Description of problem: After a update from gnome-shell 3.10.4-2@updates to 3.10.4-4@updates-testing, gnome-shell crashes with SIGTRAP showing a "Oh no something has gone wrong" screen, but only if launched by gdm. Mannualy logging in the VT and running a startx works normally. I confirm this problem *disappears* after downgrading to gnome-shell 3.10.4-2 (already tried going back and forward on package versions, behaviour is consistent). Already tried wiping gnome configs, still crashes. I'm unsure which log should I attach. No relevant info is shown in .xsession-errors, neither at Xorg.0.log. Version-Release number of selected component (if applicable): 3.10.4-4 How reproducible: Always if gnome-shell is started from gdm. Non-reproducible if started from startx. Steps to Reproduce: 1. Upgrade to gnome-shell 3.10.4-4 2. Login trough GDM into gnome-shell Actual results: A "Oh no something has gone wrong passing" screen is displayed, no mouse icon is show, altough buttons do react upon mouse hovering. Expected results: A working gnome-shell environment. Additional info: I'm using a three-monitor setup, with two graphic cards. Graphic driver is radeon. Kernel 3.15.0-0.rc5.git4.1.fc21.x86_64. Rest of the system is with FC20 packages.
Launching by gdm does not seem to be relevant. Starting via lightdm has the same effects. On a machine caught by this I got the following in logs: [ 80.720607] traps: gnome-shell[2166] trap int3 ip:316dc504e9 sp:7fff529d64a0 error:0 [ 81.542563] traps: gnome-shell[2240] trap int3 ip:316dc504e9 sp:7fff116d3310 error:0 [ 822.021392] traps: gnome-shell[3269] trap int3 ip:316dc504e9 sp:7fff586017b0 error:0 [ 822.930987] traps: gnome-shell[3326] trap int3 ip:316dc504e9 sp:7fff9a93e310 error:0 The first two are from gnome-shell-3.10.4-4.fc20.x86_64; two which follow are from an attempt to downgrade to gnome-shell-3.10.4-3.fc20.x86_64. Only backing down to gnome-shell-3.10.4-2.fc20.x86_64 makes this to work again. In this case this is as plain setup as it gets. One graphic card and one monitor.
Just in case on a machine I have a gnome-shell which refuses to start I removed _all_ gnome-shell extensions but this did not help at all. In many gnome-shell crashes once I got in logs the following: [ 1717.422050] traps: gnome-shell[3854] trap int3 ip:3cda8504e9 sp:7fffc67f1660 error:0 [ 1717.439453] gnome-settings-[3830]: segfault at 18 ip 000000396941be59 sp 00007fff32ef6d60 error 4 in libgnome-desktop-3.so.8.0.0[3969400000+40000] [ 1719.733770] traps: gnome-shell[3934] trap int3 ip:3cda8504e9 sp:7fff5e1ba320 error:0 I also tried to follow https://wiki.gnome.org/Projects/GnomeShell/Debugging after loading gnome-shell-debuginfo. Clearly I could only attach gdb to a crashed gnome-shell. Following "Acquiring a backtrace" instruction from this page I got this: (gdb) t a a bt Thread 4 (Thread 0x7fee14a72700 (LWP 4617)): #0 0x0000003cd80ea9dd in poll () from /lib64/libc.so.6 #1 0x0000003cda8495b4 in g_main_context_iterate.isra () from /lib64/libglib-2.0.so.0 #2 0x0000003cda849a3a in g_main_loop_run () from /lib64/libglib-2.0.so.0 #3 0x00000034f3ad0376 in gdbus_shared_thread_func () from /lib64/libgio-2.0.so.0 #4 0x0000003cda86ea45 in g_thread_proxy () from /lib64/libglib-2.0.so.0 #5 0x0000003cd8c07f33 in start_thread () from /lib64/libpthread.so.0 #6 0x0000003cd80f4ded in clone () from /lib64/libc.so.6 Thread 3 (Thread 0x7fee0ffff700 (LWP 4618)): #0 0x0000003cd80ea9dd in poll () from /lib64/libc.so.6 #1 0x0000003cda8495b4 in g_main_context_iterate.isra () from /lib64/libglib-2.0.so.0 #2 0x0000003cda8496dc in g_main_context_iteration () from /lib64/libglib-2.0.so.0 #3 0x00007fee14042b7d in dconf_gdbus_worker_thread () from /usr/lib64/gio/modules/libdconfsettings.so #4 0x0000003cda86ea45 in g_thread_proxy () from /lib64/libglib-2.0.so.0 #5 0x0000003cd8c07f33 in start_thread () from /lib64/libpthread.so.0 #6 0x0000003cd80f4ded in clone () from /lib64/libc.so.6 Thread 2 (Thread 0x7fee0effd700 (LWP 4620)): #0 0x0000003cd80ea9dd in poll () from /lib64/libc.so.6 #1 0x0000003cda8495b4 in g_main_context_iterate.isra () from /lib64/libglib-2.0.so.0 #2 0x0000003cda8496dc in g_main_context_iteration () from /lib64/libglib-2.0.so.0 #3 0x0000003cda849729 in glib_worker_main () from /lib64/libglib-2.0.so.0 #4 0x0000003cda86ea45 in g_thread_proxy () from /lib64/libglib-2.0.so.0 #5 0x0000003cd8c07f33 in start_thread () from /lib64/libpthread.so.0 #6 0x0000003cd80f4ded in clone () from /lib64/libc.so.6 Thread 1 (Thread 0x7fee14a73a80 (LWP 4449)): #0 0x0000003cd80ea9dd in poll () from /lib64/libc.so.6 #1 0x0000003cda8495b4 in g_main_context_iterate.isra () from /lib64/libglib-2.0.so.0 #2 0x0000003cda849a3a in g_main_loop_run () from /lib64/libglib-2.0.so.0 #3 0x0000000000409400 in main () (gdb) call gjs_dumpstack () No symbol table is loaded. Use the "file" command. What is expected from 'gjs_dumpstack ()' and how to get that I have no idea. Downgrading to gnome-shell-3.10.4-2.fc20 immediately fixes the whole mess. Note: if further information from me is required I am going away in few days and after that I can run further tests only in the end of July.
(In reply to Michal Jaegermann from comment #2) > Note: if further information from me is required I am going away in few days > and after that I can run further tests only in the end of July. Michal, do you also get the same behaviour when lauching the gnome-session from startx? I have no idea why, but this problem does not happens when I launch it from startx.
Created attachment 901119 [details] gdb stacktrace of crash Here I dumped the stacktrace of the crash, using gnome-shell-3.10.4-3 (which is the first patch that created the bug). It's definitely linked to the magnifying glass feature.
(In reply to Bernardo Donadio from comment #3) > Michal, do you also get the same behaviour when lauching the gnome-session > from startx? I have no idea why, but this problem does not happens when I > launch it from startx. Indeed, you are right. Running startx from a level 3 brings up what looks like a normal desktop session. Still in an output of 'journalctl -xb' I see entries like: gnome-session[2182]: (gnome-settings-daemon:2335): GLib-GObject-CRITICAL **: g_object_unref: assertion 'G_IS_OBJECT (object)' failed There are also entries like this one "Window manager warning: Buggy client sent a _NET_ACTIVE_WINDOW message with a timestamp of 0 for 0x1800008 (gnome-twea)", but that seems to be hardly relevant as I was getting the same gnome-session crashes when gnome-tweak-tool was not even installed. It is not configured to autostart anyway. When doing that from a level 5, with a display-manager.service running, I can find these: May 31 13:03:12 ... gnome-session[5131]: Window manager warning: Log level 6: AT-SPI: COuldn't connect to accessibility bus. Is at-spi-bus-launcher running? May 31 13:03:12 ... kernel: traps: gnome-shell[5186] trap int3 ip:3cda8504e9 sp:7fff04d5de20 error:0 Well, yes, at-spi-bus-launcher is running. It is a process 4867 at my current layout. There is more failed assertions in the second case then in the first one. I attach relevant fragments of journalctl output for a session started with startx and for a bomb.
Created attachment 901120 [details] log messages from journalctl for a ok session started via 'startx'
Created attachment 901128 [details] log messages from journalctl for a crashed gnome-session
This reminds me. There is another bug 1053850 with a gnome-shell crash after "AT-SPI: COuldn't connect to accessibility bus. Is at-spi-bus-launcher running?". Also "traps: gnome-shell[9628] trap int3 ip:31b3a504e9 sp:7fff1b409b60" looks familar (at least "504e9"). That other bug still sits as NEW. Another one is bug 1051559.
Should this be filled upstream?
This should be fixed in gnome-shell-3.10.4-5 now, closing.