GNOME.org
 

[Evolution] More problems with evolution

Evolution Team:

Again, hello from the Sun Gnome Team.

We have found four additional situations with the evolution 8 
preview release.  The first and the last situations cause core dumps,
the other two seem to be bugs.

The first two and the last occur in both Solaris and Linux.  The third
seemsto only occur in Solaris.  

This is running on the same machine as yesterday's repor (a Sun
Ultra 10 running Solaris 8) and it's the same build of Evolution.

SITUATION #1
------------

Bring up Evolution, click on the "Inbox" icon, and select the
"Settings->Mail Configurations" menu choice.  The tool then pops up a
window saying that "evolution-mail" has crashed.  The window says
"Ooops!  The view for `evolution:/local/Inbox' have died 
unexpectedly.  :-(  This probably means that the mail component has
crashed."

We verified that this problem occurs in both Solaris and Linux.

Stacktrace of the evolution-mail core file follows:

Note that the address 0x42757474 is out of bounds in many of the functions:

--stacktrace #1 begin--

#0  0xfde32f04 in strlen () from /usr/lib/libc.so.1
#1  0xfe34f884 in g_strconcat (
    string1=0x42757474 <Address 0x42757474 out of bounds>) at gstrfuncs.c:157
#2  0xfec6f5e8 in build_hash_key (
    icon=0x42757474 <Address 0x42757474 out of bounds>, 
    subtype=0xfecc1478 "regular") at gnome-stock.c:926
#3  0xfec6f8b0 in lookup (icon=0x42757474 <Address 0x42757474 out of bounds>, 
    subtype=0xfecc1478 "regular", fallback=0) at gnome-stock.c:992
#4  0xfec715b4 in gnome_stock_or_ordinary_button (
    type=0x42757474 <Address 0x42757474 out of bounds>) at gnome-stock.c:1481
#5  0xfebfa0b8 in gnome_dialog_append_button (dialog=0x171b00, 
    button_name=0x42757474 <Address 0x42757474 out of bounds>)
    at gnome-dialog.c:468
#6  0xfebf9688 in gnome_dialog_construct (dialog=0x171b00, 
    title=0xac228 "Evolution Accounts", ap=0xac244) at gnome-dialog.c:261
#7  0x30454 in construct (dialog=0x171b00) at mail-accounts.c:332
#8  0x30e88 in mail_accounts_dialog_new (shell=0xf11f0) at mail-accounts.c:397
#9  0x36b8c in providers_config (uih=0xd81e0, user_data=0x104ba8, 
    path=0x17436c "SetMailConfig") at mail-callbacks.c:950
#10 0xfefb2614 in impl_Bonobo_UIComponent_execVerb (servant=0x13f630, 
    cname=0x17436c "SetMailConfig", ev=0xffbee9d0) at
bonobo-ui-component.c:121
#11 0xff14b4ec in _ORBIT_skel_Bonobo_UIComponent_execVerb (
    _ORBIT_servant=0x13f630, _ORBIT_recv_buffer=0xf0a00, ev=0xffbee9d0, 
    _impl_execVerb=0xfefb2560 <impl_Bonobo_UIComponent_execVerb>)
    at Bonobo-skels.c:2302
#12 0xff0ab7d4 in ORBit_POA_handle_request (recv_buffer=0xf0a00, poa=0xf1af8)
    at orbit_poa.c:507
#13 0xff0afb54 in ORBit_handle_incoming_request (recv_buffer=0xf0a00)
    at server.c:89
#14 0xff0affec in ORBit_handle_incoming_message (recv_buffer=0xf0a00)
    at server.c:159
#15 0xff0541c8 in giop_main_handle_connection (connection=0xf1040)
    at connection.c:1211
#16 0xff1db228 in orb_handle_connection (source=0xefa88, cond=G_IO_IN, 
    cnx=0xf1040) at oaf-mainloop.c:69
#17 0xfe339bb0 in g_io_unix_dispatch (source_data=0xea640, 
    current_time=0xffbeedb8, user_data=0xf1040) at giounix.c:135
#18 0xfe33cc24 in g_main_dispatch (dispatch_time=0xffbeedb8) at gmain.c:656
#19 0xfe33d9a8 in g_main_iterate (block=1, dispatch=1) at gmain.c:877
#20 0xfe33dca0 in g_main_run (loop=0xee718) at gmain.c:935
#21 0xfe91fc64 in gtk_main () at gtkmain.c:476
#22 0xff16e2fc in bonobo_main () at bonobo-main.c:275
#23 0x539bc in main (argc=4, argv=0xffbef034) at main.c:99

--stacktrace #1 end--

The gnome_dialog_construct() function (see #6 in backtrace) reads as
follows:

void
gnome_dialog_construct (GnomeDialog * dialog,
                        const gchar * title,
                        va_list ap)
{
  gchar * button_name;

  if (title)
    gtk_window_set_title (GTK_WINDOW (dialog), title);
 
  while (TRUE) {

    button_name = va_arg (ap, gchar *);

    if (button_name == NULL) {
        break;
    }

    gnome_dialog_append_button( dialog,
        button_name);
  };

  /* argument list may be null if the user wants to do weird things to the
   * dialog, but we need to make sure this is initialized */
  gnome_dialog_init_action_area(dialog);
}

Since the call to gnome_dialog_append_button (#5 in backtrace) is where
the out-of-bounds variable is appearing, it seems that the button_name
is the out of bounds variable.  It is clearly passed in via va_arg.
The calling function, gnome_stock_or_ordinary_button (#4 in backtrace)
calls gnome_dialog_construct as follows:

        gnome_dialog_construct (GNOME_DIALOG (dialog), _("Evolution Accounts"),
                                GNOME_STOCK_BUTTON_OK);

Not sure why GNOME_STOCK_BUTTON_OK (the 3rd argument to gnome_dialog_construct
is
the va_arg one...see above) is causing an out-of-bounds problem.  Any tips on
how to dig further into this issue?

SITUATION #2
------------

This problem occurs when the user clicks on the "Inbox", and then clicks on
the "Compose" button.  The problem only occurs when the mail is not
configured, so a pop-up appears saying "You have not configured the mail
client.  You need to do this before you can send, receive or compose mail.
Would you like to configure it now?".  Then click on the "Yes" button and
an empty window appears.  This doesn't cause any core dump, but it is
clearly a bug.

We verified that this problem occurs in both Linux and Solaris.

SITUATION #3
------------

This problem occurs when the user clicks on the "Calendar" icon.
You cannot see what is being typed when the user enters any data
into the calendar as described below:

In the "Day" and "5 Day" views any data entered into the calendar
is not visible.  It seems to be white text on a white background.
If you highlight the cell after entering data, you can see the
text.

The "Week" and "Month" calendars work okay if you click on a date
and start typing, which causes it to add a new entry.  The text is
white and the background is gray and clear to read.  However, entries
that are carried over from the "Day" or "5 Day" view have the same
display problem as the "Day" and "5 Day" views described in the
paragraph above.  The "Week" view seems to be white text on white
background.  The "Month" view seems to be white text on light gray
background.  The "Month" view is very difficult but not quite
impossible to read.  Again if you highlight the text in the cell
after entering the data, you can see the text.

This problem only occurs on Solaris.  Any tips about how to correct
or look into this situation?

SITUATION #4
------------

While clicking on cells while in the Calendar, using the "Week"
view, a pop-up appears saying "Ooops!  The view for 
`evolution:/local/Calendar' have died unexpectedly.  :-(
This problably means that the calendar component has crashed."

We were able to recreate this problem in Linux and Solaris.  To
recreate the problem simply click in many cells and enter many
values into the "Week" view.  The problem seems to happen after
playing with it this way for about 1 minute.  If you have problems
recreating this situation, we can be more specific about how we
are creating the problem, but it seems pretty easy to create.
We verified that each core dumping situation had the same
traceback.

Stacktrace of the evolution-calendar core file follows:

--stacktrace #2 begin--

#0  0x5ab44 in e_week_view_start_editing_event (week_view=0x12beb0, 
    event_num=5, span_num=0, initial_text=0x0) at e-week-view.c:2775
#1  0x50a34 in e_week_view_event_item_button_release (wveitem=0x187388, 
    event=0xffbee4e8) at e-week-view-event-item.c:758
#2  0x504e0 in e_week_view_event_item_event (item=0x187388, event=0xffbee4e8)
    at e-week-view-event-item.c:670
#3  0xfe6c27bc in gnome_canvas_item_marshal_signal_1 (object=0x187388, 
    func=0x50444 <e_week_view_event_item_event>, func_data=0x0, 
    args=0xffbee158) at gnome-canvas.c:543
#4  0xff20f160 in gtk_signal_real_emit (object=0x187388, signal_id=78, 
    PARAMS=0xffbee158) at gtksignal.c:1492
#5  0xff20a704 in gtk_signal_emit (object=0x187388, signal_id=78)
    at gtksignal.c:552
#6  0xfe6cabc4 in emit_event (canvas=0x15a370, event=0x13a760)
    at gnome-canvas.c:2817
#7  0xfe6cb7c4 in gnome_canvas_button (widget=0x15a370, event=0x13a760)
    at gnome-canvas.c:3034
#8  0xff1a2934 in gtk_marshal_BOOL__POINTER (object=0x15a370, 
    func=0xfe6cb450 <gnome_canvas_button>, func_data=0x0, args=0xffbee778)
    at gtkmarshal.c:30
#9  0xff20f160 in gtk_signal_real_emit (object=0x15a370, signal_id=33, 
    params=0xffbee778) at gtksignal.c:1492
#10 0xff20a704 in gtk_signal_emit (object=0x15a370, signal_id=33)
---Type <return> to continue, or q <return> to quit---
    at gtksignal.c:552
#11 0xff286b0c in gtk_widget_event (widget=0x15a370, event=0x13a760)
    at gtkwidget.c:2860
#12 0xff1a2854 in gtk_propagate_event (widget=0x15a370, event=0x13a760)
    at gtkmain.c:1313
#13 0xff1a0b6c in gtk_main_do_event (event=0x13a760) at gtkmain.c:770
#14 0xff036370 in gdk_event_dispatch (source_data=0x0, 
    current_time=0xffbeedc0, user_data=0x0) at gdkevents.c:2139
#15 0xfefccc24 in g_main_dispatch (dispatch_time=0xffbeedc0) at gmain.c:656
#16 0xfefcd9a8 in g_main_iterate (block=1, dispatch=1) at gmain.c:877
#17 0xfefcdca0 in g_main_run (loop=0xf1850) at gmain.c:935
#18 0xff19fc64 in gtk_main () at gtkmain.c:476
#19 0xfecee2fc in bonobo_main () at bonobo-main.c:275
#20 0x74180 in main (argc=4, argv=0xffbef02c) at main.c:83

--stacktrace #2 end--

Line 2775 of e-week-view.c reads as follows:

        gtk_object_get (GTK_OBJECT (span->text_item),
                        "event_processor", &event_processor,
                        NULL);

gdb shows that the variables has the following values:

   span              $2 = (EWeekViewEventSpan *) 0x183550
   span->text_item   $9 = (GnomeCanvasItem *) 0x402c000
   *span->text_item  Cannot access memory at address 0x402c0000.
   event_processor   $4 = (ETextEventProcessor *) 0x0

CONCLUSION
----------

We hope that the information about OS-independant issues is
helpful, and please let us know if there is any additional
information or help we can provide in these areas.  We are
also hoping for any tips or hints regarding how to correct
the Solaris specific issues.

Thanks much!

Brian Cameron and Laszlo Peter
Sun Gnome Team Engineers
[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]