Re: [boost] error on inclusion of ../boost/math/special_functions/next.hpp

On Nov 3, 2023, at 8:55 AM, Alexander Grund via Boost wrote:

...

...

...

Test output: teeks99-dkr-dc6-17 - serialization - test_complex_text_archive / clang-linux-6.0~c++17 Rev 1c12b5c15fc82f0cb6d4bd931a057fc050df8a34 / Sun, 29 Oct 2023 21:39:02 +0000 Compile [2023-10-29 23:34:25 UTC]: fail

"clang++-6.0" -fvisibility-inlines-hidden -Wno-c99-extensions -std=c++17 -fPIC -m64 -pthread -O0 -fno-inline -Wall -fvisibility=hidden --target=x86_64-pc-linux -ftemplate-depth-255 -DBOOST_ALL_NO_LIB=1 -DBOOST_ARCHIVE_TEST=text_archive.hpp -DBOOST_ATOMIC_DYN_LINK=1 -DBOOST_BEAST_ALLOW_DEPRECATED -DBOOST_BEAST_SEPARATE_COMPILATION -DBOOST_COBALT_USE_STD_PMR=1 -DBOOST_FILESYSTEM_DYN_LINK=1 -DBOOST_LIB_DIAGNOSTIC=1 -DBOOST_SERIALIZATION_DYN_LINK=1 -DBOOST_WSERIALIZATION_DYN_LINK=1 -I".." -c -o "/var/boost/run/results/boost/bin.v2/libs/serialization/test/test_complex_text_archive.test/clang-linux-6.0~c++17/debug/debug-symbols-off/threading-multi/visibility-hidden/test_complex.o" "../libs/serialization/test/test_complex.cpp"

In file included from ../libs/serialization/test/test_complex.cpp:19: In file included from ../boost/math/special_functions/next.hpp:17: In file included from ../boost/math/special_functions/trunc.hpp:21: In file included from ../boost/math/ccmath/ldexp.hpp:15: In file included from ../boost/math/ccmath/abs.hpp:16: ../boost/math/ccmath/isnan.hpp:35:29: error: no template named 'is_integral_v' in namespace 'std'; did you mean 'is_integral'? if constexpr (!std::is_integral_v<T>) ~~~~~^~~~~~~~~~~~~ is_integral /usr/bin/../lib/gcc/x86_64-linux-gnu/5.4.0/../../../../include/c++/5.4.0/type_traits:289:12: note: 'is_integral' declared here struct is_integral ^

Looks like using clang-6 with libstdc++ 5. Do we really support such configurations (or this in particular)? Although the fix is easy (replace the `*_v` inline variables by `::value` variants) it is additional effort.

Alex _______________________________________________ Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost

Even if we replaced _v with ::value there would be more sinister errors since this function relies on having __builtin_is_constant_evaluated support (which to a sufficient degree was added in GCC-6 as __builtin_constant_p). The drone image has clang-6 with libstdc++6 so we missed it in our CI matrix. Matt