Comment 1 Florian Best 2023-01-05 13:52:09 CET

The error message seems wrong: "Daemon died" is one example. Correct is that the socket connection was closed and the send data did not contain a null byte.

management/univention-directory-manager-modules/univention-cli-client:
126 def receive_answer(sock):
127 »   """Receive complete answer from server."""
128 »   data = b''
129 »   while True:
130 »   »   buf = sock.recv(1024)
131 »   »   if len(buf) == 0:
132 »   »   »   print('E: Daemon died.', file=sys.stderr)
133 »   »   »   sys.exit(1)
134 »   »   elif buf[-1:] == b'\0':
135 »   »   »   buf = buf[0:-1]
136 »   »   »   data += buf
137 »   »   »   break
138 »   »   else:
139 »   »   »   data += buf
140 »   return data

Comment 2 Florian Best 2023-01-05 16:50:38 CET

In Bug #33224 we changed that the CLI-client shares their stdout+stderr file-descriptors with the CLI-server. The CLI-server then doesn't send the whole UDM object data as JSON object but prints directly to the stdout/stderr. But it also sends some data (most likely/often just a empty list) as JSON object back to the socket.

The change had also the side-effect that data is printed when the client process has already ended:
timeout -k 0.1 0.1 udm users/user list

Comment 3 Arvid Requate 2023-01-05 17:21:45 CET

Yes, I see, that's the bug I put in the "See also" section here.
Do you have an idea how we could make this more robust, so
that udm client doesn't interpret an error when there apparently is none?

Comment 4 Florian Best 2023-01-05 17:23:25 CET

I have to reproduce it myself to see why it is complaining. I will look at it back in ~ 2 weeks.

Comment 6 Florian Best 2024-01-10 17:03:44 CET

Any uncaught exception in univention-cli-server in server_main() or MyRequestHandler.handle() will trigger this.
→ so we should handle all exception and send a operation failed message

Another reason could be a broken pipe during "print(file=stream)" in admincli.py: Bug #32532