restart bash script while loop by itself using signal?

lqmsh · 02-10-2022, 10:25 PM

I am trying to get a bash script while loop restart itself using signal properly.
I wrote testing.sh like this below

Code:

#!/bin/bash

function restart {
    $0 "$@"
    exit 0
}
trap restart SIGINT

num=0
while true; do
    echo $num
    ((num=num+1))
    sleep 1s
done

if I run that testing.sh in terminal, as you know it will show incremental number with sleep 1s interval. And if I hit Ctrl+c, the script will restart itself showing the number starting from 0.
Everything seems fine but

Code:

$ pgrep -f testing.sh

shows many PIDs.
Is that a problem?
Because I intended to use it for swaybar status so I can update it by hitting volume control keys, etc.

Turbocapitalist · 02-10-2022, 10:34 PM

The way you have it written, the signal causes the script to spawn another process. So each time you interrupt it, there will be yet another script running. If all you wish to do is to reset the counter variable, then have the function do that, and only that:

Code:

#!/bin/bash

function restart {
    num=0
}

trap restart SIGINT

num=0
while true; do
    echo $num
    ((num=num+1))
    sleep 1s
done

What are you trying to do?

lqmsh · 02-10-2022, 10:52 PM

Quote:

The way you have it written, the signal causes the script to spawn another process.

What I wish to do is to kill itself and start again itself from the beginning.
Showing incremental numbers is just for testing, I will replace it with other command for the status of sway window manager.

NevemTeve · 02-10-2022, 11:33 PM

Your script cannot commit suicide (other than `exit`), but maybe it can execute itself:

Code:

restart() {
    exec "$myself" "${origargs[@]}"
}
myself="$0"
origargs=("$@")
trap restart INT

pan64 · 02-11-2022, 12:22 AM

Quote:

Originally Posted by lqmsh

What I wish to do is to kill itself and start again itself from the beginning.
Showing incremental numbers is just for testing, I will replace it with other command for the status of sway window manager.

This looks similar how daemon processes work (= daemon reload/restart), but usually that means you save the pid of the process and later kill and restart it from outside. Do not reinvent the wheel.

Mechanikx · 02-11-2022, 12:55 AM

Here is my dwm status bar script. Maybe you will find it useful.

Code:

#!/bin/bash

update()
{
    kill -SIGTERM $(jobs -p)
}

trap update SIGTERM

let loop=0

while true
do

    if [[ $loop%3600 -eq 0 ]]; then
        weather=$(curl -s wttr.in?format=1 | awk 'FNR == 1 {print $2}')
        let loop=0
    fi

    date=$(date "+%a %b %d %Y %I:%M:%S %P")
    mem=$(free -h | awk 'FNR == 2 {print $7}')
    load=$(cat /proc/loadavg | awk '{print $1, $2, $3}')
    total=$(df -h | awk 'FNR == 4 {print $2}')
    used=$(df -h | awk 'FNR == 4 {print $3}')
    vol=$(amixer | awk 'FNR == 6 {print $5}')
    devs=""

    for device in $(grep '/dev/sd[b-z][1-9]' /proc/mounts | awk '{print $1}' | cut -d '/' -f3)
    do
        devs+=" [$device]= |"
    done

    if [ -z "$devs" ]; then
        status=" $weather | $vol | $load | $used/$total | $mem | $date"
    else        
        status=" $devs $weather | $vol | $load | $used/$total | $mem | $date"
    fi

    xsetroot -name "$status"
    let loop=$loop+1
    sleep 2 &
    wait

done

And here is how I change the volume:

Code:

super + Up
    amixer sset Master 5%+; kill -SIGTERM $(pgrep dwm-status.sh)

super + Down
    amixer sset Master 5%-; kill -SIGTERM $(pgrep dwm-status.sh)

super + Right
    amixer sset Master toggle; kill -SIGTERM $(pgrep dwm-status.sh)

The important part is to background the 'sleep' command then have the status script wait and if the volume is changed to kill the 'sleep' process so the status will be updated in real time.

My bash scripting isn't the greatest so there's probably room for improvement.

pan64 · 02-11-2022, 01:17 AM

Quote:

Originally Posted by Mechanikx

The important part is to background the 'sleep' command then have the status script wait and if the volume is changed to kill the 'sleep' process so the status will be updated in real time.

I don't know how it will work if you send sigterm during the execution of the loop (and "currently" not sleeping).
probably you will find useful: https://askubuntu.com/questions/5757...ng-bash-script

Mechanikx · 02-11-2022, 01:39 AM

Quote:

Originally Posted by pan64

I don't know how it will work if you send sigterm during the execution of the loop (and "currently" not sleeping).
probably you will find useful: https://askubuntu.com/questions/5757...ng-bash-script

Thanks. I will check it out.

Mechanikx · 02-11-2022, 03:52 AM

Quote:

Originally Posted by pan64

I don't know how it will work if you send sigterm during the execution of the loop (and "currently" not sleeping).
probably you will find useful: https://askubuntu.com/questions/5757...ng-bash-script

I wrote a small script to see what would happen:

Quote:

#!/bin/bash

var='Hello'

update()
{
var='Goodbye'
kill -SIGTERM $(jobs -p)
}

trap update SIGTERM

while true
do
echo $var
done

And this is what happens when it's sent a sigterm:

Code:

...
Hello
Hello
Hello
kill: usage: kill [-s sigspec | -n signum | -sigspec] pid | jobspec ... or kill -l [sigspec]
Goodbye
Goodbye
Goodbye
...

So kill just errors out with a usage message because $(jobs -p) doesn't return a pid (since there's no background process). Otherwise the script functions as it should.

pan64 · 02-11-2022, 04:08 AM

Quote:

Originally Posted by Mechanikx

So kill just errors out with a usage message because $(jobs -p) doesn't return a pid (since there's no background process). Otherwise the script functions as it should.

or it will kill a df, awk or other processes.
by the way, you may also try the pkill command:

Code:

kill -SIGTERM $(pgrep dwm-status.sh)
# replace to
pkill -SIGTERM dwm-status.sh

Mechanikx · 02-11-2022, 04:19 AM

Quote:

Originally Posted by pan64

or it will kill a df, awk or other processes.
by the way, you may also try the pkill command:

Code:

kill -SIGTERM $(pgrep dwm-status.sh)
# replace to
pkill -SIGTERM dwm-status.sh

I don't think I'm getting what you're saying.

Code:

kill -SIGTERM $(pgrep dwm-status.sh)

Sends a sigterm to dwm-status.sh. Which in turn traps that signal and then update() sends a sigterm to any background process launched and in this case, if there is one (might not be sleeping) would be sleep.

I'm unclear as to why this might kill df, awk, or other processes since I'm only killing a bg process and the only one would be sleep.

pan64 · 02-11-2022, 04:25 AM

I suggested to use a single pkill command instead of kill and pgrep together.

The other kill inside your shell script kill -SIGTERM $(jobs -p) may kill other processes too.

boughtonp · 02-11-2022, 06:55 AM

Quote:

Originally Posted by lqmsh

Because I intended to use it for swaybar status so I can update it by hitting volume control keys, etc.

It seems you should elaborate on this original problem. Sway support key bindings and swaybar appears to have stop/continue behaviour already - so what exactly is it you want to do that needs to be different to the way everyone else is doing it?

Mechanikx · 02-11-2022, 10:54 PM

Quote:

Originally Posted by pan64

I suggested to use a single pkill command instead of kill and pgrep together.

The other kill inside your shell script kill -SIGTERM $(jobs -p) may kill other processes too.

I took your advice as well as read that SO thread and made some modifications:

Code:

#!/bin/bash

my_pid="$$"

update()
{
    pkill -P "$my_pid" sleep
}

trap update SIGTERM

let loop=0

while true
do

    if [[ $loop%3600 -eq 0 ]]; then
        weather=$(curl -s wttr.in?format=1 | awk 'FNR == 1 {print $2}')
        let loop=0
    fi

    date=$(date "+%a %b %d %Y %I:%M:%S %P")
    mem=$(free -h | awk 'FNR == 2 {print $7}')
    load=$(cat /proc/loadavg | awk '{print $1, $2, $3}')
    total=$(df -h | awk 'FNR == 4 {print $2}')
    used=$(df -h | awk 'FNR == 4 {print $3}')
    vol=$(amixer | awk 'FNR == 6 {print $5}')
    devs=""

    for device in $(grep '/dev/sd[b-z][1-9]' /proc/mounts | awk '{print $1}' | cut -d '/' -f3)
    do
        devs+=" [$device]= |"
    done

    if [ -z "$devs" ]; then
        status=" $weather | $vol | $load | $used/$total | $mem | $date"
    else        
        status=" $devs $weather | $vol | $load | $used/$total | $mem | $date"
    fi

    xsetroot -name "$status"
    let loop=$loop+1
    sleep 2 &
    wait

done

Code:

super + Up
    amixer sset Master 5%+; pkill -SIGTERM dwm-status.sh

...

I believe now there's no ambiguity regarding which process or processes are killed by update().

Thanks again.