Crash involving TUnixSystem::DispatchSignals(ESignals)

Sometimes my program fails involving TUnixSystem::DispatchSignals(ESignals). The stacktrace is always the following:

55Thread 1 (Thread 0x2b6ba5dddb80 (LWP 131)):
56#0 0x00002b6ba420286d in waitpid () from /lib64/libc.so.6
57#1 0x00002b6ba4194479 in do_system () from /lib64/libc.so.6
58#2 0x00002b6ba41947b0 in system () from /lib64/libc.so.6
59#3 0x00002b6ba1e26a89 in TUnixSystem::StackTrace() () from /cvmfs/cms.cern.ch/slc6_amd64_gcc530/cms/cmssw-patch/CMSSW_8_0_26_patch1/extern
al/slc6_amd64_gcc530/lib/libCore.so
60#4 0x00002b6ba1e2898c in TUnixSystem::DispatchSignals(ESignals) () from /cvmfs/cms.cern.ch/slc6_amd64_gcc530/cms/cmssw-patch/CMSSW_8_0_26_
patch1/external/slc6_amd64_gcc530/lib/libCore.so
61#5
62#6 0x00002b6b9ce51890 in BTagCalibrationReader::eval_auto_bounds(std::string const&, BTagEntry::JetFlavor, float, float, float) const () f
63rom /cvmfs/cms.cern.ch/slc6_amd64_gcc530/cms/cmssw/CMSSW_8_0_26/lib/slc6_amd64_gcc530/libCondToolsBTau.so

The error is not repeated the next time the program is run.

BTagCalibrationReader::eval_auto_bounds(std::string const&, BTagEntry::JetFlavor, float, float, float) const ()

can be seen in

http://cmsdoxygen.web.cern.ch/cmsdoxygen/CMSSW_8_0_26_patch1/doc/html/df/de6/BTagCalibrationReader_8cc_source.html#l00152

Any thoughts what could be the problem?

Viesturs

You have some sort of memory error. Try using valgrind (witht the parameter --suppressions=$ROOTSYS/etc/valgrind-root.supp) to get more information.

Cheers,
Philippe.

Could the error be caused by memory limits on a work node?

it could … it is one of the many possibilities. Did you run your example under valgrind? Was there any useful information?

Interpreting the valgrind output has always been a challenge to me. It contains a lot of obscure references and does not pinpoint to the actual problem - http://vveckaln.web.cern.ch/vveckaln/condor_debug/err139/MC13TeV_W0Jets_2.err

You can reduce the valgrind noise further by adding --suppressions=$ROOTSYS/etc/valgrind-root-python.supp
and --leak-check=no

The output shown does indeed no indicate the issue. It might be a nullptr dereference or something else. To make further progress in understanding the problem you will need to have a debug build of (at least) libCondToolsBTau.so

Cheers,
Philippe.

1 Like

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.