I found a rather dirty solution to this problem which obviously needs
more investigation. It seems that the if_name is NULL and than shifted
for one char. Eg. there g_assert(if_name == (NULL+1)) is true in
capture_ui_utils.c:118. Question is, why is if_name NULL and where is it
set?
Will continue debugging it
Radek
On Mon, 2005-03-21 at 12:30 +0100, Radek Vokal wrote:
> I've reproduced this bug on x86_64 machine as well. Running ./ethereal -
> i any -kS segfaults, backtrace on nonoptimized ethereal shows
>
> (gdb) bt
> #0 0x000000354746f532 in strcmp () from /lib64/tls/libc.so.6
> #1 0x000000000041e9ec in get_interface_descriptive_name (if_name=0x1
> <Address 0x1 out of bounds>) at capture_ui_utils.c:144
> #2 0x000000000043e1dc in main_cf_callback (event=13594976,
> data=0x600160, user_data=0xcf6f70) at main.c:1372
> #3 0x00000000004291fc in cf_callback_remove (func=0x400610340) at
> file.c:165
> #4 0x0000000000429c6a in cf_continue_tail (cf=0x600160, to_read=0,
> err=0x100018f40) at file.c:528
> #5 0x0000000000425986 in sync_pipe_do_capture (capture_opts=0x610340,
> is_tempfile=1) at capture_sync.c:516
> #6 0x000000000042468d in capture_read (capture_opts=0x100aa2b60,
> is_tempfile=55, drops_known=177306599, drops=0)
> at capture.c:201
> #7 0x00000000004401a5 in main (argc=0, argv=0x7fbffff938) at
> main.c:2116
>
> Looking into capture_ui_utils.c I really don't know why it started to
> fail while AFAIK there weren't any changes recently :(
>
> Radek
>
> On Fri, 2005-03-18 at 22:31 +0100, Pierre JUHEN wrote:
> > Still looking for the bug.
> >
> > Facts :
> >
> > Ethereal 0.10.9 recompiled in the same environment in x86_64 mode is OK
> > (bug doesn't show up)
> >
> > Ethereal 0.10.10 RPMs for i386 (Fedora Core 3) mode are OK (bug doesn't
> > show up)
> >
> > => bug is specific to the x86_64 version of 0.10.10 realease.
> >
> > Bug does occur only when doing synchronous capture.
> >
> >
> >
> > Gnu debugger analysis :
> >
> > Bug seems to occur when ethereal rebuilds the graphical interface,
> > probably in the menu rebuilding
> > process (gtk/menu.c), seemingly in walk_menu_tree_for_selected_tree_row
> > recursive calls.
> >
> > During this process, it seems that the stack is overwriten 3/5 layers up.
> >
> > Regular backtrace seem to show that it occurs when creating the menu
> > "/Statistics/Service Response Time/ONC-RPC...."
> >
> > This part is quite difficult to understand with the debugger.
> >
> > I need help.
> >
> > Thank you
> >
> > Pierre JUHEN
> >
> >
> >
> > _______________________________________________
> > Ethereal-dev mailing list
> > Ethereal-dev@xxxxxxxxxxxx
> > http://www.ethereal.com/mailman/listinfo/ethereal-dev
--
Radek Vokál <rvokal@xxxxxxxxxx>
OS Systems Engineer
IT executives rate Red Hat #1 for value
http://www.redhat.com/promo/vendor/index.html
