xfwm4 segmentation fault in new CVS on Debian Sarge

[Jeff Hill]

Well, g'dang, I must admit I didn't think to check about those old libs as
I had done an apt-get --purge on the deb's, but the old libs were still
there in /usr/lib.

Sadly, removing them didn't solve the problem (I also checked for any
other old, installed Xfce4 stuff). I redid the backtrace; it isn't exactly
identical (ran diff on new backtrace and the one I sent: line 55 changed
argv= to 0xbffffbc4 from 0xbffffe14). That doesn't sound like it'll be
much help to you. I did verify that the backtrace I sent was from core
created two hours ago, so that has to be the right one.

> make a new .xinitrc, and just put -
> /usr/local/bin/xfce-mcs-manager
> /usr/local/bin/xfwm4 --daemon
> xterm
> in it.  you should just get a big grey screen with an xterm (suitably
> decorated with xfwm4 window borders)

Didn't get any decorated borders (plain black edge). Exiting xterm, X did
quit. No crash.

> if that doesn't crash, replace the xterm line with
> /usr/local/bin/xfce4-panel, and see how that works.  you should end up
> with a grey screen and the panel.  the WM will be running too of course,
> so try to launch stuff from the panel to make sure everything there is
> working ok.

Hmm. Now I've managed to launch xterm from the panel and exit xterm
without a crash (still no decorated borders) if I quickly click the xterm
icon when the panel first comes up. Otherwise, putting the mouse pointer
over any part of the panel causes Xfce4 to crash, even if I don't try to
click on anything.

I did get the Mozilla startup screen once when I clicked on the globe icon
quickly, but Xfce4 crashed immediately (half a second maybe) after
displaying the Mozilla startup screen.

I could boot up my Debian Sid installation and run all of these tests and
backtraces again, but the Sarge install is brand new on a clean drive and
I had very little success with the Sid installation.

I guess I should mention that I couldn't build the xfce4-sample-plugin,
but it certainly doesn't seem like that should be causing the problem. The
sample plugin dies during make with:
        sample.c: In function 'xfce_control_class_init';
        sample.c:150: error: 'sample_create_options' undeclared (first
                use in this function)
        sample.c: 150: error: (Earch undeclared identifier is
                reported only once for each function it appears in.)
        make[2]: *** [libsample_la-sample.lo] Error 1
        . . . .

Using gcc version 3.3.3, gnu make 3.80, kernel 2.4.25 from kernel.org

I'm willing to try anything you suggest, but I certainly don't have a clue
what to try on my own. If helpful, I could try to setup dhis and give you
access to the machine with the Xfce4 install on it (nothing important on
the Sarge drive anyway). I've never set up dhis before, but it doesn't
sound too difficult (I have a web server I can use to setup as a dhis
server).

Thanks again for the assist,

Jeff



Brian J. Tarricone said:
> woah... that's quite a scary stacktrace.  unfortunately, that doesn't
> really tell me much about where in xfce4-panel it's crashing, as it's
> deep in freetype, called from fontconfig, called from pango, called from
> gtk, called from the gtk/glib main loop.  yuck.
> 
> that's rather suprising/odd that your original output said that xfwm4
> crashed, but the core file is from xfce4-panel.  i guess perhaps check
> the date on the core file, make sure it makes sense with when you
> experienced the crash.  (i just want to rule out the possibility that
> this was from something earlier and unrelated.)  also i suppose it's
> possible that they both crashed and xfwm4's core file got overwritten
> (though i would think in that case we should get a 'core.PID'-type
> file).
> 
> there's some more invovled troubleshooting you could try, but before i
> get into that, i want to make sure that you don't have any old xfce
> 4.0.x files lying around (especially the libraries).  if you've had any
> .debs installed, remove them.  assuming you've been installing CVS in
> /usr/local, make sure there are no /usr/lib/libxfce* libs (unless the
> .debs install to /usr/local, then it's somewhat moot).
> 
> if that's squared away and you still have problems, give this a try...
> 
> make a new .xinitrc, and just put -
> 
> /usr/local/bin/xfce-mcs-manager
> /usr/local/bin/xfwm4 --daemon
> xterm
> 
> in it.  you should just get a big grey screen with an xterm (suitably
> decorated with xfwm4 window borders), and if you quit the xterm, X
> itself should quit.  if you get a crash, it'll be pretty unambiguous as
> to what crashed.
> 

> 
> thanks for taking the time to debug this.
> 
> 	-brian
> 
> 
> On Tue, 13 Apr 2004, Jeff Hill wrote:
> 
>> Thank you for the quick response.
>>
>> I've tried the startxfce4 script on every version of Xfce4 I've
>> installed.
>> In all cases, Xfce4 did not even present a screen before dieing with
>> error
>> messages similar to those I get when using .xinitrc and startx.
>>
>> Ran ulimit as directed. "file core" returns:
>>         core: Elf 32-bit LSB core file Intel 80386,
>>         version 1 (SYSV), SVR4-style, SR4-style, from 'xfce4-panel'
>>
>> So while I had the error message saying xfwm4 seg fault, it was
>> xfce4-panel, apparently (obviously, I'm out of my depth here).
>>
>> I assum you want only the backtrace and not all the lines gdb spits out
>> at
>> first that start "Reading symbols ...". (I snipped those lines as they
>> would have made this e-mail very much longer -- hope that was right.)
>> So:
>>
>> gdb /usr/local/bin/xfce4-panel core
>> >--snipped GDB copyright, warranty ...--<