is_member_func_test fails for GCC 4.8.1

[joaquintides]

is_member_func_test fails for GCC 4.8.1 as shown in

https://www.boost.org/development/tests/master/developer/output/igaztanaga-master-gcc-4-8c++11-boost-bin-v2-libs-type_traits-test-is_member_func_test-test-gcc-4-8c+-dbg-dbg-symbl-off-vsblt-hdn.html

The problem here is that is_member_function_pointer.hpp dispatches to a C++11 implementation or a C++03 one depending on whether BOOST_TT_HAS_ASCCURATE_IS_FUNCTION is defined or not:

type_traits/include/boost/type_traits/detail/config.hpp

Lines 84 to 87 in 17af798

	#if !defined(BOOST_NO_CXX11_VARIADIC_TEMPLATES) && !BOOST_WORKAROUND(BOOST_GCC, <= 40800)\
	&& !BOOST_WORKAROUND(BOOST_MSVC, < 1900) && !BOOST_WORKAROUND(__clang_major__, <= 4)
	# define BOOST_TT_HAS_ASCCURATE_IS_FUNCTION
	#endif

And BOOST_GCC is 40801 for the offending compiler. I guess the fix is as simple as changing BOOST_WORKAROUND(BOOST_GCC, <= 40800) to BOOST_WORKAROUND(BOOST_GCC, < 40900).

[Kojoley]

Attention, this is on the master branch and 1.69 release will be closed in less than two weeks.

Also, there is a strange part in the fail: other GCC 4.8 runner (teeks99-02-mg4.8-Docker-64on64 gcc-4.8~c++11) does not fail; Travis CI has 4.8 job and it did not fail too. It looks like there is some thing in MinGW GCC.

[jzmaddock]

Yeah, it's strange that it passes CI. The intent was to exclude gcc-4.8 from the C++11 branch so the fix is the right one in any case.

[jzmaddock]

The travis job is gcc-4.8.5 - so it's possible the fix is no longer required at that patch level I guess.

[joaquintides]

I'm afraid the commit has not been merged to master, and Boost is shipping today.