Bug 687763 - libxul apps (Firefox, thunderbird) regressed by automatic g_type_init() ctor
libxul apps (Firefox, thunderbird) regressed by automatic g_type_init() ctor
Status: NEW
Product: glib
Classification: Platform
Component: gobject
unspecified
Other Linux
: Normal normal
: ---
Assigned To: gtkdev
gtkdev
:
: 692215 (view as bug list)
Depends on:
Blocks:
  Show dependency tree
 
Reported: 2012-11-06 14:53 UTC by Colin Walters
Modified: 2015-03-30 22:47 UTC (History)
6 users (show)

See Also:
GNOME target: ---
GNOME version: ---


Attachments

Description Colin Walters 2012-11-06 14:53:54 UTC
First, they definitely emit a warning because they use g_slice_set_config to disable the slice allocator, but the gobject-init_ctor() ends up creating a hash table whch initializes gslice now.

There are also possible indications they dlopen()/dlclose() libgobject, but that would be a separate bug if we find it.

Introduced by https://bugzilla.gnome.org/show_bug.cgi?id=686161
Comment 1 Allison Lortie (desrt) (extended vacation) 2012-11-06 15:28:02 UTC
The dlopen()/dlclose() suspicion was based on some bad guessing on my part.  The problem that I was tracking was really a completely unrelated issue in evolution-data-server (as is used by Ubuntu's Thunderbird, via an addon).
Comment 2 Allison Lortie (desrt) (extended vacation) 2013-01-21 17:03:13 UTC
*** Bug 692215 has been marked as a duplicate of this bug. ***
Comment 3 Allison Lortie (desrt) (extended vacation) 2013-01-21 17:03:31 UTC
Copied from other bug:


So here's an idea: we could come up with some symbol name like
'g_get_slice_allocator_config()' or something, and dlsym() that out of the
process on libglib startup.  If the user provides that function, we call it to
determine what our configuration should be.

I think it's probably good to continue to allow this (and Mozilla, for example,
is quite upset to see it gone).  I also can't think of a better way to do it...
Comment 4 Matthias Clasen 2013-01-21 18:09:10 UTC
My opinion: An api to configure gslice is really not a great idea. And if this is just about disabling it, then I think G_SLICE=always-malloc is a fine solution.

That dlsym proposal sounds like the kind of hack that will come back to bite us at some point.

Note You need to log in before you can comment on or make changes to this bug.