boost::mpl::sequence_tag conflicts with mp11's one about fusion

[pdimov]

In boostorg/mp11@164aa37, I've split mpl.hpp into mpl_list.hpp and mpl_tuple.hpp, so in case of a conflict with Fusion libraries can now include just mpl_list.hpp and leave std::tuple to Fusion.

[jonesmz]

I'm running into this issue as well.

Any advice from boost::fusion?

[djowel]

Not sure how to deal with this, yet. Have anyone asked the mpl11 devs ( pings @pdimov )? Fusion's usage looks legit. Maybe we need to include the right header instead of a forward declaration.

[pdimov]

The forward declarations are correct; the problem is that both libraries try to add MPL support to std::tuple.

The only option I see is for me to split mp11/mpl.hpp into two parts, mpl_list.hpp, which will add MPL support for mp_list, and mpl_tuple.hpp, which will contain the std::tuple support. This will leave to the user the choice of whether to give std::tuple to Fusion or Mp11.

[jonesmz]

Forgive me for not knowing the internal details of either library, but it is strange to me that boost::fusion is defining something in the boost::mpl namespace at all.

Can boost::fusion be adjusted to use a different way of doing it?

[pdimov]

This is a documented MPL customization mechanism. What Fusion is doing is fine. The problem is that nobody really owns std::tuple, as that's a standard type, not part of any library, Boost or otherwise.

[djowel]

The only option I see is for me to split mp11/mpl.hpp into two parts, mpl_list.hpp, which will add MPL support for mp_list, and mpl_tuple.hpp, which will contain the std::tuple support. This will leave to the user the choice of whether to give std::tuple to Fusion or Mp11.

Perhaps fusion can also introduce a PP macro that allows one to inhibit the std::tuple specializations, as a workaround. or, hmmm, maybe just inhibit it all together when mp11 is detected (probably needs to test that tests are still passing as expected, or find ways to make it work with mpl11 right out of the box).

[djowel]

In boostorg/mp11@164aa37, I've split mpl.hpp into mpl_list.hpp and mpl_tuple.hpp, so in case of a conflict with Fusion libraries can now include just mpl_list.hpp and leave std::tuple to Fusion.

Wonderful! Thanks @pdimov