How to restart ES from a script?

[cyperghost]

@TMNTturtlguy
connect via SSH
in prompt type ps -a
That list all process with tty1 output

then create the file es-restart with sudo command
sudo touch /tmp/es-restart

check existance with
ls /tmp

into the ps list... take the pid from ES with the highst number (that's the latest processes) and type
kill 1234 your PID here should not be 1234 ;)

after that

ps -a
again... What do you see?

[meleu]

@cyperghost maybe ps -a | grep emulationsta can give a cleaner output for him. :)

[cyperghost]

@meleu well that are only 2 or three more processes - clean enough but thank you ;)

@TMNTturtlguy I'm also not a coder - I made my master degree in chemistry

[TMNTturtlguy]

@cyperghost thanks for explaining!

To answer your question from the previous post - OMX is the screensaver running, part of a new update to ES.

here is my output - when i ran the kill line, it killed the screensaver.

[cyperghost]

yes that's it...

You see so it works...
You killed ES with PID 941

This was terminated

and relaunched as new
ES with PID 1500

meleus script would kill
938 and 941
you can check with (the -e parameter shows list of killed PIDs)

pkill -f -e "/opt/retropie/supplementary.*emulationstation([^.]|$)"

after meleus script all emulationstation PIDs are gone >> dead!

[BuZz]

@cyperghost What's the problem with sending a SIGINT to all processes rather than one ?

@TMNTturtlguy es-restart will get removed by the ES launch script hence why it disappears.

[cyperghost]

@BuZz Well ... You are the coding wizard :)

I think if you show us a way to solve this proper please go on.
Maybe you can make this waterproof? Or tell us something behind the scene of RetroPie?
Your are welcome!

[BuZz]

@cyperghost I'm not sure what the problem is ? :)

[edit] is it still the same issue as in the OP ? I'd have to test, but it should be ok to killall on ES, which would also kill the current running script of course as it's a child process itself.

[cyperghost]

@BuZz ... and you ask me? :D

[TMNTturtlguy]

@cyperghost @meleu

That did it! Success....I am now back to running without getting the errors. @cyperghost your script does work well. I am going to upload the update to github. If you are willing it would be great for you to test. It is easy for you to delete the script after testing if you don't find it useful. Thanks for all the help. Will be interested to see how updates to ES will affect this.

[cyperghost]

@BuZz
The problem is that you can't kill ES with killall command.
It will return to bash prompt and you are not able to restart in graphical mode on your TV (via SSH yes that works)

So I seem to kill the latest child process that forces ES to restart IF a es-restart is present in /tmp

meleus script kills two pids... and then ES is gone like killall.

Maybe now you understand my "EUREKA" ;)

[BuZz]

@cyperghost perhaps he accidently killed the parent bash script. Killing all ES processes or just one shouldn't make a difference. I'll test though.

[cyperghost]

@BuZz
Please test... maybe you understand my EUREKA today, okay?
This was really tricky

[meleu]

@cyperghost @TMNTturtlguy
Some background I would like to share with you guys:

• when you type emulationstation in the shell, it executes /usr/bin/emulationstation

• the /usr/bin/emulationstation is a shell script that calls /opt/retropie/supplementary/emulationstation/emulationstation.sh.

• the /opt/retropie/supplementary/emulationstation/emulationstation.sh calls the /opt/retropie/supplementary/emulationstation/emulationstation and this is the real emulationstation binary.

• the /opt/retropie/supplementary/emulationstation/emulationstation.sh has a logic to restart ES, restart system and shutdown system based on the existence of some specific files in /tmp. The code of this script can be seen here.

• in my failed attempt to restart ES from commandline, I was trying create the /tmp/es-restart and kill only the ES binary. According to the emulationstation.sh it's enough to restart ES. But as we saw, my approach is killing the emulationstation.sh too, and that is what is "crashing".

I'll try something more concise this evening, but if @cyperghost has some free time now and would like to investigate more, my suggestion is to find a way to kill only that very specific binary: /opt/retropie/supplementary/emulationstation/emulationstation.

I used those metacharacters on my pkill regex to try to match some edge cases (like running a custom branch ES or launching ES with some parameter).

P.S.: now that I see @BuZz is aboard, I'm sure that the elegant solution is coming. :)

[BuZz]

@cyperghost I think the problem is just the emulationstation.sh is being killed (which handles the restarting). Should be simple enough to avoid killing that. eg specifying the process name exactly rather than pattern matching.

[cyperghost]

@BuZz Please go on.... I think it is not as easy
Because I manually lauched /opt.../ emulationstation.sh and that did not workd.
@TMNTturtlguy discovered the /tmp thing today +1 for him

[BuZz]

@cyperghost It should be. I would guess at something like

touch /tmp/es-restart; killall -e "/opt/retropie/supplementary/emulationstation/emulationstation"

should be fine.

[cyperghost]

let me test...

[BuZz]

or to work with emulationstation running from any path. (eg with emulationstation-kids etc).

touch /tmp/es-restart; pkill -f "/emulationstation$"

[cyperghost]

Works fine

touch /tmp/es-restart; killall -e "/opt/retropie/supplementary/emulationstation/emulationstation"