[PULL] hw/xwin fixes + a warning fix

[Colin Harrison]

Hi,

Looking into it further the code at the end of miPointerSetPosition() maybe
the cause of this...

    /* In the event we actually change screen or we get confined, we just
     * drop the float component on the floor
     * FIXME: only drop remainder for ConstrainCursorHarder, not for screen
     * crossings */
    if (x != trunc(*screenx))
        *screenx = x;
    if (y != trunc(*screeny))
        *screeny = y;

this was introduced at the same time the function switched to doubles.

If I revert those traps and then call the function...

void winEnqueueMotion(int x, int y)
{
  int valuators[2];
  ValuatorMask mask;
  double dx = (double)x;
  double dy = (double)y;

  miPointerSetPosition(g_pwinPointer, POINTER_RELATIVE, &dx, &dy);
  x = trunc(dx);
  y = trunc(dy);
  valuators[0] = x;
  valuators[1] = y;

  valuator_mask_set_range(&mask, 0, 2, valuators);
  QueuePointerEvents(g_pwinPointer, MotionNotify, 0,
                     POINTER_ABSOLUTE | POINTER_SCREEN, &mask);

}

I now get no problem...I could however do better by just using doubles for x
and y in the hw/xwin function headers and get rid of those hacky casts and
trunc()'s. Will now test that + use of POINTER_RELATIVE looks dodgy!

Thanks,
Colin