Further test cases and fixes for #11211

[daminetreg]

Relates to : https://svn.boost.org/trac/boost/ticket/11211

This removes the BOOST_CONSTEXPR macro from constructors generated by DEFINE_STRUCT_INLINE as they were generating the same error.

I also added some test cases for this. It runs under

• clang v3.5.1 with std=c++11 & std=c++14
• g++ v4.9.2 with std=c++11 & std=c++14

While I was doing the changes, I noticed that we may get more errors with BOOST_CONSTEXPR, as this was added in alot of place, and I'm not sure if we won't get more surprises as those in #11211.

@Flast did you make an auto-replace where BOOST_FUSION_GPU_ENABLED appeared or did you put them individually ? If it was made with auto-replace I think we have to take a tour on each place where it appears and determine if it makes sense, as GPU_ENABLED is not the same use case I think.

[Flast]

@Flast did you make an auto-replace where BOOST_FUSION_GPU_ENABLED appeared or did you put them individually ?

I inserted the macros by hand; i.e. no auto generated.

GPU_ENABLED is not the same use case I think.

Yes, I think so.

[daminetreg]

Ok nice @Flast, this means we don't have to fear more problems. 👍

Thanks @djowel for merging. 😄

[djowel]

No need to thank me. I should be the one to thank you guys!

[Flast]

@daminetreg I got a error under C++98 mode with trailing comma.

[daminetreg]

Oh yes you are right. It's in else case when no variadics support, then this way of writing doesn't work, actually it should never compile when BOOST_PP_VARIADICS is disabled. I tested with BOOST_PP_VARIADICS enabled all the time for this PR.

I'll make myself a script to avoid forgetting such combinations. I'll make a PR to fix it.