Screen change script argument

[vorph1]

Added a new server cli argument --screen-change-script that is invoked with the screen name as the first parameter.

My main use case for barrier is for VMs with GPU passthrough so I use this to switch my monitor input ie.

#!/bin/bash
case "$1" in
    "WIN10")
        notify-send "Changing to guest output"
        sudo ddcutil -d 1 setvcp 0x60 0x01
        ;;
    "desktop")
        notify-send "Changing to host output"
        sudo ddcutil -d 1 setvcp 0x60 0x03
        ;;
esac

This is for linux server only.

[AdrianKoshka]

So, if I understand, this allows barrier to execute arbitrary code upon the condition of screen change?

[vorph1]

Yup, that's correct

[p12tic]

This is C++ code, use std::system.

[AdrianKoshka]

I don't know how I feel about allowing barrier to execute arbitrary code. What is your opinion @p12tic ?

[vorph1]

It's running a user provided script with user configured screen names so it shouldn't be a security issue if that's your concern. Also since it's only available as a command line argument it's available only to users that run barriers manually so they should know what they're doing :)

[p12tic]

I agree with @vorph1. If attacker has access to command line arguments he already can do whatever he wants. Looking from this side, there's one issue though. info->m_screen as an argument to std::system is not escaped. Something like QProcess would be a better idea instead of std::system.

[AdrianKoshka]

It's not like barrier runs in a privileged mode anyway

[vorph1]

This would add a Qt dependency for the sever app.

[jwestfall69]

You could use execl() instead of system().

[p12tic]

Hm, execl() is linux-only. Maybe something from https://www.boost.org/doc/libs/1_64_0/doc/html/process.html would be a better cross-platform option.

[jwestfall69]

execl() is part of POSIX.

[p12tic]

OK, I missed the part that this is linux-only. Looks good then.

[AdrianKoshka]

Ready to be merged then?

[p12tic]

Yes, let's do it.