Justin Li


Killing Children

2009-05-20

It is commonly known that a lot of computer jargon are tongue in cheek, especially when they were first developed. For examples, processes are compared to people, so when they spawn other processes, the original one is called the parent and the other one the child. When you stop a process, it's called killing it. There are also times when the process is dead but still taking up resources, and very naturally we call them zombies...

Recently I ran into the problem of killing child processes. I want the output of a command in a variable, so something like this:

output="$(command...)"

The problem is two fold: the command itself spawns new children, and the command I'm running may not terminate. What I want to do is set a timer, then when the time is up I would kill the process. To kill it though, I would have to kill the youngest (inner most) child. I couldn't find a simple way to do this, so this is what I came up with:

output="$(command...)" &t=1d=0while (( "$t" < 120 && "$d" == 0 )); do    sleep 1    if jobs | grep -v 'Done' | grep -v 'Exit 1' | grep 'parse' > /dev/null; then        t="$(( $t + 1 ))"    else        d=1    fidoneif (( "$d" == 0 )); then    ps -AH | grep -A 10 "^ *$(jobs -l | awk '{print $2}')" | grep -B 10 ps | grep -v ps | awk '{print $1}' | while read pid; do        kill -9 "$pid"    done    output=''else    output="$(command..)"fi

My question is, is there a simpler way?

comments powered by Disqus