Hi folks,
I ported the ABRT kernel oops detector to journald some time ago, because of NoDefaultSyslog change.
I wanted to do the same with the ABRT Xorg stack trace detector (just because I do not like the current implementation and it is possible now [2]), but I am not able to trigger the Xorg's stack trace dumper. I tried a couple of signals, but all my efforts led to a core dump file caught by the ABRT core dump hook.
I thought I have the 'NoTrapSignals' option set to 'true', but 'grep NoTrapSignals -r /etc/ /usr/share/X11/' returns no results.
Does Xorg handle the fatal signals on its own (it seems it does [3])? If so, how can I trigger it?
Otherwise, I would love to remove the ABRT Xorg stack trace detector from Fedora.
Regards, Jakub
1: http://fedoraproject.org/wiki/Changes/NoDefaultSyslog 2: http://who-t.blogspot.cz/2014/03/viewing-xorglog-with-journalctl.html 3: https://bugzilla.redhat.com/show_bug.cgi?id=1035508#c1
On Thu, Oct 23, 2014 at 02:54:38AM -0400, Jakub Filak wrote:
Hi folks,
I ported the ABRT kernel oops detector to journald some time ago, because of NoDefaultSyslog change.
I wanted to do the same with the ABRT Xorg stack trace detector (just because I do not like the current implementation and it is possible now [2]), but I am not able to trigger the Xorg's stack trace dumper. I tried a couple of signals, but all my efforts led to a core dump file caught by the ABRT core dump hook.
I thought I have the 'NoTrapSignals' option set to 'true', but 'grep NoTrapSignals -r /etc/ /usr/share/X11/' returns no results.
Does Xorg handle the fatal signals on its own (it seems it does [3])?
yes, see OsSighandler() and OsInit(). http://cgit.freedesktop.org/xorg/xserver/tree/os/osinit.c the signal handler calls xorg_backtrace() and eventually abort()
Cheers, Peter
If so, how can I trigger it?
Otherwise, I would love to remove the ABRT Xorg stack trace detector from Fedora.
Regards, Jakub
1: http://fedoraproject.org/wiki/Changes/NoDefaultSyslog 2: http://who-t.blogspot.cz/2014/03/viewing-xorglog-with-journalctl.html 3: https://bugzilla.redhat.com/show_bug.cgi?id=1035508#c1
On Mon, 2014-10-27 at 14:01 +1000, Peter Hutterer wrote:
On Thu, Oct 23, 2014 at 02:54:38AM -0400, Jakub Filak wrote:
Hi folks,
I ported the ABRT kernel oops detector to journald some time ago, because of NoDefaultSyslog change.
I wanted to do the same with the ABRT Xorg stack trace detector (just because I do not like the current implementation and it is possible now [2]), but I am not able to trigger the Xorg's stack trace dumper. I tried a couple of signals, but all my efforts led to a core dump file caught by the ABRT core dump hook.
I thought I have the 'NoTrapSignals' option set to 'true', but 'grep NoTrapSignals -r /etc/ /usr/share/X11/' returns no results.
Does Xorg handle the fatal signals on its own (it seems it does [3])?
yes, see OsSighandler() and OsInit(). http://cgit.freedesktop.org/xorg/xserver/tree/os/osinit.c the signal handler calls xorg_backtrace() and eventually abort()
Thank you!
Is that true, that the Xorg server drop a core file by default in Fedora? I want to make sure I didn't miss any important detail.
It would mean that ABRT can stop watching logs and searching for Xorg backtraces.
Regards, Jakub
Cheers, Peter
If so, how can I trigger it?
Otherwise, I would love to remove the ABRT Xorg stack trace detector from Fedora.
Regards, Jakub
1: http://fedoraproject.org/wiki/Changes/NoDefaultSyslog 2: http://who-t.blogspot.cz/2014/03/viewing-xorglog-with-journalctl.html 3: https://bugzilla.redhat.com/show_bug.cgi?id=1035508#c1