#250 Potential ~2 pixel frame buffer offset on MGA

Ian Romanick

When running with DRI enabled (with or without any GLX programs
running) the top ~2 scan lines of the screen appear to be corrupt.
They typically contain black pixels, or occasionally fragments of
valid-looking frame-buffer contents (i.e. the colors are right, but
the position is either wrong or weird).

Removing the "glx" and "dri" modules from the XF86Config file fixes
the problem -- the top of the screen looks normal again.

HOWEVER: Doing an "xmag" of the very top of the screen in EITHER
configuration (with or without DRI) results in a Bad Match error from
the X server and a crash in xmag. That is, run xmag, put the mouse
cursor at the very top of the screen and click. I see:

X Error of failed request: BadMatch (invalid parameter attributes)
Major opcode of failed request: 73 (X_GetImage)
Serial number of failed request: 258
Current serial number in output stream: 258
Segmentation fault (core dumped)

So I suppose it's possible that the bug is, in fact, a
double-corrected offset in the XFree 4.0.1 frame buffer handling.
Maybe the DRI code is using the (ostensibly) correct offset and
clobbering the stuff that the XFree driver is putting in the wrong

Any insights? My configuration:

DRI CVS date: 18 Nov 2000
Card: Matrox G400 OEM, 32MB
Kernel: 2.4.0-test11-pre5

Note that I don't have a sourceforge account. Followup email can be
directed to: andy@plausible.org. I'm also a silent member of the
dri-devel list, so I'm likely to notice questions posted there.


  • Ian Romanick
    Ian Romanick

    • assigned_to: nobody --> idr
  • Ian Romanick
    Ian Romanick

    • status: open --> closed
  • Ian Romanick
    Ian Romanick

    Logged In: YES

    Neither the original submitter or I could reproduce this bug
    on DRI more recent than 2001 (or so).