On Friday 12 December 2014 16:11:25 Thomas Lübking wrote: > On Freitag, 12. Dezember 2014 08:00:45 CEST, Martin Gräßlin wrote: > > I'd suggest to do a platform check as on Wayland it cannot work > > (grab keyboard fails). > > You're certainly right in that the guarding is entirely superfluous on > wayland, but grabbing still "works". Despite the platform window > ::setKeyboardGrabEnabled() returns false for wayland, QWidget simply > ignores that and assigns the grabber.
oh that's much better than I thought. I only knew that the QPA prints out warnings like "not supported" so I expected that code would just break. > > What's worse: looking up the Qt code, QWidget only maintains the grabber as > static variable, the grabbing state is never re-tested. > > Eeeewww... this is gonna be more complex, I fear. Maybe one could say that misconfiguring the X-Server is out of scope. So if it's grabbed it's assumed to stay grabbed. > > I wonder whether the grabbing state can actually be tested except by > approaching a grab from a sidearm process. In doubt, the only possible > hardening would be to continuously - and the only test whether it worked > would be to invoke QWindow::setMouseGrabEnabled(bool grab) as well. QWindow::setMouseGrabEnabled should be fine. The QXcbWindow implementation looks properly to me, like actually checking whether the grab succeeded. > > Stay tuned ;-) > > Cheers, > Thomas
signature.asc
Description: This is a digitally signed message part.
