[Savage40] Small video glitches with savage snapshot

[Bjorn Knutsson]

On 8 Apr 2005 12:57, Alex Deucher wrote:
> 
> --- Bjorn Knutsson <bjorn+savage at update.uu.se> wrote:
> > On 8 Apr 2005 06:24, Alex Deucher wrote:
> > > --- Bjorn Knutsson <bjorn+savage at update.uu.se> wrote:
> > > > I downloaded and installed:
> > > > 
> > > >
> > >
> >
> http://dri.freedesktop.org/snapshots/savage-20050407-linux.i386.tar.bz2
> > > > 
> > > > I disabled BCIforXv (from xorg.0.log):
> > > > 
> > > > (**) SAVAGE(0): Option: Disable use of the BCI for Xv
> > > > 
> > > > But both glitches are still present.
> > > > 
> > > > As before, if I drop Tim's 1.1.27t instead, I do not see these
> > > > glitches.
> > > > 
> > > > The glitch relating to the top lines I only seem to get in xine,
> > > > while
> > > > the pattern happens in both xine and mplayer. I should note,
> > however,
> > > > that xine is my preferred viewer, so I've "tested" this much more
> > > > exhaustively with xine. I'll switch to using mplayer for a while
> > and
> > > > see if I catch it there.
> > > > 
> > > > The shape of the pattern depends on the width of the video
> > stream,
> > > > and
> > > > will be resized if you resize the content of the window, i.e.,
> > it's
> > > > like it's part of the original video, rather than a glitch that
> > > > overwrites the current window.
> > > > 
> > > > For one width, it looks somthing like this:
> > > > 
> > > > ****     ****     ****
> > > >     ****     ****     ****
> > > >         ****     ****     ****
> > > >             ****     ****     ****
> > > > 
> > > > across the video window. For another width, they're boxes:
> > > > 
> > > >     ****    ****    ****    ****
> > > >     ****    ****    ****    ****
> > > >     ****    ****    ****    ****
> > > >     ****    ****    ****    ****
> > > > 
> > > > One easy way to provoke them to appear is to scroll a
> > firefox-window
> > > > next to the video window. The exact bit pattern of the pattern
> > will
> > > > change as you scroll, which makes me think that the pattern is
> > > > actually a rectangle of the current screen that for some reason
> > gets
> > > > mapped into the video window, with different skews depending on
> > the
> > > > video window width. (We get the rectangles if the video window
> > width
> > > > is an integral multiple of the rectangle width.)
> > > 
> > > Can you try disabling the DRI and see if that fixes the video
> > problem?
> > > comment out the load dri line in the modules section of your
> > config.
> > > It may be a FIFO/bandwidth issue that shows up when 3D is
> > configured. 
> > 
> > I commented the line out, same thing.
> > 
> > > Also what chip do you have?
> > 
> > 01:00.0 VGA compatible controller: S3 Inc. SuperSavage IX/C SDR (rev
> > 05)
> > 
> > It's a T23.
> 
> I dunno.  The supersavages are the least supported since we know the
> least about them.  I dunno what might have changed that would cause the
> problem.  the Xv code is pretty much the same.  It may be a tiling
> issue.  try adding:
> option "disabletile" "true"
> to your config to use a linear framebuffer.

I'll try this and get back to you.

To amend my earlier report, I've now seen garbling of other windows,
e.g., Firefox, while running video. I've attached a small snapshot of
part of my screen, hopefully the list software isn't configured to
drop attachments.

The blue window to the left is the video window (mplayer), while the
rest is part of the Firefox window. The green bar just under the tabs
is similar to the corruption I previously saw in the Xine window, and
its appearance correlates with the video window being present. I've
never seen anything like it with the old 1.1.27t driver.

I guess it could be a problem somewhere else but Xv, but the problems
only seem to occur in conjunction with video windows being present
onscreen.

/Bjorn