On Sep 29, 10:50pm, Brighten Godfrey wrote:
> On several Personal DECstation 5000/20s, a friend and I have noticed
> strange, annoying behavior of the mouse: whenever the system load is high,
> the mouse acts as though it is clicking, at random. (i.e. you move the
> mouse over an xterm and it might select text even if you're not clicking
> the mouse button.) When the system load is low, the mouse acts normally.
> 
> Keyboards, mice, and cables have been swapped, and this behavior has been
> observed on several systems under different window managers. The mouse has
> model #VSXXX-BB (a hockey puck mouse with two rollers), and the keyboard
> is model #LK501-AA. We're running NetBSD 1.3.2.
> 
> Anyone else note this problem? Any solutions, ideas?

  This is a bug in the dtop.c driver - it doesn't check correctly for
"overrun" conditions on the incoming packet data from the serial bus.
This results in the mouse driver getting out of sync and interpreting
the wrong data for the mouse button status and movment.  I had a quick
fix working well enough to detect the overflow and not pass bad data to
the dtop.c mouse handler, but then tried to switch to using DMA instead
of polling.  That is sort of working, but not nearly as reliablly as the
polled method.

  I'm not sure if I still have the work-around for the polling mode
around anywhere.  I don't have the pmax at home connected up at the
moment, which is where I was working on this.  If I still have them and
can dig them up, I can provide the diffs for this if anyone wants to
try them out.  I might have to figure out the changes from scratch again,
assuming I can remember enough about what I did.

  I've also had some sporadic cases where the PROM switches the device
address of the mouse and keyboard.  NetBSD assumes the mouse and keyboard
will be at fixed addresses, and gets very confused.

Michael

-- 
Michael L. Hitch			mhitch@montana.edu
Computer Consultant
Information Technology Center
Montana State University	Bozeman, MT	USA