Hi,
I look for a mean to kill a process and all its child in a shell script.

With the kill command I can do that:
kill -- -pid_process

that kill a process and all in the same group and its childs.
But that work only with job launch in interactive shell.

Any idea are mostly appreciate.

Cedric

Since any way you are calling the function through a shell script you can write a few lines like

PPID=`pgrep <yourprogram>`
ps -ef | grep $PPID | awk ' { print $2 } '

This gives you a idea.

Havent had the time to check it out

but i think you get the idea.

Thanks for your idea.
I quickly do a sh function, maybe we can simplify it but it's work.


#print all the childs, grandson, great-grandson, ... pids
#warning not use it on your father because you create child and this do never return
get_child_pid_rec () {
local pids=

for pid in $*
do
local child_pids=
local child_pids_rec=

child_pids=$(ps -ef | awk -v ppid_p=$pid 'BEGIN { pids = "" } $3 == ppid_p { pids = pids $2 " " } END { print pids }');
if [[ $child_pids ]]
then
child_pids_rec=$(get_child_pid_rec $child_pids);
fi
pids="$pids $child_pids $child_pids_rec";
done
echo -n "$pids "
}

Can't you do the same by putting a "-" before the PID of the parent when kill'ing it?
Or does that only work for process groups?

No as I say in the first post.

A "simple test" :
THE SCRIPTS:
user ~/tmp/testshell $ more lauch_main.sh
#!/bin/bash
./main.sh &
user ~/tmp/testshell $ more main.sh
#!/bin/bash
./prod.sh & # >&3 &
while :
do
echo "I am in the main"
sleep 10
done
user ~/tmp/testshell $ more prod.sh
#!/bin/sh
i=0
while :
do
echo $i
sleep 1
i=$(($i + 1))
done
user ~/tmp/testshell $

and THE EXECUTION:
user ~/tmp/testshell $ ./launch_main.sh
user ~/tmp/testshell $ ps -ef f | tail
ucdo8305 17849 17846 0 08:36 pts/5 S 0:00 \_ zsh
ucdo8305 17891 1 0 08:37 ? S 0:00 /usr/X11R6/bin/xterm.real -name Terminal -sb -sl 1500 -display 10.193.224.73:0.0
ucdo8305 17894 17891 0 08:37 pts/8 S 0:00 \_ zsh
ucdo8305 18242 17894 0 08:47 pts/8 R 0:00 \_ ps -ef f
ucdo8305 18243 17894 0 08:47 pts/8 S 0:00 \_ more
ucdo8305 18244 17894 0 08:47 pts/8 S 0:00 \_ tail
ucdo8305 18200 1 0 08:46 pts/5 S 0:00 /bin/bash ./main.sh
ucdo8305 18201 18200 0 08:46 pts/5 S 0:00 \_ /bin/sh ./prod.sh
ucdo8305 18241 18201 0 08:47 pts/5 S 0:00 | \_ sleep 1
ucdo8305 18227 18200 0 08:47 pts/5 S 0:00 \_ sleep 10
user ~/tmp/testshell $ ps -ef f| more | head -n1
UID PID PPID C STIME TTY STAT TIME CMD
user ~/tmp/testshell $ kill -- -18200
kill: kill -18200 failed: no such process
user ~/tmp/testshell $ kill 18200 18201
user ~/tmp/testshell $

but if you launch directly ./main.sh & and try to kill the group it works.
I don't know exactly why, IMO we may have a look on the shell (I use bash 3.xxx)

Regards