I work in the testing of a GUI app for Windows and macOS and I’m having issue with our Windows runners.

In macOS, all the actions/commands stated in the runner are translated as if I do it in my local machine. So my GUI opens and the keybind simulation works as expected.

However, in Windows, seems like the runner works in a “background” level and it’s not capable of opening the GUI application or execute keybinds unless an active user is connected remotely to the runner (so an active session is undergoing).

I found a workaround for opening the app by having an active subprocess in the runner machine with a socket which I instruct to open and close the app. I suspect I could the same for the keybind but this feels… “hacky” / unnecessary.

Is there any solution to this issue? Is there any way to config the runner to work as if an user was already logged into the computer and executed the commands?