nautilus crashed with SIGSEGV in g_str_hash()

Status changed to 'Confirmed' because the bug affects multiple users.

I can reproduce this in precise as well. Here is a slightly better backtrace:

1) gdb nautilus
2) run Computer://
3) Double click a partition in the nautilus window
4) Following backtrace:

(gdb) run Computer://
Starting program: /usr/bin/nautilus Computer://
[Thread debugging using libthread_db enabled]
[New Thread 0xb68feb70 (LWP 14782)]
[New Thread 0xb60fdb70 (LWP 14783)]
[New Thread 0xb58d3b70 (LWP 14784)]
Initializing nautilus-gdu extension

GLib-GObject-CRITICAL **: g_value_get_pointer: assertion `G_VALUE_HOLDS_POINTER (value)' failed

Program received signal SIGSEGV, Segmentation fault.
g_str_hash (v=0x0) at /build/buildd/glib2.0-2.31.4.tested/./glib/ghash.c:1661
1661 /build/buildd/glib2.0-2.31.4.tested/./glib/ghash.c: No such file or directory.
 in /build/buildd/glib2.0-2.31.4.tested/./glib/ghash.c
(gdb) bt
#0 g_str_hash (v=0x0)
    at /build/buildd/glib2.0-2.31.4.tested/./glib/ghash.c:1661
#1 0xb7374a38 in g_hash_table_lookup_node (hash_return=<synthetic pointer>,
    key=0x0, hash_table=0x8783ee8)
    at /build/buildd/glib2.0-2.31.4.tested/./glib/ghash.c:385
#2 g_hash_table_lookup (hash_table=0x8783ee8, key=0x0)
    at /build/buildd/glib2.0-2.31.4.tested/./glib/ghash.c:1058
#3 0xb50493b9 in gdu_pool_get_by_object_path (pool=0x832acf0, object_path=0x0)
    at gdu-pool.c:2217
#4 0xb50479b6 in device_changed_signal_handler (proxy=0x87e4090,
    object_path=0x0, user_data=0x832acf0) at gdu-pool.c:1528
#5 0xb7444f8a in g_cclosure_marshal_VOID__POINTER (closure=0x87ce5a0,
    return_value=0x0, n_param_values=2, param_values=0x87ec5c0,
    invocation_hint=0xbfffee90, marshal_data=0x0)
    at /build/buildd/glib2.0-2.31.4.tested/./gobject/gmarshal.c:609
#6 0xb502017a in marshal_dbus_message_to_g_marshaller (closure=0x87ce5a0,
    return_value=0x0, n_param_values=3, param_values=0x87d4048,
    invocation_hint=0xbfffee90, marshal_data=0x0) at dbus-gproxy.c:1734
#7 0xb7442dac in g_closure_invoke (closure=0x87ce5a0, return_value=0x0,
    n_param_values=3, param_values=0x87d4048, invocation_hint=0xbfffee90)
    at /build/buildd/glib2.0-2.31.4.tested/./gobject/gclosure.c:774
#8 0xb74540c5 in signal_emit_unlocked_R (node=0x82be698, detail=2861,
    instance=0x87e4090, emission_return=0x0, instance_and_params=0x87d4048)
    at /build/buildd/glib2.0-2.31.4.tested/./gobject/gsignal.c:3302
#9 0xb745b942 in g_signal_emit_valist (instance=0x87e4090, signal_id=220,
    detail=2861, var_args=0xbffff044 "")
    at /build/buildd/glib2.0-2.31.4.tested/./gobject/gsignal.c:3033
#10 0xb745bad3 in g_signal_emit (instance=0x87e4090, signal_id=220,
    detail=2861)
    at /build/buildd/glib2.0-2.31.4.tested/./gobject/gsignal.c:3090
#11 0xb5020820 in dbus_g_proxy_emit_remote_signal (message=0x8427ce0,
    proxy=<optimized out>) at dbus-gproxy.c:1787
#12 dbus_g_proxy_manager_filter (user_data=0x87c98d0, message=0x8427ce0,
    connection=<optimized out>) at dbus-gproxy.c:1354
#13 dbus_g_proxy_manager_filter (connection=0x87d0568, message=0x8427ce0,
    user_data=0x87c98d0) at dbus-gproxy.c:1195
#14 0xb693e441 in dbus_connection_dispatch ()
   from /lib/i386-linux-gnu/libdbus-1.so.3
#15 0xb50189ad in message_queue_dispatch (source=0x878fb98, call...

Found the problem, it is actually caused by ubuntuone-client (specifically libsyncdaemon). Due to a recent change from bug #859635 , the library is registering a dbus message marshaller for VOID__POINTER, however the type registered is actually VOID__BOXED, which happens to be the same type as DeviceChanged from UDisks (DBusGObjectPath is a boxed type). So there is an unnecessary (it is a standard type, and just maps to g_cclosure_marshal_VOID__POINTER anyway) and incorrect entry in the marshal_table of dbus-glib, and we get the error message as above:

GLib-GObject-CRITICAL **: g_value_get_pointer: assertion `G_VALUE_HOLDS_POINTER (value)' failed

when g_cclosure_marshal_VOID__POINTER is called, and then everything dies because object_path is NULL.

I think the best fix to this would be to just remove the marshaller registration in ubuntuone-client/libsyncdaemon/syncdaemon-authentication.c. (tested a rebuild with it commented out and no crash, but I don't use ubuntuone so didn't test any of that code, but again, it should be unnecessary since it just maps to a standard marshaller).

Status changed to 'Confirmed' because the bug affects multiple users.

bug #908261 is a duplicate (will close it once it's retraced)

This bug was fixed in the package ubuntuone-client - 2.99.1-0ubuntu1

---------------
ubuntuone-client (2.99.1-0ubuntu1) precise; urgency=low

  * New upstream release.
    - Avoid using hint files with size 0. (LP: #825366)
    - Use g_return_if_fail to avoid NULL for important args (LP: #908805)
    - Don't register a new VOID__POINTER marshaller for dbus as it breaks now
      (LP: #908787, LP: #908261)
  * debian/watch:
    - Point at the release milestone
 -- Rodney Dawes <email address hidden> Wed, 04 Jan 2012 09:44:36 -0500