Problems with Debian Lenny

If you followed this blog, you would know that I have been testing this distribution for some time now. It's still yet to be released, but I am generally satisfied with it's quality. Etch was also nice and I still have it installed and updated. However, I needed newer versions of a few applications badly (Erlang, Emacs, Epiphany) and libraries (glib 2.10). It also has some annoying bugs lurking around (like sometimes Xorg fails to resume after waking from hibernate-disk). Most issues I list below about Lenny are either notorious upstream bugs, or are related to my aging hardware, so they shouldn't be taken too seriously.

  • hibernate-disk (1.99-1)+MPD [0.13.2]+via686a [Linux-image-686/2.6.26+17]: sound output doesn't resume after hibernation, and MPD seems to hang with 100% CPU usage until either it is killed or some other application does proper sound output. Fix: either killall mpd&&{sleep 5;mpd;} or aplay/dd something (perhaps even patch mpd). Workaround: put this in a resume script.
  • hibernate-disk+Xorg [1:7.3+18]+S3 Trio3D/2X (not tested on the Virge): the DPMS option (and hence any power saving mode) sometimes gets disabled after a resume from hibernate-disk. Fix: xset +dpms. Workaround: resume script?
  • Flash Player [10.0.12.36]+Epihany [2.22.3] (Firefox?): the base idle CPU usage is gradually increasing after having watched a lot of flash videos even after closing all windows and tabs. I'm not sure if they have ever produced a version in which this nuisance wasn't present... I'm migrating to GNU Gnash (and player2.swf...), so I can only report back later on. Fix: killall epiphany-gecko&&epiphany-gecko & . Workaround: none.
  • Flash Player+via686a sound: infrequently (happened to me about twice) the sound subsystem can die after browsing various flash-enabled sites for extended periods of time (I think I was also playing sound via MPD). Reinstalling the kernel module does not seem to help afterwards. This condition may get triggered by other factors as well, but it's too rare to diagnose. Fix: soft or hard reset. Workaround: install flashblock, unknown.
  • Xorg+any of S3 Trio3D/2X and S3 Virge 375: Xv output is faulty under mplayer for example (and under other software as well), the movie box flickers badly on blue background. Fix: none. Workaround: use X11 output. (Note: this may be due to a hardware quirk)
  • Xorg+S3 Trio3D/2X (not tested on the Virge): when running in 640x512 (the display resolution is quarter of the total virtual screen in this case) and doing intensive video output for an extended amount of time (playing videos) the screen occasionally flashes and shows about three white vertical lines on a black background (looks pretty random) for about half a second. This glitch can also be linked to hard lock-ups, especially prominent under heavy load (like switching windows rapidly and more while watching the video). This is an exceptionally rare event. Fix: reset. Workaround: use a normal resolution. (Note: this may be due to a hardware quirk)
  • Fluxbox+Pidgin+Epiphany+Claws Mail+Flash Player+unknown: very rare (happened to me about twice) weird keyboard focus loss bug after a lot of events trigger at the same time (a new message, a sound effect, very high CPU usage and others at the same time) possibly due to a race condition and improper locking (this is just a guess). ALT-Tab, ALT-Fx (desktop switch) and CTRL-ALT-Fx (character terminals) still work, no other key seems to get through (if I recall right, but the last incident has happened some months ago). Fix: kill Xorg from a character terminal. Workaround: unknown. (Need to test: does sending sighup to fluxbox fix this?)
  • Javascript with Epiphany or Fluxbox or xorg: if a script modifies the lengthy title rapidly in a tight loop than the whole system becomes unresponsive. Moving the mouse pointer is flawless, but no mouse click or key action is obeyed. The only update on the screen is the rapid flickering of the window title on the toolbar. Not even CTRL+ALT+Fx works! I have not tried ALT+CTRL+Backspace, but I predict a similar lack of response. The last action I took was clicking on the small x on the tab bar to close the tab. After about two minutes hung in this zombie state, the tab closed and the dozens of actions I have triggered got activated at once. All ran fine afterwards. Though I'm running an outdated single-core processor, this may still result in a DoS on a newer system. Workaround: unknown. Fix: unknown, but it could help if they set a quota for certain window manager actions in the script engine, or if they added some fair scheduling in the window manager. This is a good question anyhow.

Comments

Popular posts from this blog

Tftp secret of TL-WR740N uncovered

Hidden TFTP of TP-Link routers

Haskell for embedded: C output, compilers, monads, Timber