Corbin, I agree that any transition effect belongs in the compositor. But we have two exclusive methods here to achieve screen rotation 1) the windows are rotated by the compositor-the one thing to note here is that the xserver currently does not allow mouse inputs to be redirected, even if visually the windows are rotated. 2) the xserver itself does the rotation by talking to the driver ( using xrandr)
I guess the best way would be to show the transition using the compositor and then use xrandr to do the rotation (if the driver issue can be fixed and prevent the flicker as Keith pointed out). Comments? On 13-Feb-09, at 10:07 PM, Corbin Simpson <[email protected]> wrote: > Keith Packard wrote: >> On Fri, 2009-02-13 at 19:44 -0800, Bipin George Mathew wrote: >>> Currently, when we use Xrandr to rotate the screen, the screen goes >>> blank for a second or two and then shows the screen in the new >>> orientation. >> >> That's just a bug; there's no reason to have it blank for a long >> period >> of time as we're not setting a new mode, just changing the frame >> buffer >> pointer. Fixing the driver API to keep this from reprogramming the >> video >> hardware would make the switch instantaneous at least, even if it >> doesn't have a cool transition effect. > > After giving it a bit more thought, I'm actually going to have to say > that it *must* be up to a compositor to do any effects. After all, > since > modes aren't square, windows have to be shifted around and such, and > it > just seems like incredibly bad form to do that at any level lower than > the window manager. Also, even on a square mode, the actual viewing > area > that would be consistently viewable would be circular and there > would be > undefined areas of screen space at certain points. > > ~ C. _______________________________________________ xorg mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/xorg
