consistent and reliable client exit code and information

[Antoine Martin]

There can be many reasons why we close a connection.
Unless this is simply a network failure which is an error and should be reported as such (and the client process should terminate with a non-zero exit code), we should generally be able to tell the user why the disconnection occurred.
Most of the code is already there, it just needs to be constified so these values can be relied on.

[Antoine Martin]

See r7075.

Please test, ie:

• attaching from another client with sharing turned off, the previous client gets disconnected with this message server requested disconnect: new client (this session does not allow sharing) and the exit code is 0 (EXIT_OK)
• stopping a server (either control-C or via xpra stop) whilst the client is connected gives: server requested disconnect: server shutdown (and still 0)
• killing the vfb from underneath the server (naughty!) kills it and the client just sees a Connection lost and sets the exit code 1 (EXIT_CONNECTION_LOST)
• similarly, severing the connection between the client and server eventually causes a server ping timeout - waited 60 seconds without a response, followed by the same Connection lost and sets the exit code to 2 (EXIT_TIMEOUT)

Some of the other disconnection messages can be a bit harder to trigger, see #614 for example.

Will that do?

[Antoine Martin]

As of r7088, the server-side messages are nicer and the client tries to tell the server when it is exiting and why, ie:

• when receiving a control-C:

client Protocol(SocketConnection(/home/antoine/.xpra/desktop-10)) has requested disconnection: exit on signal SIGINT

• when exiting from the system tray:

client Protocol(SocketConnection(/home/antoine/.xpra/desktop-10)) has requested disconnection: client exit

[Antoine Martin]

Should be closed for v0.14.x, raising to critical.

[Antoine Martin]

Should be closed against milestone 0.14

[Antoine Martin]

Bump!

[Scott Bennett]

Verified and tested in 0.14.12, thank you

[migration script]

this ticket has been moved to: ​https://github.com/Xpra-org/xpra/issues/623