Comment 0 for bug 303224

Revision history for this message
Gro-Tsen (david-madore) wrote :

Binary package hint: nvidia-glx-177

In a nutshell: garbled Emacs display when running under compiz with nvidia drivers.

More precisely: when Emacs (running in its own X11 window) attempts to refresh the window (e.g., when scrolling up or down a page, jumping to top, control-L or something), it often happens that only the upper portion of the window is refreshed. This also sometimes happens with gnome-terminal (typically when scrolling), but much less reproducibly. With Emacs, it occurs so frequently that it makes the editor almost unusable.

Software & hardware config summary:
        Ubuntu 8.10 Intrepid
        Kernel 2.6.27-9-generic
        x86-64 system (with 64-bit userland) on a Core 2 Duo
        nVidia driver version 177.80 OR 177.82 OR 180.08 (I tried all three)
        nVidia Quadro NVS 290 rev 161 graphics card
        compiz 1:0.7.8-0ubuntu4.1
        Pretty much any Emacs (emacs22-x, emacs22-gtk, emacs-snapshot-gtk)

To reproduce: launch Emacs, in its own window, on some text file (of more than one page), with the compiz window manager running; jump to the bottom of the file (M-<), and then then to the top (M->) or refresh the screen (C-l): quite often, only the upper portion of the window is refreshed, not the bottom, with the limit not necessarily along a line boundary. Only the display is affected: jumping to a different workspace and back will redraw the window correctly.

The bug does not occur with metacity, so it's probably composite-related.

Probably nVidia's fault, because I couldn't produce a similar effect with an Intel 945GME with the X.org/intel driver (and 32-bit but otherwise similar software config). I sent a bug report of similar content to the nVidia team (aka /dev/null).

I tried removing --lose-bindings from the compiz options, and disabling DamageEvents in the xorg.conf, neither seemed to have any effect.

Ubuntu bug #269904 (similar problem affecting Firefox &al) is possibly related to this, but _not identical_ (because the latter bug has a workaround which does not help for this one). Although it's hard to tell since the bug report has been much polluted.

On the other hand, Ubuntu bug #239917 is _not_ related (at least the main part of it is not; the bug I describe is perhaps alluded to).