Process exit code not set when service fails to start/stop gracefully (v7)

[ramonsmits]

The host uses (old) version 0.8.0 of the Topshelf project which seems to incorrect set the exit code when an errors happens during start/stop. An error could be as described in #119.

Magnum.StateMachine.StateMachineException: Exception occurred in Topshelf.Internal.ServiceController`1[[NServiceBus.Hosting.Windows.WindowsHost, NServiceBus.Host, Version=7.0.0.0, Culture=neutral, 

--8<-----8<-----8<-----8<-----8<-----8<-----8<-----8<-----8<-----8<---

   at NServiceBus.Hosting.Windows.WindowsHost.Start() in C:\Build\src\NServiceBus.Hosting.Windows\WindowsHost.cs:line 30
   at Magnum.StateMachine.EventActionList`1.Execute(T stateMachine, Event event, Object parameter) in c:\Code\Topshelf\src\Topshelf\Internal\Hosts\ServiceHost.cs:line 0
   --- End of inner exception stack trace ---
   at Magnum.StateMachine.ExceptionActionDictionary`1.HandleException(T stateMachine, Event event, Object parameter, Exception exception) in c:\Code\Topshelf\src\Topshelf\Internal\Hosts\ServiceHost.cs:line 0

This PR is basically a workaround for the Topshelf error by calling Environment.Exit(int) ourselves. This results in the windows service to actually fail which results in Windows Service Recovery to kick in.

[ramonsmits]

@Particular/nservicebus-maintainers @andreasohlund @sfarmar This seems to be a bug in Topshelf, this is a workaround.

Another option might be to update to a newer 1.0+ version of Topshelf but probably that has already deemed to much work as I also see Particular/NServiceBus#2201 about removing Topshelf.

Any reason that we use 0.8.0, as it seems to be very old.

[ramonsmits]

Ok, digged a bit deeper and noticed this: https://github.com/Particular/Topshelf/commits/TopshelfForNSB which seems to be our fork.

[ramonsmits]

WIthout looking at the implementation of Magnum I think this is why it fails: https://github.com/Particular/Topshelf/blob/TopshelfForNSB/src/Topshelf/Internal/ServiceController.cs#L76

[seanfarmar]

@ramonsmits maybe we can get @andreasohlund to provide some context/history for this?

Otherwise, LGTM

[andreasohlund]

Can't remember, John was the only one that made it into that code base and came back alive