gcc 15 failure

[pgajdos]

Fails with gcc 15:

[   21s] + /usr/bin/cmake --build build -j8 -t check
[...]
[   67s] /home/abuild/rpmbuild/BUILD/zug-0.1.1-build/zug-0.1.1/zug/sequence.hpp:75:33: error: ‘struct zug::detail::sequence_data<ValueT, XformT>’ has no member named ‘impl’; did you mean ‘impl_’? [-Wtemplate-body]
[   67s]    75 |         impl_ = std::move(other.impl);
[   67s]       |                                 ^~~~
[   67s]       |                                 impl_
[   67s] gmake[3]: *** [test/CMakeFiles/test-sequence.dir/build.make:82: test/CMakeFiles/test-sequence.dir/sequence.cpp.o] Error 1

[pgajdos]

Also reported here:
https://bugs.gentoo.org/937320

[TheCoconutChef]

Just want to note that I spotted another problem with the operator:

    sequence_data& operator=(sequence_data&& other)
    {
        impl_ = std::move(other.impl);
        impl_.reductor.current(this);
    }

It doesn't return *this.

The problem is that I'm not sure modifying impl for impl_ actually fixes the problem since impl_ = std::move(other.impl_) triggers a move assignment on lambda with captures in almost all (possibly literally all) cases.

You end up with stuff like:

zug::skip_state<zug::detail::sequence_data<int, decltype(enumerate)> *, zug::state_wrapper<zug::no_tag, zug::detail::sequence_data<int, decltype(enumerate)> *, unsigned long>>>') has a deleted copy assignment operator
    : reductor_base<
...
copy assignment operator of 'reductor_base ... is implicitly deleted because field 'step_' has a deleted copy assignment operator
    reducing_fn_type step_;
...
copy assignment operator of 'caller<(lambda at .../enumerate.hpp:26:16)>' is implicitly deleted because field 'step' has a deleted copy assignment operator
    ReducingFnT step;
                ^
.../enumerate.hpp:26:16: note: lambda expression begins here
        return [initial = initial, step = std::forward<Step>(step)](
               ^

You can trigger these error on less recent compilers by writing:

template struct detail::sequence_data<int, decltype(enumerate)>;

[arximboldi]

Ahhhh... lambdas and their deleted assignment operators, my ethernal PITA! Thanks @TheCoconutChef , clearly the right solution should be

    sequence_data& operator=(sequence_data&& other)
    {
        impl_ = std::move(other.impl_);
        impl_.reductor.current(this);
        return *this;
    }

But if that triggers other errors it's something that probably requires a deeper look. Sadly until in two weeks I won't have time for that.

(Rushing preparing a C++Con talk, are you by any chance @TheCoconutChef attending? I appreciate a lot your contributions to immer/lager/zug a log, I'd love to hang out :)

[TheCoconutChef]

I am:

1. Super stocked to learn you're giving a talk this year. I didn't know.
2. Unfortunately not attending cppcon.

But I will for sure watch your talk.