start-stop-daemon wrong in nfsroot reproduced

[malk at sidehack.sat.gweep.net]

Hi-

I had an nfsroot I made end up w/ start-stop-daemon.REAL and
start-stop-daemon.distrib plus the link pointing to
fai-start-stop-daemon which is the fake one.

Doing FAI of boxes off this nfsroot left the fake start-stop-daemon on the
target systems which is obviously undesirable.

I recently changed fai servers and re-did fai-setup and start-stop-daemon
came out correct.  Both nfsroots were made w/ FAI 2.5.3

For the broken nfsroot, I logged the output of fai-setup to a file
and I still have the broken nfsroot laying around.

For the good nfsroot, I also logged the output of fai-setup to a file and
obviously I'm using that nfsroot now.

If anyone is interested, I could post the logs from each -- I noticed
they are a bit different.   In the broken fai-setup log, there's an error
processing lilo and a complaint about failing to configure the base
packages -- it then attempts up to 5 times to get it right.  It complained
about libdevmapper1.00 being a dependency for lilo, but libdevmapper1.00
wasn't installed.  The successful nfsroot has no such complaints and no
retries up to 5 times.  I'm guessing it's debootstrap that's doing this.

Anyway -- at least I learned from the mailing list that the fix for
left over fake start-stop-daemon is to re-make the nfsroot and also
there's been trouble reproducing the problem...  so this is my offer
to try and post any forensics you might be interested in.

Other than these wrinkles, FAI has been a great time saver.  Thanks
again for keeping FAI development going -- it would be a major loss
for debian if FAI wasn't an option -- especially considering the new
installer has had lots of issues (granted it's in development, but
FAI just seems a lot easier to use once a server is setup).

-Eric Malkowski