wouldn't necessarily be set.
The fix is to use TaskCompletionSource so we don't try to swap tasks. We need to do some additional state checks because:
- the TaskCompletionSource task may never be completed (e.g. DisposeAsync is being called without starting the connection)
- TaskCompletionSource allows setting the result only once and we should not return its task more than once (e.g. calling StartAsync after connection was successfully started and stopped)
Fixes#304
This change addressed a race condition in the ObjectMethodExecutor where
the default argument values array can become visible before it is
initialized. If a second observer accesses the array while it is being
initialized, it can observe a null value for a reference type parameter,
leading to a nullref.
The fix here is to make everything immutable and initialize it all up
front. There's no reason to create an OME without eventually running it,
so there's no downside to doing the initialization up front.