Inspired by discussion in #489, where @corranwebster observes that:
But there is something a bit dodgy about _initialize_application_home in that it isn't using self.state_location but a hard-coded path instead: if state_location changes after init then you'll end up with this situation.
And indeed _initialize_application_home should depend on self.state_location as the single point of truth for the location; the current code will create the wrong directory in the event that the user passed state_location when creating the application.
And ... segfault. We still need to skip the new test on PySide6 / Linux. I was optimistically hoping that that wouldn't be the case. (Fixed in c05852e8cedb51c3d400dff3050909f223ff86d8.)
Inspired by discussion in #489, where @corranwebster observes that:
And indeed
_initialize_application_home
should depend onself.state_location
as the single point of truth for the location; the current code will create the wrong directory in the event that the user passedstate_location
when creating the application.This PR fixes that, and adds a regression test.