We are seeing the following Stack Trace in our logs. We are currently using LaunchDarkly.EventSource version 3.2.3

Class: System.Net.WebException Message: The operation has timed out. at System.Net.HttpWebRequest+<RunWithTimeoutWorker>d__241'1[T].MoveNext () <0x101b68a10 + 0x00498> in <8294fc839f2d4b799a08e766e2dfa68e#9da8dd95652572baf7b138c1bdc40608>:0 --- End of stack trace from previous location where exception was thrown --- at System.Net.WebResponseStream+<ReadAsync>d__48.MoveNext () <0x101b13b90 + 0x00a2c> in <8294fc839f2d4b799a08e766e2dfa68e#9da8dd95652572baf7b138c1bdc40608>:0 --- End of stack trace from previous location where exception was thrown --- at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess (System.Threading.Tasks.Task task) <0x1016cc650 + 0x000d3> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification (System.Threading.Tasks.Task task) <0x1016cc5a0 + 0x0008b> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at System.Runtime.CompilerServices.TaskAwaiter.ValidateEnd (System.Threading.Tasks.Task task) <0x1016cc530 + 0x00053> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at System.Runtime.CompilerServices.ConfiguredTaskAwaitable'1+ConfiguredTaskAwaiter[TResult].GetResult () <0x1017a5c40 + 0x0001b> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at System.IO.StreamReader+<ReadBufferAsync>d__98.MoveNext () <0x101619770 + 0x0046f> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 --- End of stack trace from previous location where exception was thrown --- at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess (System.Threading.Tasks.Task task) <0x1016cc650 + 0x000d3> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification (System.Threading.Tasks.Task task) <0x1016cc5a0 + 0x0008b> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at System.Runtime.CompilerServices.TaskAwaiter.ValidateEnd (System.Threading.Tasks.Task task) <0x1016cc530 + 0x00053> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at System.Runtime.CompilerServices.ConfiguredTaskAwaitable'1+ConfiguredTaskAwaiter[TResult].GetResult () <0x1017a5c40 + 0x0001b> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at System.IO.StreamReader+<ReadLineAsyncInternal>d__61.MoveNext () <0x101618060 + 0x001eb> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 --- End of stack trace from previous location where exception was thrown --- at System.Runtime.CompilerServices.TaskAwaiter.ThrowForNonSuccess (System.Threading.Tasks.Task task) <0x1016cc650 + 0x000d3> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification (System.Threading.Tasks.Task task) <0x1016cc5a0 + 0x0008b> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at System.Runtime.CompilerServices.TaskAwaiter.ValidateEnd (System.Threading.Tasks.Task task) <0x1016cc530 + 0x00053> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at System.Runtime.CompilerServices.TaskAwaiter'1[TResult].GetResult () <0x1016cc910 + 0x0001b> in <3c7b99a36820490fb2cbc5a6fc6b06d8#9da8dd95652572baf7b138c1bdc40608>:0 at LaunchDarkly.EventSource.EventSourceStreamReader+<ReadLineAsync>d__2.MoveNext () <0x1041d44c0 + 0x00243> in <641c2bc30eae4c1d8d3db4b93ea5a14f#9da8dd95652572baf7b138c1bdc40608>:0

HI,

Our .NET client application has started throwing the following error all of a sudden (in the last few days).

System.ObjectDisposedException: Cannot access a disposed object.
Object name: 'SslStream'.
at System.Net.ConnectStream.EndRead(IAsyncResult asyncResult)
at System.Threading.Tasks.TaskFactory1.FromAsyncTrimPromise1.Complete(TInstance thisRef, Func3 endMethod, IAsyncResult asyncResult, Boolean requiresSynchronization) --- End of stack trace from previous location where exception was thrown --- at System.Runtime.ExceptionServices.ExceptionDispatchInfo.Throw() at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task) at System.Net.Http.HttpClientHandler.WebExceptionWrapperStream.<ReadAsync>d__4.MoveNext() --- End of stack trace from previous location where exception was thrown --- at System.Runtime.ExceptionServices.ExceptionDispatchInfo.Throw() at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task) at System.IO.StreamReader.<ReadBufferAsync>d__97.MoveNext() --- End of stack trace from previous location where exception was thrown --- at System.Runtime.ExceptionServices.ExceptionDispatchInfo.Throw() at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task) at System.IO.StreamReader.<ReadLineAsyncInternal>d__60.MoveNext() --- End of stack trace from previous location where exception was thrown --- at System.Runtime.ExceptionServices.ExceptionDispatchInfo.Throw() at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task) at System.Runtime.CompilerServices.TaskAwaiter1.GetResult()
at LaunchDarkly.EventSource.EventSourceStreamReader.d__2.MoveNext()

I'm configuring an LdClient as described in https://docs.launchdarkly.com/sdk/server-side/dotnet/migration-5-to-6#understanding-changes-to-networking:

var config =  Configuration.Builder("my-sdk-key")
    .Http(
        Components.HttpConfiguration()
            .Proxy(new System.Net.WebProxy(new Uri("http://my-proxy:8080")))
    )
    .Build();

However, when EventSource creates its HttpClient, it doesn't use the Http/Proxy configuration, resulting in errors. From the log:

INFO  LaunchDarkly.Sdk.DataSource Connecting to LaunchDarkly stream
INFO  LaunchDarkly.Sdk Waiting up to 5000 milliseconds for LaunchDarkly client to start...
DEBUG LaunchDarkly.Sdk.DataSource Making GET request to EventSource URI https://stream.launchdarkly.com/all
WARN  LaunchDarkly.Sdk Timeout encountered waiting for LaunchDarkly client initialization

The only workaround we've found is to set the global proxy as per instructions here: https://docs.microsoft.com/en-us/dotnet/framework/network-programming/proxy-configuration

.NET 4.7.2
LaunchDarkly.ServerSdk 6.1.0
LaunchDarkly.EventSource 4.1.2
LaunchDarkly.CommonSdk 5.1.0

Thank you for fixing the framework references in the main nuget packages. Can you also fix the ones in this project?

dotnet-eventsource/src/LaunchDarkly.EventSource/LaunchDarkly.EventSource.csproj

Assembly 'Assets/netstandard2.0/LaunchDarkly.EventSource.dll' will not be loaded due to errors:
Unable to resolve reference 'LaunchDarkly.Logging'. Is the assembly missing or incompatible with the current platform?
Reference validation can be disabled in the Plugin Inspector.

To reproduce, initialize .net-client 3.4.0-alpha1 with a low read timeout (less than 3 minutes). Debug output does not show the client reconnecting to the stream after every read timeout interval is reached, and a heartbeat is received from the flag stream after 3 minutes.

The dependency on Polly-Signed has been problematic in our project, which uses Polly (the not-signed version). We've had to change from Polly to Polly-Signed in each of our projects, which is a bit of a hassle.

It looks like there's just one retry policy used here; can I suggest that the retry logic be written manually and the dependency on Polly-Signed removed? That way this library would have less of an impact on client projects.

Hi,

this is not an issue, rather a question. Also, it's not related to the .net client. I will have to excuse in advance for not finding the right repository to ask.
The topic is related to your server side SSE technology, specifically how your server handles disconnections.

When a client close, it should terminate the TPC connection using a FIN. It's not uncommon for a client connection to drop without closing correctly the connection (this is especially true on mobile devices)
How do you handle "broken" connections on your server streaming architecture? Are you using TPC keepAlive to detect unresponsive clients?

Thanks in advance.

https://www.nuget.org/packages/LaunchDarkly.EventSource/2.2.1

I've tried open a SSE stream from a blazor client.
the call ReadAsStreamAsync on the http request don't exit creating a deadlock.

I've found that the solution is calling this extension on the http request object:
request.SetBrowserResponseStreamingEnabled(true);
which is a method extension on the Microsoft.AspNetCore.Components.WebAssembly assembly.

I believe this assembly cannot be added as reference to the source code.
A solution could be to expose a method hook from the library before execute the http request to allow the blazor code to invoke the extension method.

After adding LaunchDarkly.ServerSdk version 7.0.3 (which transitively pulled in LaunchDarkly.EventSource 5.0.1) to an ASP.NET MVC app targeting .NET Framework 4.7.2, requests to the app returned this error:

Could not load file or assembly 'LaunchDarkly.Logging, Version=1.0.1.0, Culture=neutral, PublicKeyToken=d9182e4b0afd33e7' or one of its dependencies. The located assembly's manifest definition does not match the assembly reference.

It seems that this is because LaunchDarkly.ServerSdk depends on LaunchDarkly.Logging 2.0.0, while LaunchDarkly.EventSource depends on LaunchDarkly.Logging 1.0.1. To resolve this, I had to add a binding redirect to the Web.config:

<dependentAssembly>
  <assemblyIdentity name="LaunchDarkly.Logging" publicKeyToken="d9182e4b0afd33e7" culture="neutral"/>
  <bindingRedirect oldVersion="0.0.0.0-2.0.0.0" newVersion="2.0.0.0"/>
</dependentAssembly>

Is this expected behavior for this package? As far as I can tell, EventSource is the only installed package that references Logging version 1.0.1. We will be adding LaunchDarkly to more apps in the future, and it would be nice if we did not have to include a binding redirect every time.

Error trying to initialize LDClient with SDK key.
Error:

System.MethodAccessException: Attempt by method 'LaunchDarkly.Common.StreamManager..ctor(LaunchDarkly.Common.IStreamProcessor, LaunchDarkly.Common.StreamProperties, LaunchDarkly.Common.IBaseConfiguration, LaunchDarkly.Common.ClientEnvironment, EventSourceCreator)' to access method 'LaunchDarkly.EventSource.ExponentialBackoffWithDecorrelation..ctor(System.TimeSpan, System.TimeSpan)' failed..

Using packages:

• LaunchDarkly.Client v5.6.3
• LaunchDarkly.EventSource v4.0.0

Even though the retry field is read it seems like it is not really used for the reconnection time.

Here the field is read to _retryDelay:

But _retryDelay is not really used for the reconnection time:

Instead it is only passed initially for construction of ExponentialBackoffWithDecorrelation:

The _minimumDelay in ExponentialBackoffWithDecorrelation is never updated

It would be great if the EventSource.OnMessageReceived-method, EventSource.DispatchEvent (or an intermediary) could be made virtual so that custom message processing could be implemented by inheriting from the EventSource-class and overriding this method.

As a concrete example, I'd like to extend the class with a method for adding and removing listeners for specific events (which may, in itself, be a nice additional feature), for which I'd need to hook into the message handling loop. However, at present, I can only do so by subscribing to the MessageReceived-event.

Message: Object reference not set to an instance of an object.
Type: System.NullReferenceException

   at LaunchDarkly.EventSource.EventSource.ProcessResponseContent(String content)
   at LaunchDarkly.EventSource.EventSourceService.<ProcessResponseFromReaderAsync>d__13.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.ExceptionServices.ExceptionDispatchInfo.Throw()
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at LaunchDarkly.EventSource.EventSourceService.<ConnectToEventSourceApi>d__11.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.ExceptionServices.ExceptionDispatchInfo.Throw()
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at LaunchDarkly.EventSource.EventSourceService.<GetDataAsync>d__10.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.ExceptionServices.ExceptionDispatchInfo.Throw()
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at LaunchDarkly.EventSource.EventSource.<ConnectToEventSourceAsync>d__37.MoveNext()
--- End of stack trace from previous location where exception was thrown ---
   at System.Runtime.ExceptionServices.ExceptionDispatchInfo.Throw()
   at System.Runtime.CompilerServices.TaskAwaiter.HandleNonSuccessAndDebuggerNotification(Task task)
   at LaunchDarkly.EventSource.EventSource.<StartAsync>d__32.MoveNext()

Occurs around the same time tasks are cancelled:

Due to the support of .NET Standard 1, there is a security vulnerability in the LaunchDarkly dependency chain that is rooted in EventSource. This is being picked up by SAST scanners and we are currently compelled to use rule exceptions as there is no workaround to force a package version that does not contain the vulnerability.

Ideally there would be a version of EventSource that did not support .NET Standard 1 that would either be referenced in the chain by LaunchDarkly.Common.StrongName, or simply available to override.

The result of scanning, which I believe uses dotnet retire:

* Microsoft Security Advisory CVE-2018-8416: .NET Core Tampering Vulnerability in System.IO.Compression.ZipFile/4.3.0
https://github.com/dotnet/announcements/issues/95
LaunchDarkly.EventSource/3.3.2
╚ NETStandard.Library/1.6.1
 ╚ System.IO.Compression.ZipFile/4.3.0

* Microsoft Security Advisory CVE-2018-8292: .NET Core Information Disclosure Vulnerability in System.Net.Http/4.3.0
https://github.com/dotnet/announcements/issues/88
LaunchDarkly.EventSource/3.3.2
╚ NETStandard.Library/1.6.1
 ╚ System.IO.Compression.ZipFile/4.3.0

Hello,

In using your library to process SSE events, I'm finding that as millions of events are received there can be a significant buildup in RAM. I ran it through dotMemory to profile and found that the leak appears to be due to the combined token in DoWithTimeout (AsyncHelpers.cs). The combined token is used but never disposed. Adding a combinedCancellation.Dispose() after the await task seems to cure the problem.

Thanks!

Hello.
I followed your tutorial here : #91

But I have an error

"Misused header name. Make sure request headers are used with HttpRequestMessage, response headers with HttpResponseMessage, and content headers with HttpContent objects."

public class testSEE : MonoBehaviour
{
    // found here : https://github.com/launchdarkly/dotnet-eventsource/issues/91
    private void Start()
    {
        string JSON = @"{'model':'gpt-3.5-turbo','stream':true,'messages': [{""role"": ""user"", ""content"": ""Say this is a test!""}]}";

        Configuration config = Configuration.Builder(new Uri("https://api.openai.com/v1/chat/completions"))
            .RequestHeader("Content-Type", "application/json")
            .RequestHeader("Authorization", "Bearer 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")
            .RequestBody(JSON, "application/json")
            .Build();
        EventSource eventSource = new EventSource(config);

        Debug.Log("Event source is configurated");

        eventSource.MessageReceived += EventSource_MessageReceived;
        eventSource.Closed += EventSource_Closed;
        eventSource.Opened += EventSource_Opened;
        eventSource.Error += EventSource_Error;
        Task.Run(() => eventSource.StartAsync());
    }

    private void EventSource_Error(object sender, ExceptionEventArgs e)
    {
        Debug.LogException(e.Exception);
        Debug.LogError($"error : {e.Exception.Message}");
    }

    private void EventSource_Opened(object sender, StateChangedEventArgs e)
    {
        Debug.Log($"Opened: {sender}, {e}");
    }

    private void EventSource_MessageReceived(object sender, MessageReceivedEventArgs eventArgs)
    {
        var eventName = eventArgs.Message.Name;
        var eventData = eventArgs.Message.Data;
        Debug.Log(eventData);
        Debug.Log(eventName);
        // do whatever you want to do with the message
    }

    private void EventSource_Closed(object sender, StateChangedEventArgs e)
    {
        Debug.Log($"close: {sender}, {e}");
    }
}

Do you know how I can fix this ?

Steps to reproduce:

1. run the server-side php script which will close a connection every time after it sends the data:

<?php
header('Content-Type: text/event-stream');
header('Cache-Control: no-cache'); // recommended to prevent caching of event data.

/**
 * Constructs the SSE data format and flushes that data to the client.
 *
 * @param string $id Timestamp/id of this connection.
 * @param string $msg Line of text that should be transmitted.
 */
function sendMsg($id, $msg) {
  echo "id: $id" . PHP_EOL;
  echo "data: $msg" . PHP_EOL;
  echo PHP_EOL;
  ob_flush();
  flush();
}

$serverTime = time();

sendMsg($serverTime, 'server time: ' . date("h:i:s", time()));

2. run the console app:

        static async Task Main(string[] args)
        {
            var config = Configuration.Builder(new Uri("http://localhost:8000/your_script.php"))
                .ConnectionTimeout(TimeSpan.FromSeconds(1))
                .DelayRetryDuration(TimeSpan.FromSeconds(3))
                .ReadTimeout(TimeSpan.FromMinutes(4))
                .Build();

            var evt = new EventSource(config);

            evt.Opened += (sender, eventArgs) => { Console.WriteLine("Opened"); };
            evt.Error += (sender, eventArgs) => { Console.WriteLine("Error"); };
            evt.CommentReceived += (sender, eventArgs) => { Console.WriteLine("CommentReceived"); };
            evt.MessageReceived += (sender, eventArgs) => { Console.WriteLine($"MessageReceived: {eventArgs.Message.Data}"); };
            evt.Closed += (sender, eventArgs) => { Console.WriteLine("Closed"); };

            try
            {
                await evt.StartAsync();
            }
            catch (Exception ex)
            {
                Console.WriteLine("Closed");
            }

            Console.ReadLine();
        }

3. There should be a delay = 3 seconds between reconnects, but EventSource reconnects immediately. It looks like it ignores DelayRetryDuration(...) setting.

My workaround is to change some code:

• in ExponentialBackoffWithDecorrelation.cs:

        public TimeSpan GetNextBackOff(TimeSpan retryDelay)
        {
            _reconnectAttempts++;
            return retryDelay;
        }

• in EventSource.cs:

        public async Task StartAsync()
        {
            var cancellationToken = _pendingRequest.Token;
            while (!cancellationToken.IsCancellationRequested)
            {
                await MaybeWaitWithBackOff();
                try
                {
                    await ConnectToEventSourceAsync(cancellationToken);
                    _backOff.ResetReconnectAttemptCount();
                }
                catch (Exception e)
                {
                    _logger.ErrorFormat("Encountered an error connecting to EventSource: {0}", e, e.Message);
                    _logger.Debug("", e);
                }
            }
        }

        private async Task MaybeWaitWithBackOff()  {
            if (_backOff.GetReconnectAttemptCount() > 0 && _retryDelay > TimeSpan.FromMilliseconds(0))
            {
                TimeSpan sleepTime = _backOff.GetNextBackOff(_retryDelay);
                _logger.InfoFormat("Waiting {0} milliseconds before reconnecting...", sleepTime.TotalMilliseconds);
                BackOffDelay = sleepTime;
                await Task.Delay(sleepTime);
            }
            else {
                _backOff.IncrementReconnectAttemptCount();
            }
        }

(I think the main problem is calling MaybeWaitWithBackOff(); without await.)

I noticed that even though the cancellation token is being set the stream itself is not actually closing when cancelled which is causing the socket and the stream to stay open. To resolve the issue I made the following code changes so that it will close the stream properly.

EventSourceServices.cs line 101
FROM
using (var stream = await response.Content.ReadAsStreamAsync().ConfigureAwait(false))
TO
await using (var stream = await response.Content.ReadAsStreamAsync(CancellationToken.None).ConfigureAwait(false))
await using (cancellationToken.Register(stream.Close)) // Calls stream.close when token has been cancelled

and the above code then caused an exception with the StreamReader. to resolve that issue modify lines 147-151
FROM
if (line == null)
{
// this means the stream is done, i.e. the connection was closed
return;
}
TO
if (line == null)
{
reader.Dispose(); // <----- this will prevent a premature end of stream exception
// this means the stream is done, i.e. the connection was closed
return;
}