Needs Ubuntu-style init script

With this change, the init script output looks like

$ sudo /etc/init.d/xinetd start
 * Starting internet superserver xinetd [ ok ]

$ sudo /etc/init.d/xinetd stop
 * Stopping internet superserver xinetd [ ok ]

/usr/share/doc/lsb-base/README.Debian.gz says:

  This package also includes the LSB-specified logging functions:

  log_success_msg message
  log_failure_msg message
  log_warning_msg message

  These functions *do not* comply with Debian policy and should only be used
by LSB packages.

I'm not entirely certain what should be used instead in this case.

Also note that that README file specifies a newer dependency than in this patch.

Strictly speaking, converting it to an Upstart job would be better

Some comments on the upstart job.

In order to maintain the old behavior, we send a SIGQUIT in the pre-stop portion of the upstart job. Per man page:
- SIGQUIT causes program termination.
- SIGTERM terminates all running servers before terminating xinetd.

Doing this is incompatible with using 'respawn', and I've opened upstart bug 605007 to address that.
'respawn' would be nice, as xinetd is a daemon that could die, and it would be nice to have upstart watching it, but the old sysv scripts had no such function.

In short, the upstart job should function almost identically to the sysvinit script.

A work-around: add "stop" to the pre-stop script. This will prevent respawning. This seems to work:

{{{
       P=`cat /pidfile`
       kill -QUIT $P
       wait $P
       stop # Prevent respawn bug in upstart
}}}

Why is the task fix commited? Does anybody plan to upload that to natty?

@ttx and @smoser,

This was marked 'Fix Committed' for maverick-alpha-3. I don't see the changes in xinetd in natty. Is this something that is planned? If yes, please mark the status as 'Triaged' or something more appropriate. If not, please mark as "Won't Fix" with reasons why. There is no reason leaving a Wishlist bug open for 4.5 years if no one is planning on fixing it. Thanks!

Dan,
  In my tests here, I added 'respawn', and made the pre-stop script look like:

pre-stop script
  xinetd_pid=$(status | awk '/stop\/pre-stop/ { print $NF }')
  [ -n "${xinetd_pid}" ] || exit 0
  kill -QUIT "${xinetd_pid}"
  wait ${xinetd_pid}
  stop # Prevent respawn bug in upstart (LP: #605007)
end script

With 'respawn' commented out, the script works like:
$ sudo status xinetd
xinetd start/running, process 490
$ sudo stop xinetd
xinetd stop/waiting
$ sudo start xinetd
xinetd start/running, process 1364

With respawn added back in, it looks like:
$ sudo status xinetd
xinetd stop/waiting
$ sudo start xinetd
xinetd start/running, process 3094
$ sudo stop xinetd
xinetd start/running, process 3275
$ sudo stop xinetd
xinetd start/running, process 3394

So I don't think the workaround works.
Initial debugging shows that nothing in that pre-stop script after 'wait' will run. I'm guessing that upstart just kills the pre-stop job after it sees that the process is gone.

So, I think the right thing to do here is to move to the upstart job that I have, and, if you see fit, open a new bug saying "respawn does not work with xinetd upstart job".

I've linked a new branch (lp:~smoser/ubuntu/natty/xinetd/bug43574) which is identical to the one I unlinked except for 'natty' as the release in debian/changelog.

Please sponsor and upload.

Thanks, this looks good. I'll upload shortly.

This bug was fixed in the package xinetd - 1:2.3.14-7ubuntu4

---------------
xinetd (1:2.3.14-7ubuntu4) natty; urgency=low

  * add xinetd upstart job (LP: #43574)
 -- Scott Moser <email address hidden> Tue, 13 Jul 2010 09:24:49 -0400