As I known, the undocument interface BOOST_PP_IS_EMPTY is not a general implementation to condider emptiness. There is a undefined behaviours when token concatenation return a invalid preprocessing token. Particularly, gcc restriction reports error. For example: #include <boost/preprocessor/ facilities/is_empty.hpp> BOOST_PP_IS_EMPTY() /* result: 1 */ BOOST_PP_IS_EMPTY(a) /* result: 0 */ BOOST_PP_IS_EMPTY(+a) /* error report here, for '+' cannot be pasted */ Here is a work around, it hasnot been tested enough, but does do under gcc. #define MY_IS_EMPTY(a) BOOST_PP_EQUAL(BOOST_PP_SEQ_ SIZE(MY_IS_EMPTY_II(MY_IS_EMPTY_I(a))), 2) #define MY_IS_EMPTY_I(a) (BOOST_PP_CAT(MY_IS_EMPTY_I_, BOOST_PP_SEQ_SIZE(a))) ) #define MY_IS_EMPTY_I_0 MY_IS_EMPTY_I_HELPER BOOST_PP_LPAREN() #define MY_IS_EMPTY_I_HELPER(a) )a( /* when a is empty, my may get more sequence elements of boost proprocessor */ #define MY_IS_EMPTY_II(a) BOOST_PP_CAT(MY_IS_EMPTY_II_, BOOST_PP_SEQ_SIZE(a)) ) #define MY_IS_EMPTY_II_1 ( #define MY_IS_EMPTY_II_2 ()( It uses a intermadiate result of BOOST_PP_SEQ_SIZE with wrong parameters, as: BOOST_PP_SEQ_SIZE(a) => 0 a ... I think it maybe better to allow any concatenation tokens in the language itselft. It's hard to avoid such thing when taking the preprocessor as a MPL. I just reported it on comp.std.c, https://groups.google.com/forum/#!topic/comp.std.c/7rgt8k6ILMM. There are more things and problems I did and I met. Hope for more response.