GCC 7.3 g++.dg/tls/thread_local-order2.C fails on Centos 6

[Jan Boehme]

On 02/15/2018 08:00 PM, Jan Boehme wrote:
>   Hi,
> 
> I'm working on building GCC 7.3 on Centos6.8.
> 
> I'm running in the same issue as described in GCC bug ticket 79427:
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=79427
> 
> FAIL: g++.dg/tls/thread_local-order2.C  -std=c++11 execution test
> FAIL: g++.dg/tls/thread_local-order2.C  -std=c++14 execution test
> 
> The executable runs into abort() caused by a wrong destruction order
> tracked by counter variables due to a possible malfunction of thread_local
> 
> My libc is glibc-2.12-1.192.el6.x86_64
> binutils are binutils-2.20.51.0.2-5.44.el6.x86_64
> 
> My configure flags are:
> 
> --disable-libstdcxx-pch \
> --disable-multilib \
> --enable-shared \
> --enable-static \
> --enable-__cxa_atexit \
> --enable-tls \
> --enable-clocale=gnu \
> --enable-threads=posix \
> --enable-languages=c++ \
> --enable-ld=yes \
> --enable-gold=yes \
> CC=/usr/bin/gcc \
> CXX=/usr/bin/g++ \
> --with-build-config=bootstrap-O3
> 
> I appreciate any hint to overcome this issue.
> 

Now I found out:
#1 the same behavior can be seen when bootstrapping on Centos7 using 
these settings.
#2 the same behavior can be seen when using the C7 standard compiler:

g++ --version
g++ (GCC) 4.8.5 20150623 (Red Hat 4.8.5-16)

g++ /tmp/gcc-7.3.0/gcc/testsuite/g++.dg/tls/thread_local-order2.C -o 
/tmp/thread_local-order2.exe -std=c++11

/tmp/thread_local-order2.exe
Aborted (core dumped)

/lib64/libc.so.6
GNU C Library (GNU libc) stable release version 2.17

It seems to me that this regression? is unknown or accepted over there. 
Any encouragement to file a bug for Centos/Redhat?

Regards, Jan