ComputingConfidence

The aim is to ensure that root is emailed in case of abnormalities in cron jobs:

The right way to deal with this is to change your script so that it
pipes its results into (say) grep:

   #!/bin/sh
   killall -q fetchmail
   /usr/bin/fetchmail | grep -iv "No mail for "

That way there will be no output generated if everything works as
expected.

With egrep you can scan for more than one expected result. Or you could
use a chain of greps.

killall outputs nothing if it successfully kills things. "-q" tells it
not to report that no processes were killed. If anything else happens,
you will be informed. There may be a similar switch for fetchmail, which
would save using grep.

If expected messages are going to stderr instead of stdout you have to
use a syntax that is slightly less well known. For example, to capture
the "no process killed" message from killall (instead of using "-q"):

   killall telnet 2>&1 | grep -iv "no process killed"

One day your script will output something different: "fetchmail: Oops!
Major problem! You have eight hours to act before your hard disk is
wiped and your pets are shaved!"

Because that doesn't match the regexp, it will be output and thus mailed
to you in time for you to save the day.