Code review comment for lp:~ubuntu-core-dev/ubuntu/utopic/sysvinit/unreviewed

Revision history for this message
Martin Pitt (pitti) wrote :

I built and installed this into a relatively clean utopic VM.

  /rc/rc?.d/ before: http://paste.ubuntu.com/7519546/
  /rc/rc?.d/ after: http://paste.ubuntu.com/7519547/

I can't do the comparison on my workstation as I already ran insserv a few weeks ago manually (without this branch and the init.d scripts added back); I also didn't notice any problems since then.

I did a few reboot and powerdown/boot tests without obvious failures, but it seems prudent to look into the details as many of the scripts' effects are not immediately obvious. I looked at the differences in above pastebins:

= rc[2345].d =

The order changed a bit, but only for scripts which don't depend on each other. insserv tends to sort the scripts without/with fewer dependencies like dns-clean first, which makes sense. rc.local is still correctly the last item.

= rc[06].d =
This looks a bit odd. In the original "legacy" ordering we had

lrwxrwxrwx 1 root root 18 Mej 25 13:20 S20sendsigs -> ../init.d/sendsigs
lrwxrwxrwx 1 root root 17 Mej 25 13:20 S30urandom -> ../init.d/urandom
lrwxrwxrwx 1 root root 22 Mej 25 13:20 S31umountnfs.sh -> ../init.d/umountnfs.sh
lrwxrwxrwx 1 root root 18 Mej 25 13:20 S40umountfs -> ../init.d/umountfs
lrwxrwxrwx 1 root root 20 Mej 25 13:20 S60umountroot -> ../init.d/umountroot
lrwxrwxrwx 1 root root 14 Mej 25 13:20 S90halt -> ../init.d/halt

but these got turned into "K" links like

lrwxrwxrwx 1 root root 20 Mej 26 09:48 K01kerneloops -> ../init.d/kerneloops
lrwxrwxrwx 1 root root 15 Mej 26 09:48 K01rsync -> ../init.d/rsync
lrwxrwxrwx 1 root root 27 Mej 26 09:48 K01speech-dispatcher -> ../init.d/speech
-dispatcher

won't that change the semantics? (being called with "stop" instead of "start", etc.)

= rc1.d =

Looks fine

= rcS.d =

As expected we now have the ~ 11 scripts there which got introduced merely for ordering reason, they should be no-ops. Our 4 original links are still there in a reasonable order.

So we should investigate the S → K change for 0/6, otherwise this looks good.

« Back to merge proposal