On Sunday 23 December 2007 14:52, Frank Sweetser wrote:
> Kern Sibbald wrote:
> > Thanks for the debug info. It looks like something is going wrong in the
> > printf routine, possibly because you have longer path names than most
> > regression testers. Though I cannot completely exclude it yet, I don't
> > think that it is anything related to Fedora.
>
> It looks like you're dead on about the path lengths. If I copy the test
> suite to /tmp and run everything from there, it passes normally.
>
> > I'm made a small tweak to the bsnprintf.c routine that may fix the
> > problem (I give it about 20% probability of resolving the problem), and I
> > have committed it to Branch-2.2, so could you update your branch source
> > code and then run the tests/migration-job-test again?
>
> After updating, if I run the tests from the longer path directory, the dir
> appears to get stuck in a 100% utilization loop.
>
> > If it continues to fail, either you will need to run the debugger by hand
> > to get a bit more information for me, or I'll need to be able to login to
> > the system to run it myself. This is a bug that I would definitely like
> > to eliminate before releasing the next version.
>
> It's been awhile since I did much work with gdb, but here's the backtrace
> while it's stuck in that loop:
>
> (gdb) bt
> #0 0x00110402 in __kernel_vsyscall ()
> #1 0x004985e6 in nanosleep () from /lib/libpthread.so.0
> #2 0x08099eb4 in bmicrosleep (sec=60, usec=0) at bsys.c:71
> #3 0x0806caf4 in wait_for_next_job (one_shot_job_to_run=0x0)
> at scheduler.c:130
> #4 0x0804e592 in main (argc=0, argv=0xbfeab924) at dird.c:287
>
> Let me know if there's any other debugger work you'd like done.
Frank,
I cannot reproduce this. It is starting to look like a compiler bug. Could
you do the following things?
1. g++ --version
2. cd <your-regress-directory-where-things-fail>
(this is a directory that was pre-built and where the problem occurs)
cd build/src/lib
(edit Makefile, find the line the begins with CFLAGS =, and change
the -O2 to -O0 (minus oh zero))
cd ../..
make clean
make
make install
cd ..
export REGRESS_DEBUG=1
tests/migration-job-test
That should rebuild the source without optimization. Perhaps it will fix
the problem. I give it about a 10% chance of working. If this doesn't work,
then the only solution is to either add debug code all over the place or to
go in to detail with the debugger ...
Regards,
Kern
-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Bacula-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/bacula-devel
