The magazine_chain_pop_head() problem

[Yann Droneaud <yann droneaud fr>]

(I've already send this message, but it seems it was lost in moderation)

Hi there,

Is there anybody aware of a random bug in glib's gslice which trigger a
segmentation fault under Linux ?

Specifically in the function magazine_chain_pop_head():

#0  magazine_chain_pop_head (mem_size=440) at gslice.c:492
492	      (*magazine_chunks)->data = chunk->next;

The problem was reported earlier by Mark Ellis <mark mpellis org uk> back in
April, see http://mail.gnome.org/archives/gtk-list/2010-April/msg00016.html

And the RedHat Bugzilla would found at least 64 open bugs matchings "magazine_chain_pop_head":
https://bugzilla.redhat.com/buglist.cgi?quicksearch=magazine_chain_pop_head

Still in the same bugzilla, one can found 159 closed bugs mostly as duplicate.

Affected programs are: evolution, totem, empathy, etc.

Sadly I haven't be able to reproduce the bug at will and didn't manage to found anything looking like
a race condition in the code (but the gslice slab allocator is quite a complex code for me).

So I'm stuck.

If this bug could not be reproduced for debug, debug support should be added/enabled in the code
to try to discover the culprit. Any ideas ? ;)

Regards.

-- 
Yann Droneaud