(this is specifically against 1.1.99.903 which isn't in the dropdown).

I've been running this built with everything else in 7.2rc3 under NetBSD
3.1_STABLE with wsmouse for a while and it worked fine.  When I updated to
NetBSD-4.0_BETA2 I ran into this problem.  I poked around in git and didn't see
a fix or another bug report, but it's possible I missed something.

Periodically (sometimes it would take a week, once it happened, even after a
system restart it would happen a few windows open on every startup, then
magically stop after some inconsistant amount of time), the pointer would go to
the left side of the screen and unless I moved it very very slowly at a very
consistant pace, it would get partway right on the screen and then go back over
to the left.

I originally thought something was broken with wsmouse in 4.0_BETA2 so I went in
and started putting debugging printfs into various points in xserver.  When
wsmouse turned out to be working ok, I went through and added more printfs
throughout the server and xf86-input-mouse until I finally found that it was the
acceleration in xf86input.c that was causing this.

What I saw was that the call to miPointerAbsoluteCursor() in
xf86PostMotionEvent() was sending an unreasonable low number for the x
coordinates (something along the lines of -27xxxxxxxx, my last debugging file
with this has since been replaced).  Once it started doing this, it would start
happening often (moving the mouse fast was an easy way to trigger it).

I was able to narrow it down to the code that I've #ifdef'd out in the attached
patch.  It clearly doesn't fix it because it just disabled mouse acceleration.

Unfortuantely, once I started adding the printfs in (without #ifdefing)
everything started working fine again, so I was never really able to get any
furthur.  I was also confused by the bouncing around inside modules and didn't
figure out where local->d[xy]remaind and device->ptrfeed got assigned to go see
if there was something odd there.

My best guess is that something isn't initialized, perhaps something specific to
the wsmouse driver that is in other drivers or other parts of the code, and I'm
hoping that someone more familiar with this than I am will see the report and it
will point out where it is.

Any more information I can offer, please let me know.