libstdc++.so: undefined reference to `backtrace' (FreeBSD,...)

[Gerald Pfeifer <gerald at pfeifer dot com>]

Since probably May 8th (if https://gcc.gnu.org/ml/gcc-testresults/2018-05/ 
has it right) the gcc-testresults@ list refuses the output of a nightly 
build I have been operating on i586-unknown-freebsd10.4.

Digging into this a bit, I noticed this is due to lots and lots of 
testcases failing:

                === libstdc++ Summary ===

    # of expected passes            3317
    # of unexpected failures        3732
    # of expected failures          5
    # of unresolved testcases       3692
    # of unsupported tests          775

And digging into this yet a bit further it appears to be due to 
undefined references to `backtrace' and `backtrace_symbols`:

    ~/gcc-ref10-i386/bin/g++ g++.dg/debug/trivial.C -gstabs -O2
    /home/gerald/gcc-ref10-i386/lib/gcc/i586-unknown-freebsd10.4/9.0.0/../../../libstdc++.so: undefined reference to `backtrace_symbols'
    /home/gerald/gcc-ref10-i386/lib/gcc/i586-unknown-freebsd10.4/9.0.0/../../../libstdc++.so: undefined reference to `backtrace'
    collect2: error: ld returned 1 exit status

Digging into it yet a bit further this appears to be caused by

     2018-05-08  François Dumont  <fdumont@gcc.gnu.org>
 
	* src/c++11/debug.cc [_GLIBCXX_HAVE_EXECINFO_H]: Include execinfo.h.
	[_GLIBCXX_HAVE_EXECINFO_H](_Error_formatter::_M_error): Render
	backtrace.

Indeed on the FreeBSD 10.4 system hosting the tester 
/usr/include/execinfo.h exists and has the following:

    size_t backtrace(void **, size_t);
    char **backtrace_symbols(void *const *, size_t);

And looking at the respective man page I see

    NAME
      backtrace — fill in the backtrace of the currently executing thread

    LIBRARY
      Backtrace Access Library (libexecinfo, -lexecinfo)

This (linking in the library) appears to be the missing piece?
Should this be linked in transparently on FreeBSD platforms?

Gerald