Re: Strange X11 crashes on 5.0

[Michael <macallan%netbsd.org@localhost>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Aug 6, 2009, at 9:43 AM, George Harvey wrote:

> Hi,
>
> I have 5.0 running on an Ultra 5 and I'm seeing some X11 apps, eg.
> sylpheed, firefox, core dump when started from a remote X server  
> running
> on Mac OS X 10.5.7 (leopard). Other apps, eg. xterm, xload, Tcl/Tk,  
> work
> fine. However, the apps that crash when run from the Mac work OK when
> run from X on the Sun console. I've also tried the same apps built on
> other NetBSD systems (sparc & sgimips) and on those platforms they do
> work when run from the Mac. So it looks like the problem is specific  
> to
> apps running on sparc64 with the display on the Mac.
>
> The apps that crash all fail in XListInputDevices so I built a version
> of libXi with debugging symbols and see the following with gdb (using
> gtk-demo as an example):
>
> Core was generated by `gtk-demo'. Program
> terminated with signal 10, Bus error.

Yay, another 64bit bug. And here I thought they finally learned...

> #0  0x000000004130a168 in XListInputDevices (dpy=0x45009000,
>    ndevices=0xffffffffffffafac) at XListDev.c:180
> 180                         B->class = ButtonClass;
> (gdb) bt
> #0  0x000000004130a168 in XListInputDevices (dpy=0x45009000,
>    ndevices=0xffffffffffffafac) at XListDev.c:180
> #1  0x000000004077c4bc in gdk_device_get_state ()
>   from /usr/pkg/lib/libgdk-x11-2.0.so.0
> #2  0x0000000040752870 in gdk_display_open ()
>   from /usr/pkg/lib/libgdk-x11-2.0.so.0
> #3  0x0000000040727c44 in gdk_display_open_default_libgtk_only ()
>   from /usr/pkg/lib/libgdk-x11-2.0.so.0
> #4  0x0000000040ab9fa4 in gtk_init_check ()
>   from /usr/pkg/lib/libgtk-x11-2.0.so.0
> #5  0x0000000040ab9fdc in gtk_init ()
>    from /usr/pkg/lib/libgtk-x11-2.0.so.0
> #6  0x0000000000119af4 in main ()
> (gdb)
>
> At this point I'm getting out of my depth, I can't see anything  
> wrong in
> XListDev but I'm not familiar with X11 internals so I don't know  
> what I
> should be looking for. Any suggestions?

The most common cause for SIGBUS on sparc64 is an unaligned access.  
Happens when someone passes an int * as a long * and that int happens  
to be on the odd end of a 64bit word. On other 64bit architectures  
you'd only get memory corruption out of it so it would probably crash  
later for no obvious reason.

have fun
Michael

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iQEVAwUBSnsOmcpnzkX8Yg2nAQLzhwf/a5lVPeQEVjSiLpZMDhe2UYktsdIMd7po
5w8lkrEIJ1scp7PNso8OxJ/OwSxM0Uzh612Jx53A7Wfcc64Y1EyyeHJ9o8nCTCxY
HvbN7sfe0JOhEhdib57jKnRXgjxIp4d2ZBnYg7sEvPv0wm44ASOI83L/hp+4SRtK
iMVmk+ZPf0bQ+QUJ0nrxRL+HIkGelWzDy8E9NSeMOj2o6S/NsYSMU/yj+C19pPsk
luKfe3pAmKhN7rmRFC2Kf2Teur9wBEskkK97w2//ccqCge+pEFRSLy4qDQwzrDZd
CYzIKiOsYDNqwArgQNr+3nm5f+eKi9V2nXATU3KSpy749CKhYK2QKA==
=2M5D
-----END PGP SIGNATURE-----