Linux 2.6 backtrace problems

Linux 2.6 backtrace problems

[Timo Sirainen]

[-- Attachment #1: Type: text/plain, Size: 1161 bytes --]

Googling this problem doesn't show many people complaining, so I'm 
wondering if it's just something wrong with my system.

So, simply put:

int main() { abort(); }

Running it and getting backtrace gives:

#0  0xb7ebdee9 in raise () from /lib/tls/libc.so.6
#1  0xb7fcbedc in ?? () from /lib/tls/libc.so.6
#2  0xbffff3f0 in ?? ()
#3  0xb7ebf781 in abort () from /lib/tls/libc.so.6
#4  0x00000000 in ?? ()
#5  0x00000020 in ?? ()
..
#37 0xb8000c40 in ?? () from /lib/ld-linux.so.2
#38 0xb7fe919c in ?? ()
#39 0x08048217 in ?? ()
#40 0xbffff3d8 in ?? ()
#41 0xb7ff23cf in _dl_lookup_versioned_symbol () from /lib/ld-linux.so.2
#42 0xb7ea97f8 in __libc_start_main () from /lib/tls/libc.so.6
#43 0xb7fcbedc in ?? () from /lib/tls/libc.so.6

Not very helpful. Sometimes it gives almost usable backtrace by losing 
only the first 5 functions, but sometimes it's entirely corrupted.

First this happened with Linux 2.6.5, gdb 6.0 and gcc 3.something. I 
went back to 2.4 kernel. Today I thought maybe it would be fixed by 
now, but it still happens with 2.6.9-rc2-mm1, gdb 6.2.1 and gcc 3.3.4 
(from Debian unstable). CPU is Athlon XP.

Is this kernel or gdb problem?

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 186 bytes --]

Re: Linux 2.6 backtrace problems

[Daniel Jacobowitz]

On Mon, Sep 20, 2004 at 06:46:32PM +0300, Timo Sirainen wrote:
> Googling this problem doesn't show many people complaining, so I'm 
> wondering if it's just something wrong with my system.
> 
> So, simply put:
> 
> int main() { abort(); }
> 
> Running it and getting backtrace gives:
> 
> #0  0xb7ebdee9 in raise () from /lib/tls/libc.so.6
> #1  0xb7fcbedc in ?? () from /lib/tls/libc.so.6
> #2  0xbffff3f0 in ?? ()
> #3  0xb7ebf781 in abort () from /lib/tls/libc.so.6
> #4  0x00000000 in ?? ()
> #5  0x00000020 in ?? ()
> ..
> #37 0xb8000c40 in ?? () from /lib/ld-linux.so.2
> #38 0xb7fe919c in ?? ()
> #39 0x08048217 in ?? ()
> #40 0xbffff3d8 in ?? ()
> #41 0xb7ff23cf in _dl_lookup_versioned_symbol () from /lib/ld-linux.so.2
> #42 0xb7ea97f8 in __libc_start_main () from /lib/tls/libc.so.6
> #43 0xb7fcbedc in ?? () from /lib/tls/libc.so.6

You are using a C library which was built without frame pointers and
shipped without debug info.  If this is a Debian system, install the
Debian packages of gdb and libc6-dbg and let us know if it's still
there.


-- 
Daniel Jacobowitz

Re: Linux 2.6 backtrace problems

[Timo Sirainen]

[-- Attachment #1: Type: text/plain, Size: 731 bytes --]

On 21.9.2004, at 16:01, Daniel Jacobowitz wrote:

>> #0  0xb7ebdee9 in raise () from /lib/tls/libc.so.6
>> #1  0xb7fcbedc in ?? () from /lib/tls/libc.so.6
..
> You are using a C library which was built without frame pointers and
> shipped without debug info.  If this is a Debian system, install the
> Debian packages of gdb and libc6-dbg and let us know if it's still
> there.

It does seem to work, but I'm not really happy with that solution 
either. I didn't need to use libc6-dbg with 2.4 kernel and using it 
seems to be a bit of a pain.

It's a bit strange that raise(SIGABRT) works, kill(getpid(), SIGABRT) 
works, etc. It's only abort() that's the problem. But if it's only a 
libc problem I'll submit a Debian bugreport.

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 186 bytes --]

Re: Linux 2.6 backtrace problems

[Daniel Jacobowitz]

On Tue, Sep 21, 2004 at 04:31:53PM +0300, Timo Sirainen wrote:
> On 21.9.2004, at 16:01, Daniel Jacobowitz wrote:
> 
> >>#0  0xb7ebdee9 in raise () from /lib/tls/libc.so.6
> >>#1  0xb7fcbedc in ?? () from /lib/tls/libc.so.6
> ..
> >You are using a C library which was built without frame pointers and
> >shipped without debug info.  If this is a Debian system, install the
> >Debian packages of gdb and libc6-dbg and let us know if it's still
> >there.
> 
> It does seem to work, but I'm not really happy with that solution 
> either. I didn't need to use libc6-dbg with 2.4 kernel and using it 
> seems to be a bit of a pain.
> 
> It's a bit strange that raise(SIGABRT) works, kill(getpid(), SIGABRT) 
> works, etc. It's only abort() that's the problem. But if it's only a 
> libc problem I'll submit a Debian bugreport.

I'm not talking about using LD_LIBRARY_PATH=/usr/lib/debug.  You should
be able to just install the libc6-dbg package and debug the same copy
of glibc you were using before.  Backtraces should "magically" begin to
work.

If that does not happen, please do submit a Debian bug report.

-- 
Daniel Jacobowitz