Fix console output problem, dotnet test appeared to hang #102
Conversation
|
|
||
| namespace Akka.MultiNode.TestAdapter.Internal.Sinks | ||
| { | ||
| public class FrameworkHandleMessageSinkActor: TestCoordinatorEnabledMessageSink |
There was a problem hiding this comment.
New message sink for IFrameworkHandle
There was a problem hiding this comment.
Can you add an XML-DOC comment here to explain this?
| } | ||
| }; | ||
| opt.OutputDataReceived = OutputHandler; | ||
| opt.ErrorDataReceived = OutputHandler; |
There was a problem hiding this comment.
Capture error outputs from test processes
| @@ -49,8 +52,14 @@ public TcpLoggingServer(IActorRef sinkCoordinator) | |||
|
|
|||
| Receive<Tcp.Received>(received => | |||
| { | |||
| var message = received.Data.ToString(); | |||
| sinkCoordinator.Tell(message); | |||
| if (received.Data.Count >= BufferSize) | |||
There was a problem hiding this comment.
This bug pertains to a lot of "unknown message" in our test. Although we don't have any limit on TCP message size, our buffer size is by default limited to 512 bytes, and since we pump messages as soon as we get them, if the string message is longer than 512 characters, the message will be chopped up. This happens a lot with FAIL-EXCEPTION messages since the stack trace can be quite big.
There was a problem hiding this comment.
Can we increase the message size used by the MNTR in Akka.IO.Tcp?
| @@ -112,15 +112,15 @@ public static MultiNodeTestRunnerMessageType DetermineMessageType(string message | |||
| var matchRunnerLog = RunnerLogMessageRegex.Match(messageStr); | |||
| if (matchRunnerLog.Success) return MultiNodeTestRunnerMessageType.RunnerLogMessage; | |||
|
|
|||
| var matchFailureReason = NodeFailureReasonRegex.Match(messageStr); | |||
There was a problem hiding this comment.
Failure with exception need to be tested first, else the pass test will eat the exception messages because it contains the word "FAIL"
|
|
||
| namespace Akka.MultiNode.TestAdapter.Internal.Sinks | ||
| { | ||
| public class FrameworkHandleMessageSinkActor: TestCoordinatorEnabledMessageSink |
There was a problem hiding this comment.
Can you add an XML-DOC comment here to explain this?
| @@ -49,8 +52,14 @@ public TcpLoggingServer(IActorRef sinkCoordinator) | |||
|
|
|||
| Receive<Tcp.Received>(received => | |||
| { | |||
| var message = received.Data.ToString(); | |||
| sinkCoordinator.Tell(message); | |||
| if (received.Data.Count >= BufferSize) | |||
There was a problem hiding this comment.
Can we increase the message size used by the MNTR in Akka.IO.Tcp?
|
Looks good overall but left some remarks |
|
Increasing the buffer size would not actually fix the problem, because we'd never know how big the stack trace will be, a very deep stack trace with lots of inner exceptions |
|
@Arkatufus yeah but 512 is exceptionally tiny. |
|
I'll bump it up to 10 KB then? that should be a fairly safe number |
Closes #100