Bugs/Issues in GKrellM or elsewhere that affect GKrellM

• Can't change fonts in gtk+-2.4 gkrellm had been using GdkFont which is eventually to be phased out and that may have contributed to an overlooked bug in gtk. Regardless, as of gkrellm 2.2.0, Pango is used to draw text and this fixes the font changing problem. After installing 2.2.x, subsequent compiles of current versions of some plugins will generate a few warnings which in all cases I know of so far are harmless.
• gkrellm won't go in the slit with gtk+-2.4 You can read about the problem here . I've made a change in 2.2.0 based on this and now have some feedback that it is fixed in Blackbox under Gtk 2.4.1. But if anyone still has problems, let me know.
• I'm not sure if this is fixed in kernel 2.4.23, but if you run noflushd under Linux it's possible to get great quantities of zombies. It seems there is a kernel bug where nr_queued_signals can reach nr_max_signals and no more signals can be queued. Then child threads stop being reclaimed by the thread manager and zombies start to grow. The solution is to not run noflushd or upgrade the kernel. Do a google groups search on gkrellm noflushd and sort by date.
• Gkrellm may segfault if you have the gkacpi2 version 0.4 plugin installed and this plugin is part of some plugin packages so many may be affected. If you run gkrellm and get output similar to this:

  $ gkrellm
  No support for device type: ac_adapter.
  No support for device type: battery.
  No support for device type: button.
  No support for device type: fan.
  No support for device type: processor.
  No support for device type: thermal_zone.
  devices:
  0 PFN1
  1 PFN0
  Segmentation fault
  

  you will have to remove the gkacpi2 plugin before gkrellm can run. You will need to remove probably one of these:

      /usr/lib/gkrellm2/plugins/gkacpi2.so
      ~/.gkrellm2/plugins/gkacpi2.so
      /usr/local/lib/gkrellm2/plugins/gkacpi2.so
  

  The segfault is in the plugin init function, so it happens even if the plugin is not enabled.
• Some Linux SMP machines see high CPU usage when running the gkrellmd server or running gkrellm with an inet monitor. This problem does not seem to appear on single processor Linux kernels. One user with 70% usage with a SMP kernel on a 2.4GHz P4 Xeon sent me this rather interesting test:

  time cat /proc/net/tcp > /dev/null
  real  0m0.746s
  user  0m0.000s
  sys   0m0.752s
  

  So, it's a kernel issue and the only gkrellm fix can be to slow down reads of /proc/net/tcp which you can now do in gkrellm 2.1.8. If you run gkrellmd, you can slow down these reads with the inet-interval option in gkrellmd.conf. If you run a normal gkrellm, there is a new Inet update interval option.

• 2.1.6 segfaulted when reading disk data from /proc/partitions for major 8 drives sdb, sdc, ... and this is fixed in 2.1.7.
• Sensor data (temperatures, fans, voltages) from a gkrellmd server will not be recognized if the gkrellm client is running under a locale where the decimal point character is a comma instead of a period. This is fixed in 2.1.6.
• With some session managers multiple gkrellms may be started at session logins even though only one gkrellm was running when the last session ended. Most reports of this are in KDE 3 but it can sometimes happen under Gnome if there is not a clean session logout. As of gkrellm 2.1.5 these multiple gkrellm instances will be prevented by a file lock which should serve for now. The lock is not enforced when using the -p or -t options so developers won't be inconvenienced.
• When setting the skip taskbar and skip pager options, some window manager versions may not recognize the states when gkrellm restarts. As of version 2.1.3, gkrellm properly sets these properties prior to window mapping, however this continues to be a problem. So in an attempt to get these setting to work, gkrellm 2.1.10 will by default send extra _NET_WM_STATE messages at startup even though the EWMH spec at freedesktop.org says this should not be necessary. For gkrellm versions 2.1.5 through 2.1.9, to enable this behaviour you had to manually edit the set_wm_state_hack option in your user_config and this option is now gone. Some historical info on this issue is in this Bugzilla report.
• Sensor temperature readings can be wrong (the "limit" value is read instead of the current value) if you are using Gkrellweather plugin versions earlier than 2.0.4 and you are using a locale where the decimal point is a comma. Those Gkrellweather versions can mangle the i18n locale so that the decimal point character is altered from a comma to a period and this confuses the temperature sensor reading. The solution is to disable Gkrellweather or upgrade to a version >= 2.0.5 You may have to cycle gkrellm restarts and config saving to get the sensors config cleaned up.
• Some report the -w flag doesn't work and gkrellm won't go into the Fluxbox or Blackbox slit. However, it is working fine for most people. Note that giving the -w flag is ignored if you have selected the option "Set window type to be a dock or panel" in the General->Properties config. But selecting that Properties option will not work if your window manager version has not been updated to recognize the EWMH window manager hints. When window managers catch up to the spec, the right way to specify a dock mode will be the Properties option, but until then as a fallback, use -w without selecting the option.
• Theme transparency might not work with Fluxbox or Blackbox. For these window managers that can set a background in a window that covers up the root window, you need to set your background with something that sets the real root background such as Esetroot, wmsetbg, or xv -root, etc. Similarly, programs such as Nautilus can cover up the root pixmap so transparency won't work.
• GKrellM version 2.1.1 could not compile a gkrellmd server using glib 1.2. Fixed in 2.1.2.
• The gkrellmd version 2.0.3 had a big honkin' memory leak. You should have upgraded by now.