Strange stuff happens if you kill radiomail when establishing a connection

[driechers]

I seem to hit a lot of edge cases if i kill radiomail when connecting. Things I've observed:

• Radio stays keyed up but no audio is transmitted
• Vara sends audio to radio but does not key it
• Vara is not terminated after it fails

It also appears that if the app is backgrounded, it wont continue any operations with the radio, even if background refresh is enabled. This seems to also trigger similar edge cases.

I'm thinking perhaps some robustness updates to ensure vara is killed should do the trick, I'm not sure if that will unkey the radio necessarily. I am keying with ptt over rigctl.

My first thoughts are:

• Check for process still running after a delay from sigterm then send sigkill
• Check for process still running when new connections are received and kill it

[georges]

Thanks for the observations, make sense that would occur. The underlying assumption is that RadioMail manages the session, so yes if it disappears, what you described is likely to occur. varanny could act as a watchdog and take some action but the challenge is when to take such action? Killing the process after a certain timeout would risk to kill an ongoing connection, and if the timeout is too large, that would defeat the purpose entirely.

A possible option could be to have a more interactive heart beat between RaidoMail and varanny. If the heart beat stops, varanny could interpret that as a disconnect and kill the processes it just spanned. This assumes that rigctld would unkey upon termination, which hopefully is the case.

Could be a worthwhile enhancement to contemplate for more robustness.

[georges]

Another simpler approach could be to simply require the connection between varanny and RadioMail to remain open. If the socket closes, varanny assumes it should shutdown the session.

[georges]

Ah, reading through the code, it actually should be terminating the processes when connection close. The problem might be that on Windows it's not a graceful termination. What OS are you running on?

[driechers]

I see yes a heartbeat is not what I had in mind. I’m assuming varanny uses TCP and that iOS behaves like Linux and if RadioMail crashes or is killed then the OS closes the TCP session. If that is not the case and it’s UDP I could see why this would happen.

The thought behind sending a sigkill at start and end is that sigterm can fail. Vara hf is kindof junky software so I very much could see it failing to shutdown and stay running. Sigkill cannot be handled so it will terminate the process in those cases.