Hi,
I have a job that runs fine locally, but when I submit it to the GRID (i.e. everything is compiled there), the job fails on the same dataset (some NTUP_EGAMMA):
...
( ERROR ) TBufferFile::Ch... : object of class vector<vector<float> >
read too few bytes: 6 instead of 14
( ERROR ) TUnixSystem::Di... : segmentation violation
Attaching to program: /proc/20166/exe, process 20166
[Thread debugging using libthread_db enabled]
0xffffe410 in __kernel_vsyscall ()
#1 0x57c6ddf3 in __waitpid_nocancel () from /lib/libc.so.6
#2 0x57c1306b in do_system () from /lib/libc.so.6
#3 0x57bd387d in system () from /lib/libpthread.so.0
#4 0x5585a1dd in TUnixSystem::Exec(char const*) ()
from
/direct/usatlas+OSG/atlas_app/atlas_rel/15.8.0/DetCommon/15.8.0/InstallArea/i686-slc5-gcc43-opt/lib/libCore.so
#5 0x5585f6eb in TUnixSystem::StackTrace() ()
from
/direct/usatlas+OSG/atlas_app/atlas_rel/15.8.0/DetCommon/15.8.0/InstallArea/i686-slc5-gcc43-opt/lib/libCore.so
#6 0x5586048d in TUnixSystem::DispatchSignals(ESignals) ()
from
/direct/usatlas+OSG/atlas_app/atlas_rel/15.8.0/DetCommon/15.8.0/InstallArea/i686-slc5-gcc43-opt/lib/libCore.so
#7 0x5586058d in SigHandler(ESignals) ()
from
/direct/usatlas+OSG/atlas_app/atlas_rel/15.8.0/DetCommon/15.8.0/InstallArea/i686-slc5-gcc43-opt/lib/libCore.so
#8 0x55856e42 in sighandler(int) ()
from
/direct/usatlas+OSG/atlas_app/atlas_rel/15.8.0/DetCommon/15.8.0/InstallArea/i686-slc5-gcc43-opt/lib/libCore.so
#9 <signal handler called>
#10 0x585504a0 in TurnonJetMET::CalculateTurnon(TurnonJetMETHistos&, double)
Does anybody have an idea why the read-in of the file fails there, but not locally? Locally valgrind doesn't show me any problems.
The code is compiled C++ based on the SFrame framework. The error occurs in 5.22.00j as well as 5.26.00c, and also on different sites.
Has anybody seen something similar?
Thanks a lot!
Alexander
Received on Tue Aug 31 2010 - 18:38:08 CEST
This archive was generated by hypermail 2.2.0 : Tue Aug 31 2010 - 23:50:01 CEST