I just upgraded the last of my 3 computers to NetBSD 1.6, and am having
some troubles with XFree86 4.x.

After some fiddling, I have the old 1280x1024 resolution supported again
(though it oddly tried to use 1400x1050 originally...a very unusual
resolution (^&).

My present main concern is that the display sometimes has garbage on it.
I haven't yet determined if the problem is in the video signal generation
or is actual corruption of the graphics memory (perhaps due to overly
agressive optimizations in XFree86 4.x).

The effects amount to some partial ghosting of graphics in some cases.  In
other cases, it looks like the display is being interlaced, but (say) the
odd frame is always the one being sent out (so the even frames are lost
resulting in garbled graphics).

The problems are not consistant. By flipping back and forth on virtual
consoles, I can eventually get a "clean" X display.  The display stays
"clean" until I flip away and come back...

(Alternatively, exiting and restarting X has the same effect as flipping
away and back, insofar as garbage is concerned.)


The video card is an S3 ViRGE from about 5 years back and never gave me
any problems under XFree86 3.x.  One thing that I noticed in the XFree86
server log is that it seems to be trying to use a RAMDAC, but I recall
under 3.x, XFree86's docs (or config tool or something) explicitly (and
strongly) claimed that I should *NOT* try that with this card (I think
primarily because the card didn't have one...(^&)

Do these problems sound familiar to anyone?  That they only intermittantly
come up suggests to me that the problem lies in some kind of bogus
initialization, rather than in a config option that is fundamentally
wrong.


Also, in an OpenGL window with mostly black background, I've noticed
flashing white pixels.  These seem to be artifacts of the window updating
write-events (not actual corruption of data) since when the image settles
down there are never any stray white pixels that I've seen.  Any
suggestions about that one?

(Oh, and FWIW: OpenGL seems marginally faster in XFree86 4.x with my card.
Whether this is just that the particular card has a better driver in 4.x,
or if moving OpenGL into the server helped that much---or if the OpenGL
support is simply better written---I'm not sure.  But it does feel
faster.)


  ``I probably don't know what I'm talking about.'' --rauch@math.rice.edu