ahem, steps 1. and 4. were written with the old way in mind, using systemd, it'd be some kind of 'systemctl ... tftp.socket' magic as noted at the end of the description, sorry for possible confusion

Uh, why is tftp-servcer shipping both unit files and also configuration for xinetd?
But you can still enable tftp with xinetd through chkconfig, just use 
chkconfig --type xinetd tftp on (like in the case that there was an initscript and xinetd service with the same name).

I have written a support for systemd socket activation to chkconfig but it only works for .socket and @.service pairs. In cases like this I don't have the information if .service or .socket should be enabled.

What I would suggest is to put this to tftp.service 
[Install]
Also=tftp.socket

This would mean that enable on tftp.service is still noop, but it will also enable the socket.

I think we should go with this solution, because it will 
a) fix the problem with chkconfig
b) it will also make possible to just type systemctl enable tftp instead of tftp.socket
which I think is more user-friendly

In Fedora, this was fixed by http://pkgs.fedoraproject.org/cgit/tftp.git/commit/?id=1f86b3299f04684be5db03409fd4e6d392bca804. I'll go with the same fix here, since using Also= doesn't quite work in this case.

Created attachment 961633 [details]
patch

(In reply to Lukáš Nykrýn from comment #3)
> But you can still enable tftp with xinetd through chkconfig, just use 
> chkconfig --type xinetd tftp on (like in the case that there was an
> initscript and xinetd service with the same name).

um, that still doesn't work ...

[0 root@pipa02 ~]# chkconfig --type xinetd tftp on
[0 root@pipa02 ~]# tftp 127.0.0.1 -c get somenonsense
Transfer timed out.

(but it at least modifies /etc/xinetd.d/tftp )

and is a change from RHEL6 where '--type xinetd' wasn't needed


> b) it will also make possible to just type systemctl enable tftp instead of
> tftp.socket
> which I think is more user-friendly

100% agreed, it'd be much easier to use this way

(In reply to Karel Volný from comment #6)
> > chkconfig --type xinetd tftp
...
> and is a change from RHEL6 where '--type xinetd' wasn't needed

well, I mean ... after applying the patch from comment #5, '--type xinetd' won't be needed too, but it will do the things the systemd way which is inconsistent

in RHEL6 you could do

chkconfig tftp enable
chkconfig --type xinetd tftp disable

et vice versa and you'd get the desired effect

in RHEL7 if you'd do

chkconfig tftp enable
chkconfig --type xinetd tftp disable

then to my best understanding tftp(.service) will keep running

I'd tend to disregard this as obvious admin error, using the commands inconsistently, but I don't know how much do we want to keep the old behaviour (as chkconfig is a compatibility layer - those who want the new behaviour use systemctl directly I guess ...)

Well the problem here is that we are shipping both xinetd configuration file and systemd unit-file. If you would ship initscript and xinetd file in rhel6 you would be in the same situation.

I would not complicate it. After the fix chkconfig tftp (en|dis)able will work and if you are using --type you should know what you are doing.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://rhn.redhat.com/errata/RHBA-2015-2192.html