Bash: synchronisation barrier

barrier.sh

#!/bin/bash

./process_1.sh &
pid_1=`echo $!`

./process_2.sh &
pid_2=`echo $!`

./process_3.sh &
pid_3=`echo $!`


# synchronisation barrier
wait $pid_1 $pid_2 $pid_3

./process_4.sh

barrier_exitstatus.md

Bash barrier with all exit status

Demo example

In statement tasks="(sleep $i; true && echo ok) & ${tasks}":

• The boolean true represents command execution that finish by exit 0
• You can test with true; echo $?
• { (sleep 3; true && echo ok) & (sleep 2; true && echo ok) & } | grep -c ok | ( read n && (( n==2 )) ) ; echo $?
• A boolean false in this example represents command execution that finish by exit 1 (other exit code)
• You can test with false; echo $?
• { (sleep 3; true && echo ok) & (sleep 2; false && echo ok) & } | grep -c ok | ( read n && (( n==2 )) ) ; echo $?

i=0
tasks=''

for engine in 1 2 3
do
  i=$(($i+1))

  tasks="(sleep $i; true && echo ok) & ${tasks}"
done

echo "Executed command: eval '{ ${tasks} } | grep -c ok | ( read n && (( n==${i} )) ) ; echo \$?'"

# synchronisation barrier with exitstatus ckeck
exitstatus=$(eval "{ ${tasks} } | grep -c ok | ( read n && (( n==${i} )) ) ; echo \$?")
echo $exitstatus

Concrete example

• Launch several ruby tests and verify that all pass:

#!/bin/bash

i=0
tasks=''

cd engines
for engine in 1 2 3
do
  i=$(($i+1))

  tasks="(cd `bundle show $engine`; bundle exec rspec && echo ok) & ${tasks}"
done

echo "Executed command: eval '{ ${tasks} } | grep -c ok | ( read n && (( n==${i} )) ) ; echo \$?'"

# synchronisation barrier with exitstatus ckeck
exitstatus=$(eval "{ ${tasks} } | grep -c ok | ( read n && (( n==${i} )) ) ; echo \$?")
echo $exitstatus