Bash - stopping on failures

A good habit is to begin each bash script with:

set -e (stop execution if a command returns with non-zero exit status)
set -u (stop execution if an undefined variable has been used, instead of treating the variable to have an empty value)

However, the set -e commnd has some consequences. The script can end where we would not expect it:

# A sub-command returns non-zero exit status

man non_existing_entry # error (as expected)
OUTPUT=$(man non_existing_entry) # error (maybe not expected)
man non_existing_entry & # NOT error

# Result of an arithmetic expression is zero

let NUM=0 # error (not expected)
NUM=1
let NUM=NUM-1 # error (not expected)

NUM=0 # NOT error
NUM=$((0)) # NOT error
NUM=1
NUM=$((NUM-1)) # NOT error

# grep did not find anything

ip addr | grep non_existing # error (not expected)

We can avoid treating these cases as errors by surrounding the commands by set +e and set -e:

set +e
man non_existing_entry
let NUM=0
ip addr | grep non_existing
set -e

A function doing this arround a command can help to simplify the code:

noerror()
{
  set +e
  $@
  set -e
}

It can be used as follows:

noerror man non_existing_entry
ip addr | noerror grep non_existing

Written on February 12, 2016