Problem with GCC 8: Example does not compile b/c `poly::unerase_poly` is not declared. about dyno HOT 3 CLOSED

Update: Funny thing is, clang-5.0 and gcc-7 both compile it (though clang warns about string literal operators being a GNU extension):

lukas@i11pcbarth ~/tmp/mnwe/dyno $ /usr/bin/clang-5.0  --std=c++17 -I /home/lukas/src/dyno/include/ -I /home/lukas/opt/boost_install/include/ ./example.hpp 
In file included from ./example.hpp:1:
In file included from /home/lukas/src/dyno/include/dyno.hpp:8:
In file included from /home/lukas/src/dyno/include/dyno/builtin.hpp:8:
In file included from /home/lukas/src/dyno/include/dyno/concept.hpp:8:
/home/lukas/src/dyno/include/dyno/detail/dsl.hpp:124:18: warning: string literal operator templates are a GNU extension [-Wgnu-string-literal-operator-template]
  constexpr auto operator""_s() { return detail::string<c...>{}; }
                 ^
1 warning generated.

lukas@i11pcbarth ~/tmp/mnwe/dyno $ /usr/bin/g++-7  --std=c++17 -I /home/lukas/src/dyno/include/ -I /home/lukas/opt/boost_install/include/ ./example.hpp 
lukas@i11pcbarth ~/tmp/mnwe/dyno $ 

Thus, this might be a GCC-8 bug / regression? My feeling is that something in the enable_ifs for poly::unerase_poly is fishy, but I really don't understand what's going on there.

from dyno.

I've run in in to the same problem with my travis build and gcc-8.0
It seems that gcc-8.2 compiles the sources fine.

from dyno.

Yes, this seems like a GCC bug. I can reproduce with GCC 8.1. I was not able to find a workaround that did not involve making unerase_poly a non-static member function, which is not satisfactory since that requires capturing this in the function returned by virtual_impl even when this would not otherwise need to be captured. I'll close this as "wontfix" but I'm open to workarounds if somebody has one.

from dyno.