Double call to TNetXNGFile::Close() leads to a segfault

[veprbl]

Check duplicate issues.

• Checked for duplicates

Description

Double call to TNetXNGFile::Close() leads to a segfault, which is not the case for TFile::Close() in general. This leads to bugs when integrating software together like AIDASoft/DD4hep#1156.

Reproducer

TFile *_file0 = TFile::Open("root://dtn-eic.jlab.org//work/eic2/EPIC/Tutorials/pythia8NCDIS_10x100_minQ2=1_beamEffects_xAngle=-0.025_hiDiv.hepmc3.tree.root");
_file0->Close();
_file0->Close();

ROOT version

ROOT 6.26/06
xrootd-5.4.3

Installation method

from nixpkgs

Operating system

macOS

Additional context

No response

[Axel-Naumann]

Thanks, @veprbl

Copying the stacktrace from the original, linked report:

#6  <signal handler called>
#7  0x00007fe343359b6a in XrdCl::File::Close(XrdCl::ResponseHandler*, unsigned short) () from /opt/software/linux-ubuntu23.04-skylake/gcc-12.3.0/xrootd-5.6.1-4hhj2qxu72su5ur6i2x5as7tabayobgz/lib/libXrdCl.so.3
#8  0x00007fe343359c85 in XrdCl::File::Close(unsigned short) () from /opt/software/linux-ubuntu23.04-skylake/gcc-12.3.0/xrootd-5.6.1-4hhj2qxu72su5ur6i2x5as7tabayobgz/lib/libXrdCl.so.3
#9  0x00007fe3462e24d3 in TNetXNGFile::Close(char const*) () from /opt/software/linux-ubuntu23.04-skylake/gcc-12.3.0/root-6.28.04-zjvl6wz23xtpv2dj4b73bj3yxsws43i4/lib/root/libNetxNG.so.6.28.04
#10 0x00007fe346ac1377 in HepMC3::ReaderRootTree::close (this=0x55cb53cbff00) at /data/spack/stage/wdconinc/spack-stage-hepmc3-3.2.6-m4mzua36zxrtruwv25lqk5yf7iucr3pc/spack-src/rootIO/src/ReaderRootTree.cc:114
#11 0x00007fe34385b7ee in HepMC3::ReaderPlugin::~ReaderPlugin (this=0x55cb53cb5e50, __in_chrg=<optimized out>) at /data/spack/stage/wdconinc/spack-stage-hepmc3-3.2.6-m4mzua36zxrtruwv25lqk5yf7iucr3pc/spack-src/src/ReaderPlugin.cc:68
#12 0x00007fe345e35994 in std::_Destroy<HepMC3::ReaderPlugin> (__pointer=0x55cb53cb5e50) at /usr/include/c++/12/bits/stl_construct.h:151
#13 0x00007fe345e358ec in std::allocator_traits<std::allocator<void> >::destroy<HepMC3::ReaderPlugin> (__p=0x55cb53cb5e50) at /usr/include/c++/12/bits/alloc_traits.h:648
#14 0x00007fe345e350df in std::_Sp_counted_ptr_inplace<HepMC3::ReaderPlugin, std::allocator<void>, (__gnu_cxx::_Lock_policy)2>::_M_dispose (this=0x55cb53cb5e40) at /usr/include/c++/12/bits/shared_ptr_base.h:613
#15 0x00007fe36b71e887 in std::_Sp_counted_base<(__gnu_cxx::_Lock_policy)2>::_M_release() () from /opt/software/linux-ubuntu23.04-skylake/gcc-12.3.0/root-6.28.04-zjvl6wz23xtpv2dj4b73bj3yxsws43i4/lib/root/libRIO.so
#16 0x00007fe356ee7fcd in std::__shared_count<(__gnu_cxx::_Lock_policy)2>::~__shared_count (this=0x7fff8c03a978, __in_chrg=<optimized out>) at /usr/include/c++/12/bits/shared_ptr_base.h:1071
#17 0x00007fe345e21200 in std::__shared_ptr<HepMC3::Reader, (__gnu_cxx::_Lock_policy)2>::~__shared_ptr (this=0x7fff8c03a970, __in_chrg=<optimized out>) at /usr/include/c++/12/bits/shared_ptr_base.h:1524
#18 0x00007fe345e266d4 in std::__shared_ptr<HepMC3::Reader, (__gnu_cxx::_Lock_policy)2>::operator= (this=0x55cb53cb4a48, __r=...) at /usr/include/c++/12/bits/shared_ptr_base.h:1620
#19 0x00007fe345e233f0 in std::shared_ptr<HepMC3::Reader>::operator= (this=0x55cb53cb4a48, __r=...) at /usr/include/c++/12/bits/shared_ptr.h:440
#20 0x00007fe345e1ce84 in dd4hep::sim::HEPMC3FileReader::HEPMC3FileReader (this=0x55cb53cb49d0, nam=...) at /home/wdconinc/git/DD4hep/DDG4/hepmc/HepMC3FileReader.cpp:140
#21 0x00007fe345e1d6ff in (anonymous namespace)::Factory<dd4hep::sim::HEPMC3FileReader, dd4hep::sim::Geant4EventReader*(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> >)>::call(std::string) (a0=...) at /home/wdconinc/git/DD4hep/DDG4/include/DDG4/Factories.h:176
#22 0x00007fe34cccafcc in dd4hep::PluginService::Create<dd4hep::sim::Geant4EventReader*, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > > (id=...) at /home/wdconinc/git/DD4hep/DDCore/include/DD4hep/Plugins.h:102

[Axel-Naumann]

@amadio would you expect XrdClient to survive repeated calls to Close()?

[amadio]

@Axel-Naumann I will see what can be done on our side about this and fix in the next release if I find a good solution.

[linev]

There is very simple fix to make TNetXNGFile::Close() reentrant.
See PR: #13559