[Pkg-utopia-maintainers] Bug#528145: Bug#528145: dbus start script segmentation fault

[Aurelien Jarno]

On Mon, May 11, 2009 at 07:33:07PM +0200, Michael Biebl wrote:
> Deng Xiyue wrote:
> > Package: dbus
> > Version: 1.2.14-2
> > Severity: grave
> > 
> > Dbus 1.2.14-2 failes to install due to start script segmentation fault,
> > which makes it half-installed and unusable, hence severity grave.
> > Relevant aptitude log as follow:
> > 
> ..
> > Starting system message bus: dbus/etc/init.d/dbus: line 89: 20246 Segmentation fault      start-stop-daemon --start --quiet --pidfile  --user  --exec  -- --system 
> ..
> > Architecture: mipsel (mips64)
> 
> I can confirm this problem on mips(el). It seems to be a toolchain issue on this
> architectures:
> 1.) The 1.2.12-1 package from the archive works fine. If I recompile it with a
> current toolchain on sid, I the get same behaviour.
> 2.) Compiling 1.2.14-2 with the lenny toolchain produces working binaries.
> 
> I had a quick discussion with aurel on irc, which confirms that:
> 
> > [16:56] <mbiebl> aurel32: around?
> > [16:56] <aurel32> mbiebl: for a few seconds
> > [16:56] <mbiebl> need your help with http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=528145
> > [16:57] <mbiebl> looks like dbus-daemon is segfaulting on mipsen
> > [16:57] <mbiebl> a backtrace only shows this
> > [16:57] <mbiebl> [16:35] <mbiebl> (gdb) run
> > [16:57] <mbiebl> [16:35] <mbiebl> Starting program: /root/dbus/bus/dbus-daemon 
> > [16:57] <mbiebl> [16:35] <mbiebl> Program received signal SIGSEGV, Segmentation fault.
> > [16:57] <mbiebl> [16:35] <mbiebl> 0x2aab55e8 in ?? () from /lib/ld.so.1
> > [16:57] <mbiebl> [16:35] <mbiebl> (gdb) bt full
> > [16:57] <mbiebl> [16:35] <mbiebl> #0  0x2aab55e8 in ?? () from /lib/ld.so.1
> > [16:58] <Manoj> Hmm. I take it back. I don't seem to have a .pc file there
> > [16:58] <mbiebl> aurel32: interesting fact is, if I recompile dbus 1.2.12-1 on a up-to-date sid mips system
> > [16:58] <mbiebl> it also segfaults
> > [16:58] <mbiebl> whereas the package from the archive works
> > [16:58] <mrvn> mbiebl: toolchain bug?
> > [16:59] <mbiebl> almost looks like
> > [16:59] <aurel32> mbiebl: I'll try to have a look, but I probably won't have time until a few days
> > [16:59] <aurel32> but it looks like a toolchain bug yes
> > [16:59] <mbiebl> aurel32: thanks
> > [16:59] <mbiebl> please keep me posted
> 

The problem appears as dbus is creating dbus-daemon as a broken pie
binary. Disabling -pie workaround the problem. I don't know yet what has
introduced this regression though.

-- 
Aurelien Jarno	                        GPG: 1024D/F1BCDB73
aurelien at aurel32.net                 http://www.aurel32.net