samba - Jun 2022 - Broken Dependencies?

Hi,

if i understand this right, you will provide new packages?
> -----Urspr?ngliche Nachricht-----
> Von: Michael Tokarev <mjt at tls.msk.ru>
> Gesendet: Sonntag, 29. Mai 2022 18:06
> An: Andrew Bartlett <abartlet at samba.org>; Dirk Laurenz
> <samba at laurenz.ws>; 'sambalist' <samba at
lists.samba.org>
> Betreff: Re: [Samba] Broken Dependencies?
> 
> [Catching up with old(ish) emails, hopefully]
> 
> 25.05.2022 01:10, Andrew Bartlett via samba wrote:
> > On Tue, 2022-05-24 at 14:25 +0300, Michael Tokarev wrote:
> ..
> >> What do you mean "built without backtrace support"?  How
one builds
> >> with or without backtrace support, - I don't know this term.
> ...
> > In short, if libunwind is found on the system at build time, Samba
> > will
> 
> Aha. I didn't know, and apparently no one in the Debian samba team did
> know, either. I've added libunwind-dev build dependency now, and it is
picked
> up by the build procedure. Indeed, in debian we had no stack backtrace
> support in samba, now we do have it. I just tried - killing smbd with
SIGSEGV
> produces a nice backtrace now.
> 
> > do an internal stack unwind that at least gives us a first clue, but
> > ideally we need the full gdb backtrace, ideally by having the
> > gdb_backtrace script packaged so the admin can specify it.
> >
> > Long ago Debian packages used to mail the admin when Samba crashed
> > with a backtrace, I don't know when that stopped happening.
> 
> It is still the same, nothing had changed in this area. But with time,
working
> email setup is less and less common on a typical linux system, so the
reports
> aren't being emailed as often as before. Also, it depends on gdb to be
installed
> to actually *produce* a report in the first place.
> 
> > "We" should add info on installing the right packages and
enabling
> > this to https://wiki.samba.org/index.php/Bug_Reporting to give folks a
> > clue so I stop repeating myself on the lists...
> 
> Actually this wiki page isn't useful now. But the problem is: my first
conclusion
> after trying to find some info about one or another aspect of samba was
that
> the samba wiki is right to useless generally. And after several tries, it
doesn't
> occur to me anymore to think about using samba wiki at all..  hwell..
> 
> /mjt

01.06.2022 08:56, Dirk Laurenz wrote:
> Hi,
> 
> if i understand this right, you will provide new packages?
Yes and no.

The "no" is about your issue. What you face is a stack corruption,
where sack backtrace fails. Samba as shipped on Debian does have
a proper stack backtrace support, but in your case it fails because
stack is corrupt, not because there's no support.

The "yes" part is that sure I will provide new backports, for later
samba releases. Including the fix for the problem you have, once we
know the fix.

I think I'll try to reproduce the prob on a local qemu-emulated arm
machine, as time permits (recently it was a scarce resource).

Thanks,

/mjt