Crash during initialization if linked against LHAPDF 5.4.0

[Oleg.Zenin@cern.ch]

Sherpa 1.1.2-beta (​https://www.hepforge.org/archive/sherpa/Sherpa-beta.1.1.2.tar.gz) linked with LHAPDF 5.4.0 crashes during initialization (please, see the output in test-5.4.0.log attached).
The build script and an archived test directory (testdir-5.4.0.tgz) are also attached.
The test was run on LCG slc4_ia32_gcc34 platform (SLC4, i386, gcc 3.4.6).

The same test runs without a crash if Sherpa 1.1.2 is linked against LHAPDF 5.3.1.

With best regards,
​Oleg.Zenin@cern.ch

[Oleg.Zenin@cern.ch]

Crash log.

[Oleg.Zenin@cern.ch]

Build script

[Oleg.Zenin@cern.ch]

Archived directory with test scripts

[Frank Siegert]

Hi Oleg,

thanks for your very well documented bug report.

Unfortunately I don't have access to AFS, because I don't have a CERN computing login. That's why I am never able to test Sherpa against the LHAPDF packages on GENSER.

Would it be possible for you to test with a self-compiled LHAPDF according the following procedure? Because that works for me, and I don't see any significant difference to what you did.

cd /tmp
wget http://www.hepforge.org/archive/lhapdf/lhapdf-5.4.0.tar.gz
tar xzf lhapdf-5.4.0.tar.gz
cd lhapdf-5.4.0
./configure --prefix=/tmp/lhapdf-5.4.0/install
make install

cd /tmp
wget http://www.hepforge.org/archive/sherpa/Sherpa-beta.1.1.2.tar.gz
tar xzf Sherpa-beta.1.1.2.tar.gz
cd SHERPA-MC-1.1.2
./configure --enable-lhapdf=/tmp/lhapdf-5.4.0/install
make install

cd /tmp
cd SHERPA-MC-1.1.2/SHERPA/Run/LHC
# modify Run.dat for LHAPDF (cteq6l -> cteq6l.LHpdf, CTEQ6Grid -> PDFsets)
../../../bin/Sherpa # ==> works fine for me
./makelibs          # ==> works fine for me
../../../bin/Sherpa # ==> works fine for me

Cheers, Frank

PS: You are still enabling CLHEP and ROOT in the Sherpa installation, which both don't serve any purpose (anymore). CLHEP has been replaced by HepMC and ROOT was and is only there for internal things which the users don't use.

[Oleg.Zenin@cern.ch]

Output from the standalone example.

[anonymous]

Thank you for the fast reply, Frank.

I've got the same crash in the self-contained example above.
Please see the .log attached.

This might show up with a specific gcc version.
If I'm not mistaken you're using gcc4 which I haven't checked yet.

WBR,
Oleg.

[Frank Siegert]

Hi Oleg,

thanks for testing it again. I have tried the same example also on SLC4 with gcc 3.4.6, and don't see any crash.

$ cat /etc/issue
Scientific Linux SL release 4.5 (Beryllium)
Kernel \r on an \m

$ g++ --version
g++ (GCC) 3.4.6 20060404 (Red Hat 3.4.6-8)
Copyright (C) 2006 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

$ g77 --version
GNU Fortran (GCC) 3.4.6 20060404 (Red Hat 3.4.6-8)
Copyright (C) 2006 Free Software Foundation, Inc.

GNU Fortran comes with NO WARRANTY, to the extent permitted by law.
You may redistribute copies of GNU Fortran
under the terms of the GNU General Public License.
For more information about these matters, see the file named COPYING
or type the command `info -f g77 Copying'.

Do you have any environment variables set or paths included in PATH or LD_LIBRARY_PATH which would cause it to use inconsistent versions of any libraries?

The point at which it crashes (after "Setting vertices...") is completely unrelated to anything LHAPDF, at this point the PDFs are not even initialized yet.

Since I can't reproduce it even on the same operating system, I wonder how to progress. Is there any way that I could get ssh-access to a machine where this problem occurs? I would send you my public ssh key in that case.

Otherwise I could prepare a script which hopefully produces all debugging information I'd need and ask you to run it, and make its logfile available somewhere.

Cheers, Frank

[Oleg.Zenin@cern.ch]

Hi, Frank,

This indeed looks like some vague problem with the SLC4 installation on my workstation as long as everything runs fine on LXPLUS.cern.ch slc4_ia32_gcc34 (should be identical to my local platform) and slc4_amd64_gcc34 machines.

So I think this issue can be left over until anybody manages to reproduce it independently.

Thank you very much.

With best regards,
Oleg.

[Frank Siegert]

Hi Oleg,

I didn't test it on lxplus (unfortunately don't have a CERN account), but on the local SL4 machines here at IPPP. But they should provide the same testing ground.

I'd certainly be interested in a followup, if you still find something out. But as you suggest, I'll mark this as "works for me" in the meantime, until this has been confirmed.

Cheers,
Frank

[anonymous]

it even on the same ​operating system, I wonder ​how to progress. Is there any way that I could get ssh-access to a machine where this problem occurs

[Stefan Hoeche]

Milestone old deleted