Possible Bug: Complete lock-up in ASP.NET Core startup due to StubHelpers?

[rossdempster]

Hi all,

This is potentially a difficult issue to track down. I have been developing with ASP.NET Core for a few months now and (very) rarely the application locks-up completely during start up. I am currently running version 2.1.101 but don't recall which version I was running when I first encountered this issue. The time between manifestations is measured in weeks. I assume this is some kind of unlucky race condition.

I captured this screenshot from the latest occurrence. The lock-up always occurs within WebHostBuilder.Build(). The rest of the stack trace might be variable because this is the first time I've had a proper look at it.

In this instance, the code appears to be stuck in StubHelpers.AddToCleanupList (hence why I am posting this here). Resuming the application has no effect. We remain stuck in the same place. I need to restart the application to get past this point.

[janvorli]

@rossdempster are you sure that in the case you've posted here the stack trace wasn't changing when you resumed the execution and then interrupted it again?

[rossdempster]

That is what I observed yes.

I've spent the last 30 minutes trying to reproduce this by continually restarting my app under the debugger and I've just found something interesting.

If I hit CTRL+C before startup is complete (but after the window appears) I get exactly the same effect. The application appears to be running but pausing it and observing the stack shows it to be stuck at some arbitrary location. It's different each time I try it.

It's actually possible to get the program stuck here:

No matter how often I resume and break the application, that's as far as it's got!

I think I may have been trying to copy/paste something at the exact moment the window has appeared and inadvertently triggered this issue. I've just never noticed the cause and effect.

I'm not sure whether it's worth taking this any further? The observed behaviour of the code having stopped executing but the stack trace being preserved does seem to point toward an issue somewhere in the runtime, but it only seems to affect the debugger because I've not managed to reproduce it with dotnet run. The impact seems minimal.

What do you think?

[tommcdon]

@cshung can this issue be resolved?

[tommcdon]

@cshung investigated and was not able to repro. Please provide a dump of the process when it is hung and we can investigate further. Closing the issue due to lack of actionable information for now.