Bug 754951 – gnome-session-3.16.0 crashed

After an evaluation, GNOME has moved from Bugzilla to GitLab. Learn more about GitLab.
No new issues can be reported in GNOME Bugzilla anymore.
To report an issue in a GNOME project, go to GNOME GitLab.
Do not go to GNOME Gitlab for: Bluefish, Doxygen, GnuCash, GStreamer, java-gnome, LDTP, NetworkManager, Tomboy.

Bug 754951 - gnome-session-3.16.0 crashed


Summary:	gnome-session-3.16.0 crashed


Status:	RESOLVED FIXED

Product:	gnome-session
Classification:	Core
Component:	general
Version:	3.16.x
Hardware:	Other Linux

Importance:	Normal normal
Target Milestone:	---
Assigned To:	Session Maintainers
QA Contact:	Session Maintainers

URL:
Whiteboard:

Depends on:
Blocks:

Reported:	2015-09-13 11:17 UTC by Pacho Ramos
Modified:	2016-03-07 22:27 UTC

See Also:
GNOME target:	3.20
GNOME version:	---

Attachments
main: don't use g_setenv after start up (18.01 KB, patch) 2016-03-07 20:53 UTC, Ray Strode [halfline]	committed	Details \| Review

Description Pacho Ramos 2015-09-13 11:17:47 UTC

warning: Could not load shared library symbols for linux-vdso.so.1.
Do you need "set solib-search-path" or "set sysroot"?
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
Core was generated by `gnome-session'.
Program terminated with signal SIGSEGV, Segmentation fault.

+ Trace 235439

Thread 2 (Thread 0x7f014699c980 (LWP 22952))

#0 g_uri_unescape_segment
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/glib/gurifuncs.c line 88
#1 _g_dbus_address_parse_entry
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/gio/gdbusaddress.c line 480
#2 g_dbus_address_try_connect_one
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/gio/gdbusaddress.c line 767
#3 g_dbus_address_get_stream_sync
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/gio/gdbusaddress.c line 968
#4 initable_init
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/gio/gdbusconnection.c line 2581
#5 g_bus_get_sync
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/gio/gdbusconnection.c line 7057
#6 gsm_util_update_activation_environment
at gsm-util.c line 462
#7 gsm_util_setenv
at gsm-util.c line 510
#8 main
at main.c line 350

Thread 1 (Thread 0x7f014699b700 (LWP 22997))

#0 getenv
from /lib64/libc.so.6
#1 ??
from /lib64/libc.so.6
#2 g_cancellable_set_error_if_cancelled
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/gio/gcancellable.c line 314
#3 g_socket_receive_message
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/gio/gsocket.c line 4434
#4 _g_socket_read_with_control_messages_ready
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/gio/gdbusprivate.c line 139
#5 socket_source_dispatch
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/gio/gsocket.c line 3276
#6 g_main_dispatch
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/glib/gmain.c line 3122
#7 g_main_context_dispatch
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/glib/gmain.c line 3737
#8 g_main_context_iterate
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/glib/gmain.c line 3808
#9 g_main_loop_run
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/glib/gmain.c line 4002
#10 gdbus_shared_thread_func
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/gio/gdbusprivate.c line 274
#11 g_thread_proxy
at /var/tmp/portage/dev-libs/glib-2.44.1/work/glib-2.44.1/glib/gthread.c line 764
#12 start_thread
from /lib64/libpthread.so.0
#13 clone
from /lib64/libc.so.6

Comment 1 Ray Strode [halfline] 2015-09-14 17:54:53 UTC

perhaps DBUS_SESSION_BUS_ADDRESS has invalid characters in it?

Comment 2 Pacho Ramos 2015-09-14 18:15:57 UTC

I will try to see the next time it occurs... I can only see that I didn't "manually" messed with that, but I have no idea if maybe something changed its value on its own or similar :/

Comment 3 Matthias Clasen 2016-03-01 21:34:46 UTC

This is just a typical "setenv is not threadsafe" crash - one thread is calling getenv (via gettext) while the other is calling setenv. Boom. We can't do that.

Comment 4 Matthias Clasen 2016-03-03 18:28:07 UTC

Some options we discussed:

- keep a separate environ array in gnome-session and use that when launching apps

- something else I forget now

Comment 5 Ray Strode [halfline] 2016-03-03 19:30:23 UTC

- try putenv instead and see if that doesn't crash because it has some handywavy comment in its man page about being reentrant
- try to convince gettext to stop calling getenv() every time it's run.

Comment 6 Matthias Clasen 2016-03-03 20:55:00 UTC

I sent a mail to Daiki, mentioning this bug

Comment 7 Michael Catanzaro 2016-03-04 19:32:53 UTC

Why do you pin this on gettext? I don't see that in the backtrace.

Isn't it better to not use setenv() after starting threads? Changing just gettext() doesn't really fix anything, that just papers over the problem....

Comment 8 Matthias Clasen 2016-03-04 19:42:37 UTC

The only case of setenv/genenv thread-safety crashes that I have ever seen is this one, where gettenv is called from gettext. Thats why I pin this is on gettext: it is simply the most common caller of getenv, and it is a bit insidious because gettext looks like it _should_ be threadsafe, in particular when it comes in the guise of _()

Comment 9 Ray Strode [halfline] 2016-03-07 20:53:03 UTC

Created attachment 323327 [details] [review]
main: don't use g_setenv after start up

It's not threadsafe and gettext makes us crash, so maintain child
environment separate from process environment.

Comment 10 Ray Strode [halfline] 2016-03-07 20:54:19 UTC

Attachment 323327 [details] pushed as 759b2b5 - main: don't use g_setenv after start up

Comment 11 Michael Catanzaro 2016-03-07 21:13:49 UTC

Nice to see it fixed properly, planning to cherry-pick to gnome-3-20?

Comment 12 Ray Strode [halfline] 2016-03-07 22:27:05 UTC

yea, i'm going to cherry-pick the things over from master before release on monday.  note there were a couple follow up fixes that I accidentally posted to bug 754941