Re: PATCH: Fix "invisible wall" problem with relative mouse mode

["Daniel P. Berrange" <dan berrange com>]

On Tue, Feb 24, 2009 at 08:50:19AM -0600, Anthony Liguori wrote:
> Daniel P. Berrange wrote:
> >On Tue, Feb 24, 2009 at 08:25:33AM -0600, Anthony Liguori wrote:
> >  
> >>Daniel P. Berrange wrote:
> >>    
> >>>A number of bugs have conspired together to make relative mouse mode not
> >>>work very well. The core problem is that when using the relative mode the
> >>>mouse pointer will appear to hit an "invisible wall" before the true edge
> >>>of the screen.
> >>>
> >>>The problems start occur when we do the mouse grab and fall out from 
> >>>that...
> >>>
> >>>- We passed our window as the "confine-to" arg for the gdk_pointer_grab.
> >>>  This meant that the client desktop pointer would stop moving / 
> >>>  generating
> >>>  events when we hit the edge of our window. In relative mode we need to
> >>>  get continuous events no matter where the pointer is, since we're after
> >>>  relative changes, not absolute coords.
> >>> 
> >>>      
> >>What is supposed to fix this is auto warping.  If you detect that you're 
> >>close to the edge of the screen, it should warp you toward the center of 
> >>the window.
> >>
> >>When I first implemented this, I observed that confining the pointer to 
> >>the desktop instead of the window merely made the problem harder to 
> >>reproduce.  You still stop getting pointer events if the mouse hits the 
> >>edge of the screen so under the right circumstances, you'll still hit a 
> >>wall.
> >>
> >>The warping was supposed to be the proper fix although apparently it's 
> >>not working well enough.
> >>    
> >
> >The warping *does* in fact work perfectly, but only if you allow the mouse
> >to move  outside the VNC window. So this patch should enable the behaviour 
> >you originally intended. If we confined the pointer to just the VNC 
> >window, then we would have to change the warping code to watch for hitting
> >the edge of the VNC widget instead of edge of the screen. 
> >
> >To test this was actually working, I tweaked the grab so that the local
> >cursor was not hidden while grabbed - this lets you see when it hits
> >the edge and that it warps back away from it
> >  
> >        if (x == 0) x += 200;
> >        if (y == 0) y += 200;
> >        if (x == (gdk_screen_get_width(screen) - 1)) x -= 200;
> >        if (y == (gdk_screen_get_height(screen) - 1)) y -= 200;
> 
> This looks like a bug to me, no?  Shouldn't it be using the window 
> sizes?  I'm not sure what I was thinking with this one...

In this context, x & y are refering to the pointer position relative to
the root window, since they're initialized from

	int x = (int)motion->x_root;
	int y = (int)motion->y_root;

> If we just changed it to use the window width/height, shouldn't that 
> also fix the problem?

If we also changed x & y to be the window-relative instead of root-relative
coords, then yes, that'd probably work too.

Daniel
-- 
|: http://berrange.com/     -o-    http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org  -o-  http://virt-manager.org  -o-  http://ovirt.org :|
|: http://autobuild.org       -o-         http://search.cpan.org/~danberr/ :|
|: http://freshmeat.net/~danielpb/    -o-   http://gtk-vnc.sourceforge.net :|