Only set GDK_BACKEND to "x11" if GDK_BACKEND is unset and XDG_SESSION_TYPE is not "wayland"

[prurigro]

The existing logic always sets the GDK_BACKEND variable to x11, which prevents the application from using the wayland backend (or running at all in a wayland compositor if xwayland is unavailable). The reason this logic is present is because windows would fail to centre without it, so we need to keep the behaviour in x11 sessions.

The GDK_BACKEND variable can and typically should remain unset in a wayland session, and the XDG_SESSION_TYPE variable is usually set and can have a value of unspecified, tty, x11, wayland or mir.

The logic in the patch is as follows; if the system has already set the GDK_BACKEND variable, we should assume that it knows what it's doing and keep the existing value. If the GDK_BACKEND variable is empty however, we check to see if the XDG_SESSION_TYPE variable is set to wayland, and if it is we shouldn't touch the GDK_BACKEND variable. If the XDG_SESSION_TYPE variable is not set to wayland though, we can assume we're most likely in an x11 environment in those cases. If we're in an x11 environment, we set GDK_BACKEND to x11 and windows can be centred :)

I tested this in x11 and wayland with xwayland disabled using all combinations of XDG_SESSION_TYPE and GDK_BACKEND being set appropriately, inappropriately and being unset, and got the expected behaviour in all cases.

This resolves #1420

[prurigro]

I just pushed a second commit to this PR after deciding that GDK_BACKEND probably also shouldn't be overridden if XDG_SESSION_TYPE is mir, tty or some future session type. Rather than ensure it isn't wayland, it now ensures it is either unset, unspecified or x11.

I also ran another set of tests to make sure everything is still working as expected.

[leaanthony]

Thanks so much for this! I think this is a more than acceptable way of dealing with the different backends 👍