I mean it this time....

In <arch/i386/isa/pms.c>, in pmsopen(), there is a section:

  #if 0
    pms_dev_cmd(PMS_SET_RES);
    pms_dev_cmd(3);             /* 8 counts/mm */
    pms_dev_cmd(PMS_SET_SCALE21);
    pms_dev_cmd(PMS_SET_SAMPLE);
    pms_dev_cmd(100);           /* 100 samples/sec */
    pms_dev_cmd(PMS_SET_STREAM);             <== THIS MUST BE REMOVED !!!
  #endif

As noted, the PMS_SET_STREAM command doesn't work.  I guess that's why the 
whole section is '#if 0'.  If you comment out that one line, you can '#if 1' 
the whole section and it works fine -- and makes a BIG difference in the way 
the mouse behaves.

Only time will tell if that cures any of the problems I was having.  (It's 
possible that some or all of them may involve an unstable driver for this 
graphics card -- I've been meaning to get another one anyway.)

XFree86 3.3.5 has a fix (#1463) that increases the "mouse event buffer size 
for NetBSD".  I have the 3.3.5 source, but uh, haven't gotten around to 
building it yet.  Seems to me the above would reduce the flow of mouse events 
into the buffer, which would have the same net effect as increasing the buffer 
size.


--Jim Howard  <jiho@mail.c-zone.net>