0.63.0 broke property injection

[seahurricanelemonade]

I'm unable to provide code unfortunately but I can describe the scenario.

I have a test which inherits from an abstract class (B) that also inherits from an abstract class (A)

B has a proprety injected which is a Pulsar connection implementation that has SharedType.None. The pulsar connection class also has another property injected which is a Pulsar test container with SharedType.PerTestSession.

A has a property injected which is based off WebApplicationFactory with ShareType.None. The WebApplicationFactory also has a property injected which is the Pulsar test container with SharedType.PerTestSession.

Running a test and it fails to initialise the Pulsar test container in time for use by the Pulsar connection class.

       at TUnit.Core.ObjectInitializer.GetInitializationTask(Object obj, IAsyncInitializer asyncInitializer)
   at TUnit.Core.ObjectInitializer.InitializeAsync(Object obj, CancellationToken cancellationToken)
   at TUnit.Engine.Services.PropertyInitializationOrchestrator.InitializeObjectWithPropertiesAsync(Object instance, PropertyInjectionPlan plan, Dictionary`2 objectBag, MethodMetadata methodMetadata, TestContextEvents events, ConcurrentDictionary`2 visitedObjects)
   at TUnit.Engine.Services.PropertyInjectionService.<>c__DisplayClass5_0.<<InjectPropertiesIntoObjectAsyncCore>b__0>d.MoveNext()
--- End of stack trace from previous location ---
   at TUnit.Engine.Services.PropertyInjectionService.InjectPropertiesIntoObjectAsyncCore(Object instance, Dictionary`2 objectBag, MethodMetadata methodMetadata, TestContextEvents events, ConcurrentDictionary`2 visitedObjects)
Stack trace:    at TUnit.Engine.Services.PropertyInjectionService.InjectPropertiesIntoObjectAsyncCore(Object instance, Dictionary`2 objectBag, MethodMetadata methodMetadata, TestContextEvents events, ConcurrentDictionary`2 visitedObjects)
   at TUnit.Engine.Services.PropertyInjectionService.InjectPropertiesIntoObjectAsyncCore(Object instance, Dictionary`2 objectBag, MethodMetadata methodMetadata, TestContextEvents events, ConcurrentDictionary`2 visitedObjects)

[thomhurst]

So the pulsar connection class is being initialised before the test container? But it should be the other way round? Is that right?

[seahurricanelemonade]

So the pulsar connection class is being initialised before the test container? But it should be the other way round? Is that right?

Seems like that, yes

[thomhurst]

Try v0.63.3

[seahurricanelemonade]

That has fixed the issue, thanks.

I'm also experiencing another issue with disposal of the properties ever since trying to upgrade from 0.25.21.

##[error]Unobserved task exception: System.AggregateException: A Task's exception(s) were not observed either by Waiting on the Task or accessing its Exception property. As a result, the unobserved exception was rethrown by the finalizer thread. (Cannot access a disposed object.
Object name: 'DotPulsar.Internal.Connection'.)
 ---> DotPulsar.Internal.Exceptions.ConnectionDisposedException: Cannot access a disposed object.
Object name: 'DotPulsar.Internal.Connection'.
   at DotPulsar.Internal.Connection.Send(BaseCommand command, CancellationToken cancellationToken)
   at DotPulsar.Internal.Connection.<>c__DisplayClass39_0.<<ProcessIncomingFrames>b__1>d.MoveNext()
   --- End of inner exception stack trace ---

That's all the information I can get as its thrown in the finalizer thread. And it doesn't happen everytime so something makes me think there is a race condition somewhere.

[thomhurst]

I'm also experiencing another issue with disposal of the properties ever since trying to upgrade from 0.25.21.

##[error]Unobserved task exception: System.AggregateException: A Task's exception(s) were not observed either by Waiting on the Task or accessing its Exception property. As a result, the unobserved exception was rethrown by the finalizer thread. (Cannot access a disposed object.
Object name: 'DotPulsar.Internal.Connection'.)
 ---> DotPulsar.Internal.Exceptions.ConnectionDisposedException: Cannot access a disposed object.
Object name: 'DotPulsar.Internal.Connection'.
   at DotPulsar.Internal.Connection.Send(BaseCommand command, CancellationToken cancellationToken)
   at DotPulsar.Internal.Connection.<>c__DisplayClass39_0.<<ProcessIncomingFrames>b__1>d.MoveNext()
   --- End of inner exception stack trace ---

That's all the information I can get as its thrown in the finalizer thread. And it doesn't happen everytime so something makes me think there is a race condition somewhere.

Even on the latest version?

[seahurricanelemonade]

Yes, I've been experiencing this ever since upgrading from 0.25.21. It's an issue independent from the original issue where 0.63.0 broke the injection, that's resolved. But there's something weird going on with the disposals... is there anything I can provide to help you track it down?

[thomhurst]

A minimal project reproducing it would be good, but hard if it's only sometimes!

[thomhurst]

If you can make it happen with dotnet run, use --detailed-stacktrace to try and get a better stacktrace

[seahurricanelemonade]

Seems to be fixed in 0.66.6