Warning: implicit conversion loses integer precision

[jfirebaugh]

When compiling with -Wshorten-64-to-32 on platforms where std::size_t is unsigned long, many warnings of the following form are produced:

include/mpark/variant.hpp:754:24: warning: implicit conversion loses integer precision: 'std::size_t' (aka 'unsigned long') to 'unsigned int' [-Wshorten-64-to-32]
          lhs.index_ = rhs.index();
                     ~ ^~~~~~~~~~~

[viboes]

Could you give more context so that we know what lhs and rhs are?

[jfirebaugh]

Sure -- as far as I can tell (not super familiar with the implementation), this warning will occur regardless of the alternative types of the actual variant in use, because the lhs and rhs involved here are internal to the implementation. The error location is here, so:

• lhs is mpark::detail::constructor<Traits> and lhs.index_ is unsigned int index_ from base
• the type of rhs is a templated type, but I'm guessing it's also some instantation of base, meaning rhs.index() is base::index(), returning std::size_t.

[mpark]

Thanks for reporting this @jfirebaugh! I've fixed this in https://github.com/mpark/variant/commit/d8fd0fbba727f0f2f757a05ae7a86e14c3b5982c