* Disable response buffering via the IHttpBufferingFeature
- To make sure SignalR works with servers and middleware
that do perform response buffering, disable it via the
IHttpBufferingFeature for SSE.
- Added test to verify buffering is disabled
- Today we don't end the request if the application completes
but the websocket transport hasn't gotten the receive frame as yet.
This changes adds a WebSocketOptions.CloseTimeout to EndPointOptions
that allows configuring this timeout. When the timeout is reached, we abort
the connection and end the transport task so that the request can end.
- Added tests for websocket timeout and skipped tests for application timeouts
* Fix issue where multiple calls to dispose don't wait properly
- DisposeAsync returned immediately to anyone but the first caller.
This means that it was possible to end the request before properly
waiting on the transport task which means writing after dispose was possible.
- Added a test
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
Fixing a race where removing the last message from the channel would trigger draining the event queue (in channel Completion continuation) which would prevent from raising the Received event for the very message that was read from the channel.
- Creating an MSBuild project for the TS client
- Adding project references to the TS client project from projects that need the client - (ensures the correct targets dependency graph and prevents building the client multiple times and related races)
- Removing gulp tasks from individual projects (allows containing npm only in the TS client source and node tests)
- Using incremental compilation to build the TS client only when inputs change (prevents building the client multiple times or when not needed at all)
- Removing `npm install` from all the projects (takes up to 10 seconds even if there is nothing to restore) - npm packages will still be installed when running full build (if needed) or need to be installed manually
* Fixed parameter count mismatch when invoking methods with wrong case
- Hub methods were being tracked with 2 dictionaries, one for parameter names
the other for callbacks. This change introduces a single dictionary that stores
the hub name to a HubMethodDescriptor. That descriptor stores the parameter types
and method info for the bound hub method.
- The callback is now just an invoke method on the HubEndPoint itself.
- Added tests for case sensitivity in hub method names
- It was possible for the application to be torn down during
a background scan. When that happened, the timer would be disposed
before the end of the scan and would throw an ObjectDisposedException
when timer resumption happened. This change introduces a lock that avoids that
race.
Fixes#161
* Handle misbehaving user code
- Execute EndPoint logic on a threadpool thread
- Turn synchronous exceptions into async ones to unify the
error handling
- Added tests
- The connection state object is manipulated by multiple parties in a non thread safe way. This change introduces a semaphore that should be used by anyone updating or reading the connection state.
- Handle cases where there's an active request for a connection id and another incoming request for the same connection id, sse and websockets 409 and long polling kicks out the previous connection (https://github.com/aspnet/SignalR/issues/27 and https://github.com/aspnet/SignalR/issues/4)
- Handle requests being processed for disposed connections. There was a race where the background thread could remove and clean up the connection while it was about to be processed.
- Synchronize between the background scanning thread and the request threads when updating the connection state.
- Added `DisposeAndRemoveAsync` to the connection manager that handles`DisposeAsync` throwing and properly removes connections from connection tracking.
- Added Start to ConnectionManager so that testing is easier (background timer doesn't kick in unless start is called).
- Added RequestId to connection state for easier debugging and correlation (can easily see which request is currently processing the logical connection).
- Added tests
* Clean up disposal of connection state
- Removed IDisposable and added a DisposeAsync method to ConnectionState
- Added ApplicationTask and TransportTask to ConnectionState as first class
properties so that it is easy to see (in a process dump or debugger) the
outstanding tasks that Sockets is keeping track of on a per connection basis.
* Allow processing of other incoming invocations in parallel
- Don't wait on the response of an invocation to pick up
the next message from the channel.
- Unhandled exceptions should continue bubbling up correctly
* rename getid to negotiate
* also change SSE and Long Polling to require a pre-established connection
* disallow changing transports mid-connection; return a 400 response if the user attempts to do so
* Use TryRead and TryWrite
- Use TryWrite to avoid errors on channel close for /send requests
- Use TryRead until it returns false for all transports but long polling
- Remove Streaming* classes from Sockets. The main
API will be channels based and streaming transports
will use the PipelineChannel (formerly FramingChannel) to
access messages.
- Added WriteAsync and ReadAsync to Connection and hid
the IChannelConnection from public API.
- Also fixed the fact that unknown methods caused server side
exceptions.
- Changed the consumption pattern to WaitToReadAsync/TryRead to avoid
exceptions.
- React to API changes
- Fixed ChannelConnection to use IChannel<T> for
both sides of the connection. This allows use to close both the
input and the output when we are tearing down.
- Use TryComplete instead of complete to avoid exceptions thrown on
Complete(), particularly ChannelClosedException.
* Need a separate set of primitives to handle messaging
* Using Channels (not Pipelines!) to provide the data flow for messaging
* All transports are now "message" based transports
* Added an adaptor to convert message-based transports to serve
streaming endpoints
Making sure that OnConnected/OnDisconnected events are invoked correctly (e.g. if invoking OnDisconnectedAsync on hub threw we would not call OnDisconnectedAsync on lifetime manager and therefore we would continue to use/track connections that were already closed)
David Fowler
Kai Ruhnau
Alexis Nowikowski
BrennanConroy
Andrew Stanton-Nurse
Pawel Kadluczka
moozzyk
Nate McMaster
Mikael Mengistu
Pranav K
Nate McMaster
Dan Kirkham
Andrew Stanton-Nurse
Stephen Halter