On Sat, Oct 19, 2013 at 7:49 AM, Dietmar Maurer <dietmar@xxxxxxxxxxx> wrote: >> I understand. you are sending keysym, and avoid the layout and hw scancode... >> But Spice, in general, is talking to a VM, and currently only via scancodes. >> >> Quite clearly, I understand that using the agent and XTest, input-methdod or >> accessible-technology will avoid the hw layer altogether. We thought about that >> long before you proposed your extension... > > You do not need the agent and directly inject keysyms to solve this. > > All you need is to extend the spice protocol with the ability to send keycodes. > If you have that, you can use the existing code in qemu to translate keysyms > to scancodes (the code is already there for VNC). > > While this is not perfect, it would be far better than what we have now. "Far better", in what way? Wouldn't it have the same limitations, using a us layout -> scancode table? I know it's configurable at qemu side, but that's not per client, so it would be US anyway. And the fact that qemu supports keycode->scancode isn't solving the problem at the spice level. There are other servers than qemu. Also HTML key events are not x11 keysyms afaik, so you'll need an additional table. -- Marc-André Lureau _______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/spice-devel