On Wed, 27 Sep 2017 15:39:08 +0200, Marc-André Lureau wrote: > > https://bugzilla.redhat.com/show_bug.cgi?id=1285378 > > Is this really solving that bug under wayland? I thought the problem was > that there was no gdk_device_warp() or alternative for relative mouse input. Confirming my fix does not fix the Wayland bug (it even makes it worse). It does fix a different mouse bug reproducible only on x11 server. I will try to fix the Wayland issue (in fact unrelated to this patch) and altogether to disable this warp functionality on Wayland completely. > I need an explanation for the "delta +=". Why not set mouse_delta to 0 after > using the value? gdk_device_warp() is permanent, it moves the mouse pointer forever. mouse_delta_x/y are applied still to the absolute mouse position, before it is relativized to dx/dy. Therefore it needs to be remembered forever. Such one-time application of "delta" compensation could be done for dx/dy but then we would need to apply inverse transform_input() to such delta as gdk_device_warp() is done in screen coordinates and not in virtual machine coordinates needed for dx/dy. > Could you check if you can remove gdk_display_sync() with your patch? I see no difference with both the original code and with my code (using mouse_delta_x/y) whether the gdk_display_sync() call is present or not. BTW today I have found the erratic mouse motion happens for me only with virt-viewer over ssh -X (on local etherner). On local machine the erratic movement does not really happen. Steph _______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/spice-devel