notmuch zombie apocolypse

Austin Clements amdragon at MIT.EDU
Wed Jun 12 10:10:05 PDT 2013


TL;DR: This is a bug caused by GTK Emacs 24 and recent glib
(introduced somewhere between 2.33.12 and 2.36.1) when there's no
D-BUS session daemon running.  It was "fixed" in Emacs git two weeks
ago, but you can temporarily work around the problem by running DBUS
or by exporting GSETTINGS_BACKEND=memory.

Quoth David Bremner on Jun 12 at 10:31 am:
> 
> Hi All;
> 
> I've noticed some strange behaviour with zombie processes and
> notmuch-emacs.
> 
> * 0.15.2
> - start large search, quit (with q)
> - zombie is created
> - repeat as many times as you like.
> - first search to complete, all zombies are gone.
> - CPU is not pegged.
> 
> * 0.15.2+174
>   - even searches that complete create zombies
>   - cpu usage of emacs process hits 100%
> 
> Defining the former as "good" and the latter as "bad", 
> bisect reveals 08fde50bf3a4c5 as the first "bad" commit.
> 
> I am running emacs 24.3.1 on Debian jessie, with linux kernel 3.9.5-1
> from unstable.
> 
> The bug is pretty annoying since it runs down the battery on my
> laptop. In a quick test on another machine running Debian wheezy but
> still emacs 24.3.1, I don't see any zombies created at all. I'm not sure
> what the imprtant variable is, libc, glib, kernel...

This is Emacs bug #14474:
  http://lists.gnu.org/archive/html/bug-gnu-emacs/2013-05/msg00705.html
  http://lists.gnu.org/archive/html/bug-gnu-emacs/2013-06/msg00002.html
Paul Eggert found the same cause that David and I painstakingly
tracked down over IRC:
  http://lists.gnu.org/archive/html/bug-gnu-emacs/2013-05/msg00726.html
This workaround was committed to Emacs git master on May 30th in 8dea8fe0:
  http://lists.gnu.org/archive/html/bug-gnu-emacs/2013-05/msg00752.html


More information about the notmuch mailing list