Wrong screen geometry when using xrandr's `--same-as`

[The-Compiler]

I have three screens:

• Internal laptop screen (eDP)
• Portable USB-C screen (DisplayPort-0)
• External HDMI projector (HDMI-A-0)

With that, I'm trying to set up xrandr so that I have my laptop screen, and the portable screen on the left of it, mirroring what people see on the projector. Thus, I run:

xrandr --output DisplayPort-0 --left-of eDP --auto
xrandr --output HDMI-A-0 --same-as DisplayPort-0 --auto
hc detect_monitors

which leads to xrandr reporting:

eDP connected primary 1920x1080+1920+0
HDMI-A-0 connected 1920x1080+0+0
DisplayPort-0 connected 1920x1080+0+0

and hc list_monitors saying:

0: 1920x1080+1920+0 with tag "2" [FOCUS]
1: 1920x1080+0+0 with tag "1"

however, I can't get barpyrus to show up on my laptop screen anymore: When I run either barpyrus 0 or barpyrus 1, the panel shows up on the left USB-C screen (probably picking DisplayPort and HDMI?). When I do barpyrus 2, I get:

monitor_rect: Cannot parse argument "2": No such monitor: 2
Traceback (most recent call last):
  File "/usr/bin/barpyrus", line 33, in <module>
    sys.exit(load_entry_point('barpyrus==0.0.0', 'console_scripts', 'barpyrus')())
  File "/usr/lib/python3.10/site-packages/barpyrus/mainloop.py", line 45, in main
    conf = get_user_config()
  File "/usr/lib/python3.10/site-packages/barpyrus/mainloop.py", line 39, in get_user_config
    return get_config(user_config_path())
  File "/usr/lib/python3.10/site-packages/barpyrus/mainloop.py", line 24, in get_config
    exec(code, global_vars)
  File "/home/florian/.config/barpyrus/config.py", line 309, in <module>
    bar = main()
  File "/home/florian/.config/barpyrus/config.py", line 261, in main
    x, y, monitor_w, _monitor_h = hc.monitor_rect(monitor)
  File "/usr/lib/python3.10/site-packages/barpyrus/hlwm.py", line 60, in monitor_rect
    x = int(geometry[0])
ValueError: invalid literal for int() with base 10: ''

[t-wissmann]

The monitor handling of lemonbar is just wrong in some cases. If you run barpyrus 0 or barpyrus 1, can you check the -g parameter to lemonbar? The geometry should still be correct, but the panel is shown at the wrong location (e.g. compare -g of lemonbar with what xwininfo reports).

The error message for barpyrus 2 seems correct, since there are only monitors 0 and 1 (only the long exception is unfortunate indeed).

[The-Compiler]

Hmm, true! Indeed -g 1920x16+1920+0 and -g 1920x16+0+0 get passed to lemonbar, but both instances draw on the left monitor. Looks like this is LemonBoy/bar#163 probably.

The exception is a bit unfortunate indeed. Should HLWMInput.__call__ perhaps check the herbstclient exit code and bail out earlier with an exception if it's nonzero? Or are there hlwm commands being called which expect a nonzero exit code?

[t-wissmann]

I'd say __call__ should check the exit code per default. I've implemented the check in 0f78dcd.