[jaunty] corrupted console when using usplash

Moving to usplash, as I started seeing this on real hardware as well.

I tried adjusting /etc/usplash.conf (and running update-initramfs) to have xres and yres (it didn't before), but the console corruption still occurs.

This happens to me too. If I switch VTs during the boot (to cause usplash to stop early), the corruption doesn’t happen. I have narrowed down the time of the corruption to while /etc/rc2.d/S30gdm is running. The gdm initscript is supposed to close usplash before loading gdm. One stupid workaround is to put the following executable script into /etc/rc2.d/S29go-away-usplash:

#!/bin/sh
DO_NOT_SWITCH_VT=yes exec /etc/init.d/usplash start

I don’t understand what exactly is wrong with the existing code in the gdm initscript to do the same thing.

I triaged a bug recently with the same symptoms as this (bug 301737), but I assigned it to the kernel. To be honest, the bug could be in usplash, the kernel or Xorg, but I'm not really sure. I'm pretty sure it wouldn't be a GDM bug. When I first looked at that bug, I tried downgrading the likely culprits to their Intrepid versions, but to no avail.

Just a note, usplash hasn't been updated since Intrepid (and it works for most people in Intrepid), which is why I assigned bug 301737 to the kernel

The core bug here is that /lib/init/splash-functions doesn't shut down usplash cleanly, but instead just SIGKILLs it before it has a chance to clean up after itself.

However, upon fixing that I find that the introduction of splash-functions has meant that /etc/init.d/gdm no longer knows how to wait for the active console to change before logging messages, which leaves some visible litter on the console. I'm introducing a new SPLASH_ORIG_CONSOLE environment variable, set by the splash_stop function in splash-functions, which /etc/init.d/gdm can check and use for this purpose.

This bug was fixed in the package usplash - 0.5.26

---------------
usplash (0.5.26) jaunty; urgency=low

  * Fix various compiler warnings.
  * debian/splash-functions: Kill usplash more gracefully in splash_stop;
    introduces new SPLASH_ORIG_CONSOLE environment variable which can be
    checked by other init scripts (LP: #304136).
  * Notify the main loop by means of a self-pipe when SIGALRM is triggered,
    rather than attempting to animate the display from a signal handler
    (thanks to Andrea Gasparini for testing and adjustment; LP: #259761).
  * Sleep for 0.1 second rather than 1 second in init script while waiting
    for usplash to die, and adjust the loop count both there and in
    splash-functions to match the previous behaviour of giving usplash up to
    10 seconds to go away (LP: #313574).

 -- Colin Watson <email address hidden> Tue, 06 Jan 2009 12:00:21 +0000

This bug was fixed in the package gdm - 2.20.8-0ubuntu5

---------------
gdm (2.20.8-0ubuntu5) jaunty; urgency=low

  * Check new SPLASH_ORIG_CONSOLE environment variable before shutting down
    usplash ourselves (LP: #304136).

 -- Colin Watson <email address hidden> Tue, 06 Jan 2009 12:02:13 +0000