Bug#274756: libc6: abort() breaks gdb backtraces with 2.6 kernel
On Mon, Nov 01, 2004 at 10:39:38PM +0100, Joachim Berdal Haga wrote:
> >>I was about to report the same problem; I started investigating after
> >>seeing too many corrupted backtraces on a mailing list. So this is
> >>a real problem.
> >>
> >>I mention this because I'm afraid the bug will be closed after finding
> >>out why it doesn't work after installing libc6-dbg. It should work
> >>without libc6-dbg as well.
> >
> >
> >Please give a reason for this position. If you want a backtrace,
> >install the debugging package.
>
> There are two reasons, one conceptual and one practical. The conceptual
> reason is that libc6-dbg should be required to debug libc6 issues but not
> to debug not-libc-related issues in other programs. The practical reason
> is that most people don't have libc6-dbg installed and thus their error
> reports are not as good as they could be.
>
> Sure: it's just a matter of installing the package to get good
> backtraces. But it's a bit large (no problem for me, might be a problem
> on a slow connection) and, more importantly, it's not a well known
> solution. The garbage bt is more likely interpreted as a corrupted stack
> (speaking from experience).
I disagree with your conceptual issue. There are two things in
libc6-dbg: one of them is usable for debugging libc6, the other for
debugging applications which use libc6. You've made a good point for
splitting these two into separate packages! Which we probably should
have done before now anyway.
Perhaps after sarge.
The practical issue is indeed an issue of user education. My response
to this bug report should be considered a step in that direction :-)
--
Daniel Jacobowitz
Reply to: