ZeroMQServer.Start: ZMQ Poller died

[stefanmm]

Description
Program crashes when need to change theme - both from dark to light and light to dark. I am using stable Windows 11, latest version.

Expected behavior
When the time comes to change the theme, nothing happens. Icon disappears from the system tray, process is killed. This happens every day. To fix this, I start the program again and it switch theme as it should. I tried clean install and also installing via Windows store - nothing. Previous version (non "X" one) worked just fine, even on Windows 11.

Log Data
2021-10-11 16:58:43 | Error | ZeroMQServer.Start: ZMQ Poller died An existing connection was forcibly closed by the remote host.
config.yaml.txt
service.log

Commit Hash and Version
Commit: 9faf1f3
Version: 10.0.0.26

Screenshots

[Spiritreader]

ZeroMQ uses TCP sockets to communicate.
Something on your system is causing these TCP connections to be closed.

Do you

• regularily enable/disable your network adapters?
• have an application running that is opening random tcp connections and terminates them ungracefully?

Update: You can switch to the beta channel and test a fix for ZMQ being able to escape controlled exception handling

[stefanmm]

I found what the problem is - VPN client. From time to time I am connecting to the VPN (for work) and as soon as I connect to the VPN "Auto Dark Mode" crashes. This happened just now on latest Beta update.

This is the error now (beta channel) when I connect to VPN:
ZeroMQServer.Start: ZMQ Poller died SocketException > An existing connection was forcibly closed by the remote host. > Boolean TryRecv(Int32, NetMQ.Core.Command ByRef)

[Spiritreader]

I am unable to reproduce your issue, as such you would have to test out a few builds until I figure out what is causing this.
I've switched over the poller to run async on a NetMQ managed thread in an attempt to handle this gracefully (before brute-forcing it).

You can test out that new build if you want to (10.0.1.35)

Please, when testing, note the following:

• It is very important that you test whether the NetMQ Poller is still responding after switching networks / connecting to your VPN
• To do that, open the AutoDarkMode Shell (found in the install directory), and send an Alive Message.
  You should get "Response: Ok"
  

[stefanmm]

Ok I updated to the latest build (10.0.1.35) and the problem is still there.

I will send you more details about VPN connection so you can replicate on your end.

[Spiritreader]

I have made the zmq poll thread yet more resilient. In theory, it should now be able to auto-recreate sockets once they have become unavailable. I pushed this to .36

[stefanmm]

Tried, there is still the problem. Error log:
2021-10-12 15:35:57 | Error | ZeroMQServer.Start: socket no longer available, acquiring new socket. error: SocketException > An existing connection was forcibly closed by the remote host. > Boolean TryRecv(Int32, NetMQ.Core.Command ByRef) 2021-10-12 15:35:57 | Error | ZeroMQServer.Start: removing and disposing socket failed: SocketException > An existing connection was forcibly closed by the remote host. > Int32 Send(Byte[]) 2021-10-12 15:35:57 | Info | ZeroMQServer.Start: socket bound to port: 13012
The problem is that the process is killed. So, even if you try to auto-recreate sockets it can't be done since the process doesn't exists anymore.

To replicate on your end:

1. make sure Auto Night Mode is up and running
2. install Windscribe VPN Windows client (you can try any other too)
3. login with Free or Pro account (doesn't matter)
4. connect to any free/pro network from the list if you do not have your own <- at this point Auto Night Mode crashes
5. open Auto Night Mode again <- everything works again
6. disconnect from the VPN network <- everything still works
7. connect again to any network <- program crashes

[Spiritreader]

Alright, I'll try to reproduce. Thanks!

Could you please attach the entire service.log file in the meantime?

[Spiritreader]

Windscribe is telling me that it kills all sockets by default.
And moreso, it even kills sockets that target localhost (127.0.0.1) which aren't even connecting outbound to the internet.

I'm not sure if I can do anything about this if WindScribe is killing the AutoDarkModeSvc.exe process

[stefanmm]

Ok, now it makes more sense. I never got that message, but after I saw this I remembered that there is one option related to that:

I disabled it and now everything works.

I would never suspect that this could be a problem because the whole program was crashing. I will keep that option disabled, thank you for your help.

[Spiritreader]

Update: I have continued my investigation, and it looks like this is a bug in the NetMQ library we're using. So ADM does crash after all.
I am checking out if I can find a fix for this.