Compile errors with CLang7 & libc++

[abrownsword]

I updated to CLI v1.6.1 and started getting errors about the inclusion of <experimental/optional>, telling me should be included instead (this is because the <experimental/optional> header remains in place to generate an error that it shouldn't be used). Examining Optional.hpp it is apparent that when compiling for C++17, both CLI11_STD_OPTIONAL and CLI11_EXPERIMENTAL_OPTIONAL are defined and therefore both and <experimental/optional> are included. Adding a !defined(CLI11_STD_OPTIONAL) to the #if around #define CLI11_EXPERIMENTAL_OPTIONAL fixes this. A similar thing should be done with the boost optional.

[henryiii]

You are correct, fixing in coming PR. Are you sure it makes sense to drop boost optional if optional and boost are both available? They shouldn't clash, and a user may still want to use Boost.

[abrownsword]

I have no opinion on the boost support -- I don't use it and don't have it installed. If you're sure they don't clash then leaving it enabled ought to be fine.

[henryiii]

It will only activate if you have Boost. You can always disable it by defining CLI11_BOOST_OPTIONAL as 0. You can also do this before CLI11 1.6.2 comes out with CLI11_EXPERIMENTAL_OPTIONAL, if defined to 0 it will not try to activate even if present.

[henryiii]

Please verify this is now fixed in master through #159.

[nyanpasu64]

I'm still getting an error. I think #159 is wrong, it should check !defined(CLI11_STD_OPTIONAL) || CLI11_STD_OPTIONAL == 0.

CLI11/include/CLI/Optional.hpp

Line 22 in 4389b33

&& (!defined(CLI11_STD_OPTIONAL) || CLI11_STD_OPTIONAL != 0)

Right now, it includes experimental only if non-experimental already exists!

---

Changing to CLI11_STD_OPTIONAL == 0 fixes build for me.

[henryiii]

Check and see if that is better, please! Thanks.

[nyanpasu64]

i can't exactly check, considering I applied the patch directly on the release CLI11.hpp. But this is the same as the working code I have locally (Android Studio NDK r18 Beta 2).

[henryiii]

Okay, I'll assume for a 1 char fix that's safe enough. :)

[abrownsword]

Apologies, I haven't had time to sync and test this yet.

[abrownsword]

I have built from HEAD using Clang 7.0.0 on Linux and it worked perfectly. Thanks.