[C++] Segfault in json-integration-test

[asfimport]

I've upgraded to Ubuntu 18.04.1 and now I get segfaults in json-integration-test:

(gdb) run
Starting program: /home/antoine/arrow/cpp/build-test/debug/json-integration-test 
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
[==========] Running 2 tests from 1 test case.
[----------] Global test environment set-up.
[----------] 2 tests from TestJSONIntegration
[ RUN      ] TestJSONIntegration.ConvertAndValidate

Program received signal SIGSEGV, Segmentation fault.
std::string::_Rep::_M_is_leaked (this=this@entry=0xffffffffffffffe8)
    at /home/msarahan/miniconda2/conda-bld/compilers_linux-64_1507259624353/work/.build/x86_64-conda_cos6-linux-gnu/build/build-cc-gcc-final/x86_64-conda_cos6-linux-gnu/libstdc++-v3/include/bits/basic_string.h:3075
3075	/home/msarahan/miniconda2/conda-bld/compilers_linux-64_1507259624353/work/.build/x86_64-conda_cos6-linux-gnu/build/build-cc-gcc-final/x86_64-conda_cos6-linux-gnu/libstdc++-v3/include/bits/basic_string.h: Aucun fichier ou dossier de ce type.
(gdb) bt
#0  std::string::_Rep::_M_is_leaked (this=this@entry=0xffffffffffffffe8)
    at /home/msarahan/miniconda2/conda-bld/compilers_linux-64_1507259624353/work/.build/x86_64-conda_cos6-linux-gnu/build/build-cc-gcc-final/x86_64-conda_cos6-linux-gnu/libstdc++-v3/include/bits/basic_string.h:3075
#1  0x00007ffff7311856 in std::string::_Rep::_M_grab (this=0xffffffffffffffe8, __alloc1=..., __alloc2=...)
    at /home/msarahan/miniconda2/conda-bld/compilers_linux-64_1507259624353/work/.build/x86_64-conda_cos6-linux-gnu/build/build-cc-gcc-final/x86_64-conda_cos6-linux-gnu/libstdc++-v3/include/bits/basic_string.h:3126
#2  0x00007ffff731189d in std::basic_string<char, std::char_traits<char>, std::allocator<char> >::basic_string (this=0x7fffffffcf68, __str=...)
    at /home/msarahan/miniconda2/conda-bld/compilers_linux-64_1507259624353/work/.build/x86_64-conda_cos6-linux-gnu/build/build-cc-gcc-final/x86_64-conda_cos6-linux-gnu/libstdc++-v3/include/bits/basic_string.tcc:613
#3  0x00000000005f63fd in boost::filesystem::path::path (this=0x7fffffffcf68, p=...)
    at /home/antoine/miniconda3/envs/pyarrow/include/boost/filesystem/path.hpp:137
#4  0x00000000005f628a in boost::filesystem::operator/ (lhs=..., rhs=...)
    at /home/antoine/miniconda3/envs/pyarrow/include/boost/filesystem/path.hpp:792
#5  0x00000000005f1d37 in arrow::ipc::temp_path () at ../src/arrow/ipc/json-integration-test.cc:233
#6  0x00000000005f3038 in arrow::ipc::TestJSONIntegration::mkstemp (this=<error reading variable: Cannot access memory at address 0x7f00706d7427>)
    at ../src/arrow/ipc/json-integration-test.cc:241
Backtrace stopped: previous frame inner to this frame (corrupt stack?)

Reporter: Antoine Pitrou / @pitrou

_{Note: This issue was originally created as ARROW-3092. Please see the migration documentation for further details.}

[asfimport]

Wes McKinney / @wesm:
I think this is a compiler toolchain issue. If you build without an external toolchain and using the system boost libraries, does it still fail?

[asfimport]

Antoine Pitrou / @pitrou:
It seems that my lastest local changes (perhaps -D_GLIBCXX_USE_CXX11_ABI=1) had the side effect of fixing this issue, though I don't understand why.

[asfimport]

Wes McKinney / @wesm:
What's the last time you updated your toolchain? I think Anaconda has moved to the gcc5 ABI recently

[asfimport]

Antoine Pitrou / @pitrou:
Ah, well, I recreated my conda environment today in an attempt to fix other issues :) That might explain why.

[asfimport]

Wes McKinney / @wesm:
Our build instuctions in http://arrow.apache.org/docs/python/development.html are likely to have grown stale given the toolchain changes in recent months

I opened ARROW-3096 about having a look at that.

[asfimport]

Antoine Pitrou / @pitrou:
Can't reproduce anymore after I have recreated my environment and changed some build variables.