Bug 765039

User-Agent:       Mozilla/5.0 (X11; Linux i686) AppleWebKit/534.34 (KHTML, like Gecko) konqueror/4.7.2 Safari/534.34

The summary is exact enough about the problem. I describe background

I created an image of a running system to run on a second machine. I have tomcat6 and postgresql starting during boot. When I tested the disk on the original machine, everything seemed OK. But when I shifted to the second machine, first day it ran fine, but second day, tomcat and postgresql failed to have started when booted to RL 3. Networking was not essential to the use scenario, so it was not given attention to. Whenever problem occurred, two things were common. One, /sbin/runlevel result was 'unknown'. Second, when I 'su -' ed to root , I had bash error, ' bash : test : : integer expression expected '. Holding bash error as clue, I investigated.

After a long, tiring investigation, I wanted to google for a search, lo, the problem was gone. The I tested with conditions. Now I can say, if the Network is controlled by 'ifup' method, this problem does not occur. If   the cable is connected , even if the NM connection leads nowhere, like conflicting manually set IPs, still the problem does not occur. It occurs only when cable is unplugged. But I have not tested with wifi. My colleague has, and it indicates the problem occurs only with wired connection. Also, stopping of tomcat and postgresql services had no effect on the bug. Actually systemd waits for a long time before logging in  apparently for a cable plugging.

Reproducible: Always

Steps to Reproduce:
1. Setup NM to control Network. 
2. Unplug the cable.
3. Boot to any level and test runlevel. Check bash complaints for the logged in users.
Actual Results:  
1. Runlevel becomes 'unknown'

2. bash complains about 'expected integer expression'.

Expected Results:  
runlevel and bash have no business to expect cable plugging. Likewise for Tomcat and Postgresqk services.

I think it is a serious lapse. This needs urgent fixing.