Description of problem: Attempts to logout from a session do not work. Looking at a process table one finds 'metacity' process in an S state (together with other processes either in S or Z but those below metacity are usually "<defunct>"). Moreover from time to time metacity produces a core. With metacity-debuginfo, glib2-debuginfo and gtk2-debuginfo packages installed a backtrace from such core looks as follows: .... Core was generated by `metacity'. Program terminated with signal 6, Aborted. .... (gdb) where #0 0x00007fc5fab65fb5 in raise () from /lib64/libc.so.6 #1 0x00007fc5fab67b23 in abort () from /lib64/libc.so.6 #2 0x0000000000438cbb in meta_bug (format=<value optimized out>) at core/util.c:398 #3 0x0000000000423d7c in x_io_error_handler (xdisplay=<value optimized out>) at core/errors.c:268 #4 0x00000039f9c45654 in _XIOError () from /usr/lib64/libX11.so.6 #5 0x00000039f9c4c430 in _XGetXCBBuffer () from /usr/lib64/libX11.so.6 #6 0x00000039f9c4d2f2 in _XReply () from /usr/lib64/libX11.so.6 #7 0x00000039f9c40db3 in XSync () from /usr/lib64/libX11.so.6 #8 0x00000039f9c1f870 in XCloseDisplay () from /usr/lib64/libX11.so.6 #9 0x00007fc5f4580d37 in ?? () from /usr/lib64/gtk-2.0/modules/libatk-bridge.so #10 0x00007fc5fab6964d in exit () from /lib64/libc.so.6 #11 0x0000000000423d3d in x_io_error_handler (xdisplay=<value optimized out>) at core/errors.c:285 #12 0x00000039f9c45654 in _XIOError () from /usr/lib64/libX11.so.6 #13 0x00000039f9c4c430 in _XGetXCBBuffer () from /usr/lib64/libX11.so.6 #14 0x00000039f9c4d2f2 in _XReply () from /usr/lib64/libX11.so.6 #15 0x00000039f9c2a643 in XGetSelectionOwner () from /usr/lib64/libX11.so.6 #16 0x0000000000dc20a4 in check_manager_window (client=0x1ec6780) at xsettings-client.c:430 #17 0x0000000000dc25ba in _gdk_xsettings_client_process_event ( client=0x1ec6780, xev=0x7fff02ec5890) at xsettings-client.c:597 #18 0x0000000000d9c18b in gdk_xsettings_client_event_filter ( xevent=0x7fff02ec5890, event=0x1fcecf0, data=0x1ec21a0) at gdkevents-x11.c:3034 #19 0x0000000000d96496 in gdk_event_apply_filters (xevent=0x7fff02ec5890, event=0x1fcecf0, filters=0x1ec5400) at gdkevents-x11.c:349 #20 0x0000000000d977b3 in gdk_event_translate (display=0x1ec0020, event=0x1fcecf0, xevent=0x7fff02ec5890, return_exposes=0) at gdkevents-x11.c:988 #21 0x0000000000d9ab64 in _gdk_events_queue (display=0x1ec0020) at gdkevents-x11.c:2299 #22 0x0000000000d9ad3a in gdk_event_dispatch (source=0x1ecaa50, callback=0, user_data=0x0) at gdkevents-x11.c:2359 #23 0x000000000192f3fa in IA__g_main_context_dispatch (context=0x1e9c7e0) at gmain.c:2063 #24 0x0000000001932b00 in g_main_context_iterate (context=0x1e9c7e0, block=1, dispatch=1, self=<value optimized out>) at gmain.c:2696 #25 0x0000000001932fcd in IA__g_main_loop_run (loop=0x1e9c780) at gmain.c:2919 #26 0x000000000042bdf3 in main (argc=1, argv=0x7fff02ec60a8) at core/main.c:479 (gdb) f 26 #26 0x000000000042bdf3 in main (argc=1, argv=0x7fff02ec60a8) at core/main.c:479 479 g_main_loop_run (meta_main_loop); (gdb) list 474 g_free (meta_args.client_id); 475 476 if (!meta_display_open ()) 477 meta_exit (META_EXIT_ERROR); 478 479 g_main_loop_run (meta_main_loop); 480 481 meta_display_close (meta_get_display (), 482 CurrentTime); /* I doubt correct timestamps matter here */ 483 (gdb) f 11 #11 0x0000000000423d3d in x_io_error_handler (xdisplay=<value optimized out>) at core/errors.c:285 285 exit (1); (gdb) list 280 errno, g_strerror (errno), 281 display->name); 282 } 283 284 /* Xlib would force an exit anyhow */ 285 exit (1); 286 287 return 0; 288 } (gdb) f 2 #2 0x0000000000438cbb in meta_bug (format=<value optimized out>) at core/util.c:398 398 abort (); (gdb) list 393 g_free (str); 394 395 meta_print_backtrace (); 396 397 /* stop us in a debugger */ 398 abort (); 399 } 400 401 void 402 meta_warning (const char *format, ...) (gdb) Contrary to a comment at core/errors.c:284 forcing an exit does not seem to be happening. Attaching strace to a stuck metacity process shows only "restart_syscall(<... resuming interrupted call ...>" with no progress. Killing from outside a parent process, i.e. gnome-session which sits in a similiar "restart_syscall()", unstucks the whole pile. Killing whatever is below and still reacts to signals does not show any effects. Version-Release number of selected component (if applicable): metacity-2.23.21-1.fc10 gnome-session-2.23.2.2-3.fc10 glib2-2.17.0-1.fc10 gtk2-2.13.0-2.fc10 How reproducible: cores are dropped sometimes but logout does not work every time
Are you still seeing this problem with latest rawhide? This bug has been triaged
> Are you still seeing this problem No, I do not think to run into that for a while now.