Hello,

Eric is travelling all day, so I will answer.

On Tuesday 28 April 2009 08:50:11 Thomas Mueller wrote:
> >> i've got problems with a VirtualFull making bacula-sd "Kaboom!". First
> >> seen with 3.0 beta 2 - then upgraded to svn rev. 5782 with the same
> >> results.
> >
> > The next time, please, if you can, can you report the problem in during
> > the beta testing period...
>
> i tested the betas in my limited test environment and I run the nightly
> regression tests (tmu-*) . So i really tried - but failed - to find this
> kaboom. :)

OK, thanks.  Too bad it didn't show up earlier.  Oh, well that is how things 
seem to work :-(

>
> >> it stops with "Bacula interrupted by signal 7: BUS error" - what does
> >> this mean?
> >
> > This can be a double free or something alignment problem. According to
> > your log, it seems that your debug level is very high (something like
> > 900!!), that can
> > fire this problem.
>
> bacula-sd kabooms with or without debug output.
>
> >> other VirtualFull jobs - also other jobs on the same computer as the
> >> one "kabooming" - work without problems.
> >>
> >> bacula-sd debug log and traceback (don't know how to enable the
> >> "debugging symbols", do i need to pass configure options?):
> >> http://pastebin.com/m3154aaf1
> >
> > Without them, we can do nothing to correct the situation... You need to
> > be sure that binaries are not stripped. Did you compile them yourself ?
>
> found out what went wrong on compiling. DEB_BUILD_OPTIONS had to be
> exported with "nostrip" and not just defined.

I'm not a Debian expert, so that is nice to know -- and it did work.

>
> here the btraceback  (without valgrind): http://pastebin.com/m2b422faf

Yes, the dump is perfectly good.  The only problem is that it is "imposible" 
for the code to fail at that point, so it is starting to look like a compiler 
bug.

The valgrind seems to indicate that the value in jcr (a pointer) was trashed.  
How is a mystery, but a common way is when the compiler does poor 
optimization.  There are also lots of other ways it could get trashed.

>
> >> let me know if you need more information.

At this point, there are several things you can do:

- What version of the compiler are you using?   If > 4.2.4  I recommend 
compiling Bacula with a 4.2 version (we are using 4.2.4).

- can you show the options that were used on the compiler when compiling the 
file src/stored/mac.c  ?  If the optimization has more than -O2   (i.e. -O3 
or greater, reduce it to -O2 or even -O0 (oh zero) for testing).

- If the above fixes the problem, no need to continue in this list.

- Manually run the SD under the debugger, and when it crashes and gdb gets 
control.  Enter the command:

   print jcr
   print *jcr
   print jcr->dcr
   print *(jcr->dcr)

Sorry you are having these problems.  Hopefully we can resolve it shortly.

Regards,

Kern

> >
> > You can get needed information on www.bacula.org -> Support page.
> >
> > We need to know the operating system, and you need to turn on DEVELOPER
> > flag in
> > bacula/src/version.h then something like "make clean; make ; make
> > install"
>
> Operating System: Debian 4.0 Etch
> Bacula Version: SVN Rev. 8752
> DEVELOPER turned on in src/version.h: yes
> "Tape drive": Files on Disk
> Database: MySQL
> The config.out: http://pastebin.com/f1ab4f8a8
>
> > At this time, you don't need to run daemons in debug mode, perhaps you
> > can activate the valgrind mode to see if it detects something. ex:
> > valgrind --leak-check=full ${BACDIRBIN}/bacula-sd -c
> > ${BACDIRCFG}/bacula-sd.conf
>
> valgrind log: http://pastebin.com/f1eeb7fea
>
> - Thomas
>
>
> ---------------------------------------------------------------------------
>--- Register Now & Save for Velocity, the Web Performance & Operations
> Conference from O'Reilly Media. Velocity features a full day of
> expert-led, hands-on workshops and two days of sessions from industry
> leaders in dedicated Performance & Operations tracks. Use code vel09scf
> and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
> _______________________________________________
> Bacula-devel mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/bacula-devel



------------------------------------------------------------------------------
Register Now & Save for Velocity, the Web Performance & Operations 
Conference from O'Reilly Media. Velocity features a full day of 
expert-led, hands-on workshops and two days of sessions from industry 
leaders in dedicated Performance & Operations tracks. Use code vel09scf 
and Save an extra 15% before 5/3. http://p.sf.net/sfu/velocityconf
_______________________________________________
Bacula-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/bacula-devel

Reply via email to