[systemd-devel] Error message for Accept=true and non-template service file

[Lennart Poettering]

On Mon, 19.11.12 23:17, Thomas H.P. Andersen (phomes at gmail.com) wrote:

> I started looking at the remaining SysV to systemd conversions in
> fedora. For telnet there was already prepared socket and service files
> but also reports that they did not work [1]. It was not specified how
> the two files should be named and I suspect that the reported error
> could be due to using a non-template service file type with an
> Accept=true socket.
> 
> If this was the case then it would have helped with a clearer error
> message. Currently with a non-template service the socket can be
> started just fine, but on first connect we will get this:
> 
> # systemctl status telnetd.socket
> telnetd.socket - Telnet Server Activation Socket
>       Loaded: loaded (/etc/systemd/system/telnetd.socket; disabled)
>       Active: failed (Result: resources) since Mon, 2012-11-19
> 22:54:27 CET; 4s ago
>     Accepted: 1; Connected: 0
>       CGroup: name=systemd:/system/telnetd.socket
> 
> Nov 19 22:54:13 localhost.localdomain systemd[1]: Starting Telnet
> Server Activation Socket.
> Nov 19 22:54:13 localhost.localdomain systemd[1]: Listening on Telnet
> Server Activation Socket.
> Nov 19 22:54:27 localhost.localdomain systemd[1]: Unit telnetd.socket
> entered failed state
> 
> # journalctl -n4
> -- Logs begin at Mon, 2012-11-19 21:07:42 CET, end at Mon, 2012-11-19
> 22:54:27 CET. --
> Nov 19 22:54:13 localhost.localdomain systemd[1]: Starting Telnet
> Server Activation Socket.
> Nov 19 22:54:13 localhost.localdomain systemd[1]: Listening on Telnet
> Server Activation Socket.
> Nov 19 22:54:27 localhost.localdomain systemd[1]: telnetd.socket
> failed to queue socket startup job: Invalid argument

^^^

This one really should be visible in the first output too. We should add
_SYSTEMD_UNIT= fields to all messages generated on a specific unit. I
added this to the TODO list.

> Nov 19 22:54:27 localhost.localdomain systemd[1]: Unit telnetd.socket
> entered failed state
> 
> It is not very obvious what went wrong. Would it not make sense to
> detect the wrong service file type at socket start time? I could try
> to write a patch for that. (Any advice or pointers about where to best
> implement it would naturally be much appreciated)

It's probably enough to reword the message a bit. 

Does this change make sense to you?

http://cgit.freedesktop.org/systemd/systemd/commit/?id=10f70492aea211981e4bdbe58dd7ea110e05cd16

(We also should add this message to the message catalog, but I'll leave
that for later...)

Lennart

-- 
Lennart Poettering - Red Hat, Inc.