V2 and clang16 compatibility #462
-
|
I am trying out the new version of the library and it compiles successfully with gcc13 but not with clang16 & libc++16. There are compile errors like: magnitude.h:504:44: error: sorry, non-type template argument of type 'long double' is not yet supported
quantity_spec.h:641:57: error: template template argument has different template parameters than its corresponding template template parameter
quantity_spec.h:748:10: error: no viable constructor or deduction guide for deduction of template arguments of 'explode_result'
#...What is the plan in terms of clang? wait for llvm to implement or make workarounds? btw, brilliant library. |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 8 replies
-
|
I am happy that you like the library! Yes, some compilers are not yet up to speed with C++20 implementation. When I am done with V2 (I still need to clean up unit tests), release 0.8.0 from mainline, and put V2 on master branch for gcc-12+ I will try to port it to other compilers as well (clang, MSVC, and older versions of gcc). We also need a total rewrite of docs for V2. That's a lot of work so any help is appreciated. |
Beta Was this translation helpful? Give feedback.
-
|
Yes, something like that + an overload for |
Beta Was this translation helpful? Give feedback.
-
|
nice I will try to make it compile with clang |
Beta Was this translation helpful? Give feedback.
-
|
There is one other thing that needs some work lambdas in /usr/bin/cmake -E env CCACHE_DIRECT=1 CCACHE_NO_DEPEND=1 CCACHE_BASEDIR=/home/jonb/Projects/units /usr/bin/ccache /opt/clang-16.0.0/bin/clang++ -DFMT_SHARED -DUNITS_USE_LIBFMT=1 -Dgsl_CONFIG_CONTRACT_CHECKING_AUDIT -I/home/jonb/Projects/units/. -I/home/jonb/Projects/units/src/core/include -I/home/jonb/Projects/units/src/core-io/include -I/home/jonb/Projects/units/src/core-fmt/include -I/home/jonb/Projects/units/src/systems/angular/include -I/home/jonb/Projects/units/src/systems/isq/include -I/home/jonb/Projects/units/src/systems/iec80000/include -I/home/jonb/Projects/units/src/systems/si/include -I/home/jonb/Projects/units/src/systems/isq_angle/include -I/home/jonb/Projects/units/src/systems/natural/include -I/home/jonb/Projects/units/src/systems/cgs/include -I/home/jonb/Projects/units/src/systems/hep/include -I/home/jonb/Projects/units/src/systems/iau/include -I/home/jonb/Projects/units/src/systems/imperial/include -I/home/jonb/Projects/units/src/systems/international/include -I/home/jonb/Projects/units/src/systems/typographic/include -I/home/jonb/Projects/units/src/systems/usc/include -I/home/jonb/Projects/units/src/utility/include -stdlib=libc++ -I/opt/clang-16.0.0/include/ -I/opt/clang-16.0.0/include/x86_64-unknown-linux-gnu/c++/v1 -g -std=gnu++2b -fcolor-diagnostics -Wall -Wextra -Wpedantic -Wshadow -Wnon-virtual-dtor -Wold-style-cast -Wcast-align -Wunused -Woverloaded-virtual -Wcast-qual -Wconversion -Wsign-conversion -Wnull-dereference -Wformat=2 -Wdangling-else -Werror -fconstexpr-backtrace-limit=0 -MD -MT CMakeFiles/test_headers.dir/headers/src/systems/hep/include/mp-units/systems/hep/hep.cpp.o -MF CMakeFiles/test_headers.dir/headers/src/systems/hep/include/mp-units/systems/hep/hep.cpp.o.d -o CMakeFiles/test_headers.dir/headers/src/systems/hep/include/mp-units/systems/hep/hep.cpp.o -c /home/jonb/Projects/units/cmake-build-debug/headers/src/systems/hep/include/mp-units/systems/hep/hep.cpp
In file included from headers/src/systems/hep/include/mp-units/systems/hep/hep.cpp:1:
In file included from ../src/systems/hep/include/mp-units/systems/hep/hep.h:25:
In file included from ../src/systems/si/include/mp-units/systems/si/si.h:25:
In file included from ../src/systems/si/include/mp-units/systems/si/constants.h:25:
In file included from ../src/core/include/mp-units/quantity.h:27:
In file included from ../src/core/include/mp-units/bits/quantity_concepts.h:26:
In file included from ../src/core/include/mp-units/bits/reference_concepts.h:26:
In file included from ../src/core/include/mp-units/bits/unit_concepts.h:26:
../src/core/include/mp-units/bits/magnitude.h:755:34: error: constexpr variable 'first_base' must be initialized by a constant expression
static constexpr std::intmax_t first_base = get_or_compute_first_factor();
^ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../src/core/include/mp-units/bits/magnitude.h:760:28: note: in instantiation of static data member 'mp_units::detail::prime_factorization<334524384739>::first_base' requested here
magnitude<power_v_or_T<first_base, ratio{first_power}>()>{} * prime_factorization<remainder>::value;
^
../src/core/include/mp-units/bits/magnitude.h:770:47: note: in instantiation of template class 'mp_units::detail::prime_factorization<334524384739>' requested here
inline constexpr auto prime_factorization_v = prime_factorization<N>::value;
^
../src/core/include/mp-units/bits/magnitude.h:781:47: note: in instantiation of variable template specialization 'mp_units::detail::prime_factorization_v' requested here
inline constexpr Magnitude auto mag = detail::prime_factorization_v<R.num> / detail::prime_factorization_v<R.den>;
^
../src/systems/hep/include/mp-units/systems/hep/hep.h:43:57: note: in instantiation of variable template specialization 'mp_units::mag' requested here
inline constexpr struct proton_mass : named_unit<"m_p", mag<ratio{1'672'621'923'695, 1'000'000'000'000}> * mag_power<10, -27> * si::kilogram> {} proton_mass;
^
/opt/clang-16.0.0/bin/../include/c++/v1/optional:296:30: note: constexpr evaluation hit maximum step limit; possible infinite loop?
__engaged_(false) {}
^
/opt/clang-16.0.0/bin/../include/c++/v1/optional:322:8: note: in call to '__optional_destruct_base()'
struct __optional_storage_base : __optional_destruct_base<_Tp>
^
/opt/clang-16.0.0/bin/../include/c++/v1/optional:489:8: note: in call to '__optional_storage_base()'
struct __optional_copy_base : __optional_storage_base<_Tp>
^
/opt/clang-16.0.0/bin/../include/c++/v1/optional:517:8: note: in call to '__optional_copy_base()'
struct __optional_move_base : __optional_copy_base<_Tp>
^
/opt/clang-16.0.0/bin/../include/c++/v1/optional:550:8: note: in call to '__optional_move_base()'
struct __optional_copy_assign_base : __optional_move_base<_Tp>
^
/opt/clang-16.0.0/bin/../include/c++/v1/optional:582:8: note: in call to '__optional_copy_assign_base()'
struct __optional_move_assign_base : __optional_copy_assign_base<_Tp>
^
/opt/clang-16.0.0/bin/../include/c++/v1/optional:738:41: note: in call to '__optional_move_assign_base()'
_LIBCPP_INLINE_VISIBILITY constexpr optional(nullopt_t) noexcept {}
^
../src/core/include/mp-units/bits/prime.h:56:10: note: in call to 'optional({})'
return std::nullopt;
^
../src/core/include/mp-units/bits/prime.h:148:80: note: in call to 'first_factor_maybe(334524384739, 233731)'
detail::get_first_of(coprimes_in_first_wheel, [&](auto p) { return first_factor_maybe(n, wheel + p); }))
^
../src/core/include/mp-units/bits/algorithm.h:52:20: note: in call to '&f->operator()(1)'
if (auto opt = f(*first)) return *opt;
^
../src/core/include/mp-units/bits/algorithm.h:60:10: note: in call to 'get_first_of(&coprimes_in_first_wheel.__elems_[0], &coprimes_in_first_wheel.__elems_[48], {n, wheel})'
return get_first_of(begin(rng), end(rng), f);
^
../src/core/include/mp-units/bits/prime.h:148:13: note: in call to 'get_first_of(coprimes_in_first_wheel, {n, wheel})'
detail::get_first_of(coprimes_in_first_wheel, [&](auto p) { return first_factor_maybe(n, wheel + p); }))
^
../src/core/include/mp-units/bits/magnitude.h:751:41: note: in call to 'find_first_factor(334524384739)'
return static_cast<std::intmax_t>(factorizer::find_first_factor(N));
^
../src/core/include/mp-units/bits/magnitude.h:755:47: note: in call to 'get_or_compute_first_factor()'
static constexpr std::intmax_t first_base = get_or_compute_first_factor();
^I think I will go for the old version 0.7 |
Beta Was this translation helpful? Give feedback.
-
|
I'm looking into this right now. |
Beta Was this translation helpful? Give feedback.
-
|
I added some changes an hour ago already. Make sure you are updated. |
Beta Was this translation helpful? Give feedback.
I am happy that you like the library!
Yes, some compilers are not yet up to speed with C++20 implementation. When I am done with V2 (I still need to clean up unit tests), release 0.8.0 from mainline, and put V2 on master branch for gcc-12+ I will try to port it to other compilers as well (clang, MSVC, and older versions of gcc). We also need a total rewrite of docs for V2. That's a lot of work so any help is appreciated.